Diving into Data (Web Analytics) looking for insights.
Advice of Avinash:
Segmentation:
New Nirvana Rule: Never report a metric (even God’s favorite KPI) without segmenting it to give deep insights into what that metric is really hiding behind it.
The power of segmenting a metric is that you peek behind the curtain and find out more about the metric. These are the benefits that you will gain:
1. It is impossible to segment any metric without putting in the effort to understand what we are reporting and the business value that the metric represents. This is hard work but what does not kill you makes you stronger. :~)
2. Segmenting allows you to quickly hone in on areas of deeper dive from which will emerge key insights that will drive real and meaningful action.
3. Our senior executives and decision makers don’t understand all the complexity and magic that is a web experience, showing them segmented trends is a extremely effective communication tool (and the best part is you barely have to talk, the picture will tell the story).
4. You will earn a big fat bonus and promotion.
KPI Attributes:
1) What is good for Jane it might not be good for John. Every web business is unique !!
2) Do not go for perfection: Good is enough !!
3) Critical few baby, critical few !! Few KPIs:Most common: 3. Maximum: 4 KPI for each website analysis.
4) KPI Lifecycle (Half of KPI will die in one year), because your competitor become stronger. So apply Lean Six Sigma Process (Define, Measure, Analyse, Action, Improve or Eliminate). If your KPI do not give you insight, kill it, be brutal and kill it, and then define a new better KPI .
Each KPI must be:
1) Uncomplex
2) Relevant
3) Timely
4) Instantly useful
Do Not forget of Conversion Trinity
1) Relevant
2) Value
3) Call to action
Showing posts with label segment. Show all posts
Showing posts with label segment. Show all posts
Monday, September 29, 2014
Thursday, July 24, 2014
Web Analytics Segmentation: Do Or Die, There Is No Try!
Occam's Razor
by Avinash Kaushik
My
love for segmentation as the primary (only?) way of identify actionable
insights is on display in pretty much every single blog post I write.
I have said: All data in aggregate is "crap".
Because it is.
One of my earliest blog posts extolled the glorious virtues of segmentation:
Excellent Analytics Tip#2: Segment Absolutely Everything.
Many paid web analytics clickstream analytics tools, even today (!), don't allow you to do on the fly segmentation of all your data (not without asking you to change javascript script tags every time you need to segment something, or not without paying extra or paying for additional "data warehouse" solutions).
So it was with absolute delight that I wrote a detailed post about the release of Advanced Segmentation feature in Google Analytics in Oct 2008: Google Analytics Releases Advanced Segmentation: Now Be A Ninja!
Of course Yahoo! Web Analytics, the other wonderful free WA tool, had advanced segmentation from day one.
And as recently as two weeks ago I stressed the importance of effective segmentation as the cornerstone of the Web Analytics Measurement Framework.
[Update: Please read this post first in its entirety. Internalize it. Then when you are ready to get a jump start on advanced segmentation, download three super cool segments directly into your account from this post: 3 Advanced Web Analytics Visitor Segments: Non-Flirts, Social, Long Tail]
The Problem.
You can imagine then how absolutely heartbreaking it is for me to note that nearly all reporting that I see is data in aggregate.
All visits. Total revenue. Avg page views per visitors. Time on site. Overall customer satisfaction. And more. Tons of data "puking", all just aggregates.
The achingly tiny percent of time that the Analyst does segmentation it seems to stop at New vs. Returning Visitors! I have to admit I see that and I feel like throwing a tomato against the wall.
Yes new visitors and returning visitors are segments. But they are so lame that I dare you to find any insight worth, well, a tomato based on those two. You can't. Because new and returning are still two big indefinable globs!
Even if your business actually is tied to understanding the first and then subsequent visits by a person then you are far better off segmenting using Visitor Loyalty (in GA count of visits).
But I am getting off track (this whole non-segmentation business drives me bananas!).
Deep breath.
The Unbearable Lightness of Being.
Segmenting your data is key to your success and that of your company.
It is not very difficult to segment your data. Many tools include some default segments you can apply to any report you are looking at.
For example when you look at your revenue or goal performance it takes a trivial amount of effort to look at All Visits but add to that report the Paid Search Traffic and Non-Paid Search Traffic and get deeper insights.
You can tell your boss: We made 900k, and while you are obsessed with Paid Search please note that 850k of the revenue came from Organic and only $25k from Paid.
PS: Our business is in trouble because we are over-reliant on Search!
See what I mean, a bit better insights.
Among things in the above image I love analyzing Direct (to understand value of the free traffic), Visits with Conversions (to understand my BFF sources and pages and behavior), and Non-bounce Visits (to understand people who give me a chance to do business with them).
But true glory will only come from going beyond the default segments.
Because default segments are created to appeal to everyone / the lowest common denominator, and we all know that there is no such thing as "everyone".
You are unique. The top three things your business is working on are unique. The multi-channel strategy you are executing is unique. Your investment in tools vs people in your company is unique (you are 90/10 instead of 10/90!). You are struggling with your own unique challenges.
You have to have a segmentation strategy that is unique to you. And if you don't then your employment with the company needs to be re-evaluated. (Sorry.)
So how do you go about identifying unique segments for your business or non-profit?
Ask a lot of questions. Tap into the tribal knowledge. Force your leaders (ok HiPPO's) to help you define Business Objectives, Goals and Targets. [Key elements of the Web Analytics Measurement Framework.]
Let me tell you that without the above there is no hope. The first two will tell you what is important and currently prioritized. The third will tell you where to focus you analytical horsepower (based on actuals vs targets).
If you have O, G & T then it is time to select the segments to focus on, the micro-groups of data you'll focus on.
The Segmentation Selector Framework.
My humble recommendation is that as a best practice you should pick at least a couple of segments in each of these three categories:
1. Acquisition. 2. Behavior. 3. Outcomes.
You'll choose to focus on the micro group that is of value to you, and just to you, in each category. You'll apply those segments to web analytics reports where you hope to find insights (and if you choose the right segments you will!).
Let us look at each category I am recommending.
Segment Category #1: Acquisition.
Acquisition also includes search engine optimization (SEO), because it is an activity on which you spend time and money.
Ask yourself this question: "Where is my company currently spending most amount of time and money acquiring traffic?"
Bam! There's the most important segment you will focus on.
Why? If you do your analysis right you can lower cost (by identifying and eliminating the losers!) and you can increase revenue (by identifying and investing where things are going well).
See the process I followed there?
Paid search. A specific group of keywords. Television campaigns.
Email campaigns to prospective customers in Florida, New Mexico, Arizona
and Utah. Coupon affiliates. "Social media campaigns" (context). Billboard ads on side on highways. Business cards handed out at trade shows.
All of the above are examples of acquisition strategies.
When you look at your web analytics data look at All Visits AND at least one of the above.
Two acquisition segments is normal.
If you make it three then choose one acquisition strategy that your company is experimenting with.
Say you have 1/10th of one person doing some tweeting or facebooking, :), then add that one segment to your top two. This will allow your management to look at what they are focused on and also one thing that sounds cool but they have no idea if it is actually worth it.
(Short term focus) Win – Win (Long term focus)
How To Apply Segments / Analyze Data.
The reports you'll apply your acquisition segment to will depend on the Key Performance Indicators you have chosen. But a typical set of metrics you'll evaluate will hopefully represent a spectrum of success, like for example. . .
The effort will be to try and understand if for our acquisition
segment (say all my brand keywords or for email campaigns to increase
sales of the most expensive products). . . .
This is not the end of your journey but it is a darn good start!
[If you have Web Analytics 2.0 pop the CD at the back into your computer. In dashboard examples look for Stratigent_Sample_Dashboard.xls, via my friend Bill Bruno at Stratigent. It has an excellent example of segmented acquisition display, you can immediately steal it for your company!]
Segment Category #2: Behavior.
Even people who sometimes have segment their web analytics data often forget to segment by online behavior.
Many, but not all, behavior segments fall into these two buckets: People who see x pages. People who do y things.
Here are some specific examples (all of which you can create in Yahoo! Web Analytics or Google Analytics in a few seconds without having to pay anything extra for vars and slots or having to update your javascript tag or having to buy an add-on, you can also apply them to all your data including all your historical data).
Visits with more than three page views. . .
This can be so valuable on content only websites (more page views
more impressions of irrelevant display ads!) or even on ecommerce
websites (more pages views the deeper you sink your hook into the
visitor, engagement baby!).
Where do these people come from? Do they buy a lot? A little? Do they write reviews? Did we acquire them or did they just show up? If they see so many pages what type of content are they interested in (politics? naked pictures? sports?)?
So on and so forth. Segmenting one behavior, understanding its value.
Similarly another could be focusing on people how add to cart and then abandon the site.
Or people who enter the site on the home page and their behavior. . .
Or all those who did not enter the site via the home page!
Or people who use the site's product comparison chart or car configurator or, my fav, internal site search. Vs. those that don't.
Or people whose Days to Purchase (/Transaction) are 5 vs for those for whom the Days to Purchase is 1. . .
Or, cuter, those whose last visit to our website was 100, or whatever, days ago. Why? And what do they want?
Or people who visited the site more than 9 times (!) during the current time period. . .
Where are these sweet delicious people coming from? (Note: To a blog
updated only twice a month!) What do they read? What do they buy? What
can we learn from them and do more of?
Those are the types of questions you'll answer from your behavioral segments.
The more you understand what people are doing on your site, the more likely it is that you'll stop the silliness on your site (kill content, redo navigation, make cross sells better, eliminate 80% of the ads, learn to live with 19 days to conversion, don't sell too hard, and so much more).
It is also likely (I want to say guaranteed) that you'll find the delta between what you want to have happen and what your customers want. You'll choose to make happier customers, who in turn, in the naughtiest way possible, will make you happy.
And it all stars with being able to identify and focus on the right behavior segments.
Pick at least two.
But I have to admit in this segment category I truly "play" with the data a lot because it is so hard to know what the right segments are, because visitor behavior is such a complicated thing (they are constantly trying to mess with us Analysts!).
It is only after experimentation (a lot) that I end up with something sweet.
Segment Category #3: Outcomes.
But first what is it?
Segments with outcomes are people or visits where you get a order (at an ecommerce website) or you get a lead (at Organizing for America).
Those two are obvious right?
Segment out people who delivered those two outcomes. Give them a warm hug and a kiss. Now go figure out what makes them unique when compared to everyone else who showed up at your website, all those other people who you worked so hard to impress but failed to.
Take the insights and do more of what works for this group.
Or segment out everyone whose order size is 50% more than the average order size. . .
These are your "whales", people who spend a lot of money with you. Don't you want to get to know them a lot better? : )
But there is more.
Remember macro AND micro conversions!
No one is going to sleep with you on the first date. (Ok maybe a few will!)
So focus on micro conversions that lead up to a macro conversion… like people playing a product video (or on content site watching five videos!). . .
Or adding a product to their Wish List.
Or signing up to show up for a protest for your ultra liberal policies!
Or apply for a trial, or download a trial product.
You can also focus on micro conversions that all by themselves are of value to you, even if not as much as the macro conversion.
For example submitting a job application.
Or signing up for a RSS feed.
Or clicking on a link to go to a different site you want them to go to (like clicking on the amazon link to go buy my book – great outcome :)).
Of course if you are really really good you'll also segment my absolute favorite metric in the whole wide world: Task Completion Rate. It is the ultimate measure of outcome (from your customer's perspective).
Net, net. . . it is absolutely critical that you segment your data by the key outcomes important to your business. Not just because your site exists to add economic value, but also because I cannot think of another way you can earn the love of your boss or get promoted.
By understanding what it is about people who deliver outcomes you can understand what to do with all those that don't convert.
Outcomes. Outcomes. Outcomes!
Pick at least two.
If you pick three or four that is ok.
If you pick nine it might be a signal you don't know what you are doing (and you want to corner your boss in a non-HR-violation manner and ask her to help you focus on the most important).
In Summary.
Segment or die.
It is as simple as that.
The next time you start to do true analysis of your data I hope you have your minimum six segments in hand (two for each category). If you do you'll find that web analytics, this world full of web metrics and what not, suddenly becomes a lot more interesting (and you no longer feel like jumping out of your office window in frustration!).
Love, money and glory await you.
Not to mention how proud I'll be of you when I see your analysis. ; )
Ok now your turn.
Are you a segmentation God? What are some of your favorite segments? Have you used this three category framework in the past to find segments? Do you think they'll work in real life? In the context of segments what do you think is missing from this blog post? What did I overlook / not stress enough?
What's your excuse for not leveraging segmentation? (Best answer to this question win's a copy of Web Analytics 2.0!)
Please share your thoughts / wisdom / critique / guidance.
Thanks.
PS:
Couple other related posts you might find interesting:
by Avinash Kaushik
Web Analytics Segmentation: Do Or Die, There Is No Try!
My
love for segmentation as the primary (only?) way of identify actionable
insights is on display in pretty much every single blog post I write.I have said: All data in aggregate is "crap".
Because it is.
One of my earliest blog posts extolled the glorious virtues of segmentation:
Excellent Analytics Tip#2: Segment Absolutely Everything.
Many paid web analytics clickstream analytics tools, even today (!), don't allow you to do on the fly segmentation of all your data (not without asking you to change javascript script tags every time you need to segment something, or not without paying extra or paying for additional "data warehouse" solutions).
So it was with absolute delight that I wrote a detailed post about the release of Advanced Segmentation feature in Google Analytics in Oct 2008: Google Analytics Releases Advanced Segmentation: Now Be A Ninja!
Of course Yahoo! Web Analytics, the other wonderful free WA tool, had advanced segmentation from day one.
And as recently as two weeks ago I stressed the importance of effective segmentation as the cornerstone of the Web Analytics Measurement Framework.
[Update: Please read this post first in its entirety. Internalize it. Then when you are ready to get a jump start on advanced segmentation, download three super cool segments directly into your account from this post: 3 Advanced Web Analytics Visitor Segments: Non-Flirts, Social, Long Tail]
The Problem.
You can imagine then how absolutely heartbreaking it is for me to note that nearly all reporting that I see is data in aggregate.
All visits. Total revenue. Avg page views per visitors. Time on site. Overall customer satisfaction. And more. Tons of data "puking", all just aggregates.
The achingly tiny percent of time that the Analyst does segmentation it seems to stop at New vs. Returning Visitors! I have to admit I see that and I feel like throwing a tomato against the wall.
Yes new visitors and returning visitors are segments. But they are so lame that I dare you to find any insight worth, well, a tomato based on those two. You can't. Because new and returning are still two big indefinable globs!
Even if your business actually is tied to understanding the first and then subsequent visits by a person then you are far better off segmenting using Visitor Loyalty (in GA count of visits).
But I am getting off track (this whole non-segmentation business drives me bananas!).
Deep breath.
The Unbearable Lightness of Being.
Segmenting your data is key to your success and that of your company.
It is not very difficult to segment your data. Many tools include some default segments you can apply to any report you are looking at.
For example when you look at your revenue or goal performance it takes a trivial amount of effort to look at All Visits but add to that report the Paid Search Traffic and Non-Paid Search Traffic and get deeper insights.
You can tell your boss: We made 900k, and while you are obsessed with Paid Search please note that 850k of the revenue came from Organic and only $25k from Paid.
PS: Our business is in trouble because we are over-reliant on Search!
See what I mean, a bit better insights.
Among things in the above image I love analyzing Direct (to understand value of the free traffic), Visits with Conversions (to understand my BFF sources and pages and behavior), and Non-bounce Visits (to understand people who give me a chance to do business with them).
But true glory will only come from going beyond the default segments.
Because default segments are created to appeal to everyone / the lowest common denominator, and we all know that there is no such thing as "everyone".
You are unique. The top three things your business is working on are unique. The multi-channel strategy you are executing is unique. Your investment in tools vs people in your company is unique (you are 90/10 instead of 10/90!). You are struggling with your own unique challenges.
You have to have a segmentation strategy that is unique to you. And if you don't then your employment with the company needs to be re-evaluated. (Sorry.)
So how do you go about identifying unique segments for your business or non-profit?
Ask a lot of questions. Tap into the tribal knowledge. Force your leaders (ok HiPPO's) to help you define Business Objectives, Goals and Targets. [Key elements of the Web Analytics Measurement Framework.]
Let me tell you that without the above there is no hope. The first two will tell you what is important and currently prioritized. The third will tell you where to focus you analytical horsepower (based on actuals vs targets).
If you have O, G & T then it is time to select the segments to focus on, the micro-groups of data you'll focus on.
The Segmentation Selector Framework.
My humble recommendation is that as a best practice you should pick at least a couple of segments in each of these three categories:
1. Acquisition. 2. Behavior. 3. Outcomes.
You'll choose to focus on the micro group that is of value to you, and just to you, in each category. You'll apply those segments to web analytics reports where you hope to find insights (and if you choose the right segments you will!).
Let us look at each category I am recommending.
Segment Category #1: Acquisition.
Acquisition refers to the activity you undertake to attract people (or robots!) to your website.This would include campaigns you run, like pay per click marketing (PPC), email, affiliate deals, display / banner ads, facebook marketing campaigns.
Acquisition also includes search engine optimization (SEO), because it is an activity on which you spend time and money.
Ask yourself this question: "Where is my company currently spending most amount of time and money acquiring traffic?"
Bam! There's the most important segment you will focus on.
Why? If you do your analysis right you can lower cost (by identifying and eliminating the losers!) and you can increase revenue (by identifying and investing where things are going well).
See the process I followed there?
- Ask the question to identify what's important / high priority for the business.
- Create a segment (and then micro segments) for that one thing.
- Apply on the relevant reports to measure performance using key performance indicators.
- Take action. It will have an impact!
All of the above are examples of acquisition strategies.
When you look at your web analytics data look at All Visits AND at least one of the above.
Two acquisition segments is normal.
If you make it three then choose one acquisition strategy that your company is experimenting with.
Say you have 1/10th of one person doing some tweeting or facebooking, :), then add that one segment to your top two. This will allow your management to look at what they are focused on and also one thing that sounds cool but they have no idea if it is actually worth it.
(Short term focus) Win – Win (Long term focus)
How To Apply Segments / Analyze Data.
The reports you'll apply your acquisition segment to will depend on the Key Performance Indicators you have chosen. But a typical set of metrics you'll evaluate will hopefully represent a spectrum of success, like for example. . .
- How many visits did we get (to get context)
- Of those how many were new visits (if that is a focus)
- How many could we get to give us one pathetic click (bounce rate!)
- What was the cost of acquisition (if you can get total cost give yourself a gold star)
- What value could we extract at a per visit level
- How many people could we get to convert (replace total goal completions with conversion rate if you want)
- What was the total value added to our business or non-profit
This is not the end of your journey but it is a darn good start!
[If you have Web Analytics 2.0 pop the CD at the back into your computer. In dashboard examples look for Stratigent_Sample_Dashboard.xls, via my friend Bill Bruno at Stratigent. It has an excellent example of segmented acquisition display, you can immediately steal it for your company!]
Segment Category #2: Behavior.
Behavior refers to the activity people are undertaking on your website.When people show up, what is it that they are doing? Is there anything discernable / important in their behavior that is adding value to your online existence? Or, the flip side, what do we want people do to on our site, and is anyone exhibiting that behavior?
Even people who sometimes have segment their web analytics data often forget to segment by online behavior.
Many, but not all, behavior segments fall into these two buckets: People who see x pages. People who do y things.
Here are some specific examples (all of which you can create in Yahoo! Web Analytics or Google Analytics in a few seconds without having to pay anything extra for vars and slots or having to update your javascript tag or having to buy an add-on, you can also apply them to all your data including all your historical data).
Visits with more than three page views. . .
Where do these people come from? Do they buy a lot? A little? Do they write reviews? Did we acquire them or did they just show up? If they see so many pages what type of content are they interested in (politics? naked pictures? sports?)?
So on and so forth. Segmenting one behavior, understanding its value.
Similarly another could be focusing on people how add to cart and then abandon the site.
Or people who enter the site on the home page and their behavior. . .
Or people who use the site's product comparison chart or car configurator or, my fav, internal site search. Vs. those that don't.
Or people whose Days to Purchase (/Transaction) are 5 vs for those for whom the Days to Purchase is 1. . .
Or people who visited the site more than 9 times (!) during the current time period. . .
Those are the types of questions you'll answer from your behavioral segments.
The more you understand what people are doing on your site, the more likely it is that you'll stop the silliness on your site (kill content, redo navigation, make cross sells better, eliminate 80% of the ads, learn to live with 19 days to conversion, don't sell too hard, and so much more).
It is also likely (I want to say guaranteed) that you'll find the delta between what you want to have happen and what your customers want. You'll choose to make happier customers, who in turn, in the naughtiest way possible, will make you happy.
And it all stars with being able to identify and focus on the right behavior segments.
Pick at least two.
But I have to admit in this segment category I truly "play" with the data a lot because it is so hard to know what the right segments are, because visitor behavior is such a complicated thing (they are constantly trying to mess with us Analysts!).
It is only after experimentation (a lot) that I end up with something sweet.
Segment Category #3: Outcomes.
Outcomes are site activities that add value to you (business/non-profit).I find that here the problem is less that the Analysis Ninjas don't segment, rather it is that they are incredibly unimaginative.
But first what is it?
Segments with outcomes are people or visits where you get a order (at an ecommerce website) or you get a lead (at Organizing for America).
Those two are obvious right?
Segment out people who delivered those two outcomes. Give them a warm hug and a kiss. Now go figure out what makes them unique when compared to everyone else who showed up at your website, all those other people who you worked so hard to impress but failed to.
Take the insights and do more of what works for this group.
Or segment out everyone whose order size is 50% more than the average order size. . .
But there is more.
Remember macro AND micro conversions!
No one is going to sleep with you on the first date. (Ok maybe a few will!)
So focus on micro conversions that lead up to a macro conversion… like people playing a product video (or on content site watching five videos!). . .
Or signing up to show up for a protest for your ultra liberal policies!
Or apply for a trial, or download a trial product.
You can also focus on micro conversions that all by themselves are of value to you, even if not as much as the macro conversion.
For example submitting a job application.
Or signing up for a RSS feed.
Or clicking on a link to go to a different site you want them to go to (like clicking on the amazon link to go buy my book – great outcome :)).
Of course if you are really really good you'll also segment my absolute favorite metric in the whole wide world: Task Completion Rate. It is the ultimate measure of outcome (from your customer's perspective).
Net, net. . . it is absolutely critical that you segment your data by the key outcomes important to your business. Not just because your site exists to add economic value, but also because I cannot think of another way you can earn the love of your boss or get promoted.
By understanding what it is about people who deliver outcomes you can understand what to do with all those that don't convert.
Outcomes. Outcomes. Outcomes!
Pick at least two.
If you pick three or four that is ok.
If you pick nine it might be a signal you don't know what you are doing (and you want to corner your boss in a non-HR-violation manner and ask her to help you focus on the most important).
In Summary.
Segment or die.
It is as simple as that.
The next time you start to do true analysis of your data I hope you have your minimum six segments in hand (two for each category). If you do you'll find that web analytics, this world full of web metrics and what not, suddenly becomes a lot more interesting (and you no longer feel like jumping out of your office window in frustration!).
Love, money and glory await you.
Not to mention how proud I'll be of you when I see your analysis. ; )
Ok now your turn.
Are you a segmentation God? What are some of your favorite segments? Have you used this three category framework in the past to find segments? Do you think they'll work in real life? In the context of segments what do you think is missing from this blog post? What did I overlook / not stress enough?
What's your excuse for not leveraging segmentation? (Best answer to this question win's a copy of Web Analytics 2.0!)
Please share your thoughts / wisdom / critique / guidance.
Thanks.
PS:
Couple other related posts you might find interesting:
- Excellent Analytics Tip#2: Segment Absolutely Everything
- Google Analytics Releases Advanced Segmentation: Now Be A Ninja!
- Excellent Analytics Tip#4: Make Your Analysis/Reports "Connectable"
- "Dear Avinash": Be Awesome At Comparing KPI Trends Over Time
- 10 Fundamental Web Analytics Truths: Embrace 'Em & Win Big
The Difference Between Web Reporting And Web Analysis.
Occam's Razor
by Avinash Kaushik
Someone asked me this very simple question today. What's the difference between web reporting and web analysis?
My instinct was to use the wry observation uttered by US Supreme Court Justice Potter Stewart in trying to define
po rn: "I know it when I see it."
That applies to what is analysis. I know it when I see it. : )
That, of course, would have been an unhelpful answer.
So here I what I actually said:
I always find pictures help me learn, so here are some helpful pictures for you. . .
This is web reporting:
And so is this, even if it looks cuter:
And while you might be tempted to believe that this is not web reporting, with all the data and the colors and even some segments, it is web reporting:
See the common themes in all the examples above?
The thankless job of web reporting, illustrated vividly above, is to punt the part of interpreting the data, understanding the context and identifying actions to the recipient of the data puke.
If that is your role, then the best you can do is make sure you have take the right screenshots out of Site Catalyst or Google Analytics, or charge an extra $15 an hour and dump the data into Excel and add a color to the table header.
So what about web analysis?
The job of web analysis mandates a good understanding of the business priorities, creation of the right custom reports, application of hyper-relevant advanced segments to that data and, finally and most importantly, presentation of your insights and recommended action using the locally spoken language.
See the difference? It's a different job, requires different work, and of course radically different skills.
Examples of web analysis? I thought you would never ask. . .
This is a good example of web analysis:
[And not only because it is my work! Learn more about it here: Action Dashboard.]
Notice the overwhelming existence of words. That's not always sufficient, but I humbly believe always necessary.
When you look to check if you are looking at analysis or reporting look for Insights, Actions, Impact on Company. All good signs of analysis.
Here's another example of really good web analysis:
[Click on the image above for a higher resolution version.]
Ignore how well or badly the business is doing. Focus on approach taken.
Here are some things that should jump out. . . . A deliberate focus on only the "movers and shakers" (not just the top ten!). Short table: just the key data. Most of the page is taken up with words that give insights and specific actions to take.
Another example that I particularly like, both for the style of presentation and how rare it is in our world of web analytics. . .
[Click on the image above for a higher resolution version.]
No table, no rows, no pies. And yet data holds center stage with clearly highlighted actions.
Normally, we all do the column on the left (it might look different, but we have it). Unfortunately we don't appreciate is the power of the middle column ("segmentation reveled"). That is super important because it gives the recipients exposure to the hard work that you have done and in a very quiet ways increases their confidence in your work. Guess the outcome of that? They take the actions you are recommending!!
Analysts constantly complain that no one follows any of their data-based recommendations. How do you expose your hard work? In a garish Las Vegas show girl fashion where all the "data plumes" are, unsexily in this case, hanging off the body? Or, in quite concise ways? Only one of those two work.
One more? Okay here you go. . .
[Click on the image above for a higher resolution version.]
Diana has loads of observations, supported by visuals (sometimes it really helps to show the search results or the emails or the Facebook ad) with highlights (actually lowlights) in red, and finally recommendations.
And note the tie to outcomes (another common theme in all examples above). In this case, the search improvements are tied to the increase in donations I can make because of sales of my book. 1.5 extra smiles per month! (All my proceeds from both my books go to charity.) A good way to get attention from the "executive" and get him or her to take action.
Do that. A lot. Be creative. Yes it is hard work. But then again glory is not cheap, is it?
Exceptions to the rule.
Not every output you get from your Analyst, or "Analyst" :), with loads of words on it, instead of numbers, will be analysis. Hence my assertion that "I know it when I see it." Words instead of data pukes is just a clue, read the words to discern if it actually is analysis or a repeation of what the table or graph already says!
In the same vein not every output that is chock full of numbers in five size font, with pies and tables stuffed in for good measure, is a representation of web reporting. It is hard to find the exceptions to this rule, but I have seen at least two in nine years.
Top 10 signs that you are looking at / doing web analysis.
Let's make sure this horse is really and truly dead by summarizing the lessons above and using a set of signs that might indicate that you are looking at web analysis. . .
In the list above, and in the examples in this post, you see my clear, and perhaps egregious bias for business analysis and business outcomes and business actions and working with many parts of the business and business context. But I've always believed that if you and I can't have an impact then why are we doing what we do?
I hope you've had some fun learning how to distinguish between web reporting and web analysis. It is a fact of life that we need both. The bigger the company, the more they want data pukes, sorry, reporting.
But if you have "Analyst" in your job title then you perhaps now have a stronger idea of what is expected of you to earn that title. If you have hired a "web analysis consultant" and are paying them big Rupees then you know what to expect from them. Don't settle for data pukes, push them harder. Apply the rules above. Send their "analysis" back. Ask for more. Raise your expectations!!
I hope now "you'll know it when you see it," and have more datagasms!
Okay, it's your turn now.
How would you answer the question about the difference between web reporting and web analysis? What signs do you look for when evaluating the work of your Analyst or Consultants?
Please share your thoughts via comments below.
Thanks.
PS: In case you are curious here's the current official definition of po rn, as outlined in Miller v. California:
by Avinash Kaushik
The Difference Between Web Reporting And Web Analysis
Someone asked me this very simple question today. What's the difference between web reporting and web analysis? My instinct was to use the wry observation uttered by US Supreme Court Justice Potter Stewart in trying to define
po rn: "I know it when I see it."
That applies to what is analysis. I know it when I see it. : )
That, of course, would have been an unhelpful answer.
So here I what I actually said:
If you see a data puke then you know you are looking at the result of web reporting, even if it is called a dashboard.Would you agree? Got an alternative, please submit via comments.
If you see words in English outlining actions that need to be taken, and below the fold you see relevant supporting data, then you are looking at the result of web data analysis.
I always find pictures help me learn, so here are some helpful pictures for you. . .
This is web reporting:
And so is this, even if it looks cuter:
And while you might be tempted to believe that this is not web reporting, with all the data and the colors and even some segments, it is web reporting:
See the common themes in all the examples above?
The thankless job of web reporting, illustrated vividly above, is to punt the part of interpreting the data, understanding the context and identifying actions to the recipient of the data puke.
If that is your role, then the best you can do is make sure you have take the right screenshots out of Site Catalyst or Google Analytics, or charge an extra $15 an hour and dump the data into Excel and add a color to the table header.
So what about web analysis?
The job of web analysis mandates a good understanding of the business priorities, creation of the right custom reports, application of hyper-relevant advanced segments to that data and, finally and most importantly, presentation of your insights and recommended action using the locally spoken language.
See the difference? It's a different job, requires different work, and of course radically different skills.
Examples of web analysis? I thought you would never ask. . .
This is a good example of web analysis:
[And not only because it is my work! Learn more about it here: Action Dashboard.]
Notice the overwhelming existence of words. That's not always sufficient, but I humbly believe always necessary.
When you look to check if you are looking at analysis or reporting look for Insights, Actions, Impact on Company. All good signs of analysis.
Here's another example of really good web analysis:
[Click on the image above for a higher resolution version.]
Ignore how well or badly the business is doing. Focus on approach taken.
Here are some things that should jump out. . . . A deliberate focus on only the "movers and shakers" (not just the top ten!). Short table: just the key data. Most of the page is taken up with words that give insights and specific actions to take.
Another example that I particularly like, both for the style of presentation and how rare it is in our world of web analytics. . .
[Click on the image above for a higher resolution version.]
No table, no rows, no pies. And yet data holds center stage with clearly highlighted actions.
Normally, we all do the column on the left (it might look different, but we have it). Unfortunately we don't appreciate is the power of the middle column ("segmentation reveled"). That is super important because it gives the recipients exposure to the hard work that you have done and in a very quiet ways increases their confidence in your work. Guess the outcome of that? They take the actions you are recommending!!
Analysts constantly complain that no one follows any of their data-based recommendations. How do you expose your hard work? In a garish Las Vegas show girl fashion where all the "data plumes" are, unsexily in this case, hanging off the body? Or, in quite concise ways? Only one of those two work.
One more? Okay here you go. . .
[Click on the image above for a higher resolution version.]
Diana has loads of observations, supported by visuals (sometimes it really helps to show the search results or the emails or the Facebook ad) with highlights (actually lowlights) in red, and finally recommendations.
And note the tie to outcomes (another common theme in all examples above). In this case, the search improvements are tied to the increase in donations I can make because of sales of my book. 1.5 extra smiles per month! (All my proceeds from both my books go to charity.) A good way to get attention from the "executive" and get him or her to take action.
Do that. A lot. Be creative. Yes it is hard work. But then again glory is not cheap, is it?
Exceptions to the rule.
Not every output you get from your Analyst, or "Analyst" :), with loads of words on it, instead of numbers, will be analysis. Hence my assertion that "I know it when I see it." Words instead of data pukes is just a clue, read the words to discern if it actually is analysis or a repeation of what the table or graph already says!
In the same vein not every output that is chock full of numbers in five size font, with pies and tables stuffed in for good measure, is a representation of web reporting. It is hard to find the exceptions to this rule, but I have seen at least two in nine years.
Top 10 signs that you are looking at / doing web analysis.
Let's make sure this horse is really and truly dead by summarizing the lessons above and using a set of signs that might indicate that you are looking at web analysis. . .
- #1. The thing that you see instantly is not data, but rather actions for the business to take.
#2. When I see Economic Value I feel a bit more confident that I am looking at the result of analysis. Primarily because it is so darn hard to do. You have to understand business goals / outcomes (so harrrrrd!) and then work with Finance to identify economic value, and then you have to configure it in the tool and then apply advanced segments, and then figure out how things are doing. That is love. I mean that is analysis! Or at least all the work that goes into being able to do effective analysis.
#3. In the same vein, if you see references to the Web Analytics Measurement Model (or better still, see it in its entirety on one slide up front), then you know that the Ninja did some analysis.
#4. Any application of algorithmic intelligence, weighted sort, expected range for metric values (control limits), or anything that even remotely smells of ever so slightly advanced statistics is a good sign. Unknown unknowns are what it's all about!
Also mere existence of statistics is not sufficient. All other rules above and below still apply. :)
#5. If you see a Target mentioned in the report / presentation, then the Analyst did some business analysis at least. See the top right of the picture immediately above.
#6. Loads and loads and loads of context! Context is queen! Enough said.
#7. I have never seen web analysis without effective data/user segmentation. I think this statement is in both my books. . . "All data in aggregate is crap." Sorry.
#8. If there is even a hint of the impact of actions being recommended then I know that is analysis. It is hard to say: I am recommending that we shift this cluster of brand keywords to broad match. It is harder to say: I am recommending. . . and that should increase revenue by $180,000 and profit by $47,000. Look for that.
#9. If you see more than three metrics in a table you are presented with then you might not be looking at analysis.
#10. Multiplicity! If you see fabulous metrics like Share of Search (competitive intelligence) or Task Completion Rate (qualitative analysis) or Message Amplification (social media) then they are good signs that the Analyst is stepping outside Omniture / WebTrends. I would still recommend looking below the surface to ensure that they are not just data pukes, but the good thing is these are smarter metrics.
User Contributions:
#11. From Carson Smith: If someone looks at your analysis / report / presentation / dashboard and has to ask "and… as a result?", then it might be reporting. What happened should be obvious.
[I love applying the "Three Layers of the So What" test to any analysis I present or see. I ask "so what" three times. If at the end of it there is no clear action to be taken then I know it is just web reporting, not matter how great it looks or how much work went into it. Ask "as a result?" or "so what?" to your work!]
#12. From Chuck U: 1) If it can be automated, it's probably not analysis 2) If your data warehouse team says they can automate it for you, then it's definitely not analysis. [#awesome! -Avinash]
In the list above, and in the examples in this post, you see my clear, and perhaps egregious bias for business analysis and business outcomes and business actions and working with many parts of the business and business context. But I've always believed that if you and I can't have an impact then why are we doing what we do?
I hope you've had some fun learning how to distinguish between web reporting and web analysis. It is a fact of life that we need both. The bigger the company, the more they want data pukes, sorry, reporting.
But if you have "Analyst" in your job title then you perhaps now have a stronger idea of what is expected of you to earn that title. If you have hired a "web analysis consultant" and are paying them big Rupees then you know what to expect from them. Don't settle for data pukes, push them harder. Apply the rules above. Send their "analysis" back. Ask for more. Raise your expectations!!
I hope now "you'll know it when you see it," and have more datagasms!
Okay, it's your turn now.
How would you answer the question about the difference between web reporting and web analysis? What signs do you look for when evaluating the work of your Analyst or Consultants?
Please share your thoughts via comments below.
Thanks.
PS: In case you are curious here's the current official definition of po rn, as outlined in Miller v. California:
(a) whether the 'average person, applying contemporary community standards' would find that the work, taken as a whole, appeals to the prurient interest,
(b) whether the work depicts or describes, in a patently offensive way, sexual conduct specifically defined by the applicable state law, and
(c) whether the work, taken as a whole, lacks serious literary, artistic, political, or scientific value.
Tuesday, May 06, 2014
Dear Avinash: Your Digital Marketing + Analytics Challenges Answered.
Dear Avinash: Your Digital Marketing + Analytics Challenges Answered
By Avinash Kaushik
Every once in a while I take a pause and answer your questions, your burning questions (!), about digital marketing and analytics.
I'd requested you to submit questions on my Facebook or Google+ pages and am always delighted at the wide range of challenges you share.
For some answers, What is Avinash's typical day like? Hour-by-hour report please. :)?, you'll have to wait for my biography (thanks for asking Simo Ahava!). I'll answer a selection of other questions in this post.
We will cover questions in four areas: business/strategy challenges, analytics/technical challenges, career/self-development questions and rampant speculation.
Many of these questions have multiple possible answers, I invite you to participate in the discussion by adding your own answers via comments.
Let's do this!
^ Business/Strategy Challenges
Glenn Walker
Prioritize by where you will make money for your client quickly. Even the worst analytics configuration in the world will most likely allow you to measure cart and checkout abandonment rate. If you can fix that, more revenue will immediately flow into your client's bottom-line. They will show affection towards you. Bank it. Next, it should be easy to measure bounce rates for landing pages (you would have to have zero code on the site not to be able to do this). Find campaigns where they are spending most money, lower the bounce rate and reduce acquisition cost. Earn more affection.
When you feel you have enough, use it to buy time/money to go fix the configuration problems.
The mistake we make is that we obsess about every big, small and insignificant analytics implementation challenge and try to fix it because we want 99.95% comfort with data quality. Six years go by. Nothing changes for the business. We wonder why data people are not loved. :)
Don't make that silly mistake.
Strategy two…
Book two hours with the senior most company leaders who will talk to you, and create the Digital Marketing and Measurement Model .
If you have the DMMM, you have your priorities clearly laid out.
If there is anything you can measure, even with your broken analytics implementation, do that first. Add value to the business. Then prioritize fixes to the analytics implementation based on what your DMMM indicates is important to measure.
Strategy three…
Leverage the Digital Analytics Ladder for Magnificent Success to help you prioritize where to focus next.
Identify where your company is currently, what the next optimal step is in the ladder and give it all your attention in terms of data analysis or analytics code fixes.
Three different strategies to help you figure out what you should do next, even with a horrible analytics implementation. What they have in common is they encourage you to extract whatever value from the data you can first, prove your worth to the business, and then focus on analytics code fixing.
Stephen Cornelius
But if you would like a quick collection of tips: Multichannel Analytics- Tracking Online Impact Of Offline Campaigns The enabler of tying offline activity to online is ensuring you have a weak or strong primary key. The post provides more detail.
I believe these two posts with a collection of some of my favorite metrics will inspire you: 1.Best Metrics For Digital Marketing: Rock Your Own And Rent Strategies 2.Best Web Metrics / KPIs for a Small, Medium or Large Sized Business.
Joseph Boisseaux
Who is to blame?
First, I blame the analytics vendors. Vast majority of Adobe Analytics / Google Analytics remain last-click based. (Yes, yes, yes, with GA you can dive into Multi-Channel Funnels reports to move beyond last-click.) Many other tools remain 100% last-click based. If they won't take this seriously, how will they users ever see the light?
Second, I blame you and myself and all other analysts. Even when we have free solutions like MCF in GA with free attribution modeling tools, we don't really use them. Yes it does take a small mental shift, but if, the smart ones (!!), won't make the shift how can we blame anyone else? Are all your reports and presentations beyond last click?
Finally, and only lastly, I blame the management teams. They still tend to think of digital as a fulfillment channel. They have still not embraced the strategy for optimizing for marketing portfolios and still obsess about optimizing silos (they learned this from their TV, Print etc. strategies). They are actively losing money and actively creating upset customers. But they don't realize the cost. I blame their entrenched thinking.
Oh, and nothing is weird about France. Pretty much every company here is using last-click (at least until I visit them :)).
If you would like to move beyond the stupidity, sorry, of last-click:
How should you choose one? I'll share the same advice with you I'd shared about choosing a web analytics tool in Sept 2006… Get the nicest free tag management tool you can find. The Google Tag Manager is a good one, you don't need to use Google Analytics to use it. Deploy it. Enjoy it. Revel in its glory. At some point you'll bump into a small issue. Note down the limitation. If it is not a deal-breaker, keep using the tool, keep benefiting from it. Then you'll find something else. Make a note of that one.
At some point, three years from now, because you'll evolve your sophistication, you'll have five limitations and now it has reached a big problem point for your company. You now have your "what do you need to know before your choose a third party tool" list. You will make the smartest possible decision for your company because your selection will be based on your experience with a free tool that you actually used rather than reading competitive FUD literature, and you found actual problems you could not live without.
You're welcome.
^ Technical Analytics Challenges
Mehdi Oudjida
What has actually happened is that Google team has announced that they are removing the query from the referrer on ad clicks by users who use secure (SSL) search on Google.com. So analytics packages et. al. won't have access to this data.
But you as the advertiser will still see the data in the Search Terms report inside your Adwords account. You will be able to measure performance of your bids just as you did in the past.
For automated reporting you can also use the AdWords API Search Query Performance Report or the AdWords Scripts Report service.
There is a small bummer here for sure. I like to analyze my AdWords keyword performance using custom reports , especially using dimensions like Matched Search Query, inside Google Analytics and in context of other campaigns I'm running.
I can't do this anymore and I'm sad about that. But, I'm adapting to the new reality and playing with available options.
Isak Easa
Ok, only partially. But, here's an extremely detailed posts that looks at five different data sources to help you make the best of keyword data that is available in other places to optimize your SEO (or even PPC) strategies: Search: Not Provided: What Remains, Keyword Data Options, the Future.
The post covers what is still there in your analytics tools, competitive intelligence tools, Google Webmaster tools, AdWords Keyword tool, and SEO tools.
And you can definitely do brand vs non-brand analysis using these options. It is not perfect, but it is also very far from insufficient.
Alexander Velinov
Most of the time the way Google Analytics (or WebTrends or whomever) knows where someone came from is by parsing the information in the URL. If someone comes from a link, that information gets provided to Analytics, you can see where the visitor came from. If there is nothing in the referring string, that visit is marked as Direct.
If you are deliberately sending traffic, say via a campaign or an activity you are undertaking, it is best to pass that in the referral string. That way Analytics knows it was your handiwork to send that traffic. It will put that data in Campaigns section of the Acquisition report.
Here's an example. I post on twitter, http://goo.gl/W6P01k, the link brings you back to my website, and you'll see this url:
http://www.kaushik.net/avinash/best-web-metrics-digital-marketing-own-rent-strategies/?utm_source=social-media&utm_medium=twitterfbgp&utm_campaign=aktw
The URL parameters help GA put the data in the right place and classify it as a campaign. Like so…
I can now see the value of my social media campaigns clearly, and segment them by Twitter, LinkedIn, Google+ and then lastly Facebook (Oh, I love you Twitter, I love you so much!).
So a very long way of saying that if you do anything to generate traffic, always use campaign tracking parameters. Always. Typically this will apply to Paid Search, Affiliates, Email Marketing, Social Media, and Display campaigns.
If you have a lot of them, aggregate them up. In my case above I can see individual campaigns or just create and advanced segment for social-media.
Fruition Internet Marketing
I'd outlined the simplest possible option in a post on how I measured the impact of one of our radio campaigns on our digital existence and profits. Here it is: Excellent Analytics Tip #12: Unsuspected Correlations Are Sweet!
The graph above is the end result, fascinating results. Please read the blog post for all the details.
Your second option is to ensure that you invest in various techniques that allow you to create a primary key to tie your offline campaign data with online behavior and outcomes. More details are in my post on tracking online impact of offline campaigns .
Finally, the hardest option, and the one that is most rewarding and perhaps even the most accurate, is to measure effectiveness of offline campaigns by leveraging controlled experiments .
My example in the post is about measuring the value of catalog and email campaigns, but the technique you would follow would be the same. For additional inspiration seek our media-mix modeling techniques.
Measuring multi-channel campaigns and outcomes takes some effort, but if you are willing you can totally do it.
Martin Penner
All web analytics tools by default don't measure time on page for a bounced visit. So if many people come to your site and leave instantly from your home page then their time in the system is N/A (not available). Of the sessions where time is measured (because a click was made on a link that goes into your site), it is entirely possible that for a good percentage of people they land on your site, go do something else, for whatever reason, later see the tab open and make a click and go deeper into your site causing a higher time period to be recorded for your site. They can't leave the tab/page open for a lot time, after 29 minutes of inactivity the visitor session is terminated. There could be other such reasons causing your high home page time on page.
Check if you have a high bounce rate, if so you don't need a lot of people to exhibit weird behavior for your time on page metric to get messed up.
Bonus reading: Standard Metrics Revisited: #4 : Time on Page & Time on Site
Denis Pinsky
I do think the analytics industry is all it can to classify your traffic as cleanly as possible. There are other shadier ways to solve this problem, they will break privacy laws and breach user trust and so I'm glad no legitimate analytics solution is doing anything like that.
Recently the single biggest reason for a spike in Direct traffic is the massive increase in use of mobile applications by all of us. A huge chunk of social media consumption is via dedicated mobile apps. And I don't think you need me to share with you the number of mobile apps, and numbers on mobile app adoption. Mobile apps don't pass a referrer, the visitor gets classified as Direct.
So, for every single campaign you execute, very link you share via social media, and every single action you might undertake on mobile, make sure you are using campaign tracking parameters .
All the traffic you generate will now be classified correctly. The ones others generate might not be, but there is not much you can do about it.
If you want to go one more step further and really ensure all things at your end of the responsibility spectrum are covered, check that you have your analytics code implemented completely and correctly.
Bernardo Contopoulos
This is a little bit of a complex problem, so you are indeed better off working with an authorized consultant who can evaluate your unique circumstances are help you implement the right solution very quickly. Here's a list: www.bit.ly/gaac You can also try to "figure out out" :), but I'm afraid if you don't have the technical chops (and that is ok) it will simply take you too long.
All that said there are two solutions that might work.
You can use custom variables, with scope set to Visitor or User, to anonymously identify people who have received your new subscriptions and measure increased content consumption by those people. You can also of course use this strategy to differentiate between new and old subscriptions. Oh, and if you want to analyze behavior of new subscribers from a specific time period, say everyone in Jan 2016, you can use the spiffy cohort analysis option in advanced segmentation . Truly sexy stuff.
Another more advanced strategy might be to leverage the User ID option with the new Universal Analytics roll-out by the team at Google. This will allow you to do some pretty spiffy things related to tracking people and do so across devices (which my above recommendation will not do).
Bonus: User ID implementation guide .
Dan Chow
Sampling kicks in most frequently when you are looking at the data across a very large time period and use my favorite Google Analytics features like Custom Reporting and Advanced Segmentation.
Also please remember that while the default sampling is applied at 250k, you can change this (look at the top right of any given report) to anywhere from 1k to 500k.
For a more detailed and specific answer: How sampling works in Google Analytics
With sampling what GA is trying to do is not have you wait for five hours to get a perfect answer, or have your query time out, both of which happen commonly in other tools for large datasets. It is very quickly trying to five you a good enough answer. It uses very advanced strategies to ensure it is a good enough answer.
Sometimes that is simply not sufficient. Either our peers don't know how to use sampled data, or have a psychological barrier to overcome. In those cases, please use the techniques outlined above or pay for the Premium version.
Bonus: Web Analytics Data Sampling 411
^ Career / Self-Development Questions.
Kaja Sousek
So with that as a background, what do you do?
If you have enough influence (and you can have that even without a big title), then try to take charge of as much of the digital effort as you can and prove to them that by being serious you can win big. Pick the area with the most amount of revenue or cost, use data and digital savvy to improve revenue even more or reduce cost a lot. That will attract attention.
If you have very little influence, try to pick a small area. Say, email marketing. Rock it. Prove how well it can work. Perhaps the right light will shine on your effort and your management team will take you seriously, and then the digital business.
If you have no influence, keep doing the best you can but get your resume ready and find another job. This is not always an option, you might be in a geographic location where this in not an option at all. But if it is an option, in this type of a scenario without any influence for the sake of your personal passion and ambition you are better off some place else where you can add value and achieve professional success.
Webbing Yourway
If you consider disadvantage to be having a limit on how high your salary can be and how high your influence on the business side can be then yes, I do believe that having a job that is only focused on implementation is a disadvantage.
But if you are at your happiest doing a job that is technically challenging and allows you to solve difficult data collection and data processing challenges, then it is not a disadvantage. You are doing what makes you happy. Is there anything more important?
As to why I consider it to be a disadvantage (with the above mentioned definition)… Analysis is an incredibly difficult challenge not because it is hard to use the tools, it is hard because you have to be comfortable with ambiguity, you have to deeply understand business strategy, you can't just stop at data puking rather you have to identify actions to take (which means big network of people relationships and business savvy) and compute impact and then recommend things that will work (or you are out of a job). These jobs also mean, for better and for worse, more interfacing with senior management and influencing them (in your technical job you won't as much, even as your job is important), and that does matter a lot.
So, those jobs will pay more, will allow you to drive more change than a job that is simply implementation and tools training.
For more on this, and salary structures and job promotion options, please see this post: Analytics Career Advice: Job Titles, Salaries, Technical & Business Roles.
Kara Martens
I do believe that tools training can take your career forward, but less far than you might desire. You want to actually get good at analytics. The business of analysis. Transforming data into insights. And all that good stuff.
In that case seek books, blog posts, certifications that teach you how to think about analysis. This blog is a good start, :), but there are others. I link to some in the right navigation. You are welcome to consider my book Web Analytics 2.0 (which is not tool centric).
In terms of certification, I'm biased but I do recommend the Web Analytics Master Certification program at Market Motive (my start-up) that focuses on the art and science of analysis (and not reporting or a particular tool).
This blog post shares other practical tips, books and certification options: Web Analytics Career Guide: From Zero To Hero In Five Steps!
Josh Thomas
Please also see the post above titled Zero to Hero, I believe you'll find it to be of value. And please see the Unmissable Articles listed on the bottom right of this post.
^ Rampant Speculation
Suzanne van Tienen
Not people-based, a euphemism I use to refer to small groups of "persons" where you can't identify any one person. Person-based, where you can track a person and their behavior across devices and channels. Digital first. Digital and real-world in the near future.
How likely is this?
See my reply above to Joseph Boisseaux where he, rightly, complains about all of still being stuck with last-click attribution. And switching away from that is actually really easy, and businesses still refuse.
So person-based analysis will take a long time. Initially it will just be technical challenges (it is really hard to implement a unique user_id tied to one person, no matter easy analytics tools say it is). Then there will be challenges related to privacy and government rules (unclear at the moment, and if they become clear what their impact might be).
Does this mean you should not try?
No. You are making wrong decisions already by not focusing on person-based analysis. Every little step you take away from visit-based analysis makes you less wrong every day. And that is totally worth shooting for!
Let's end on that note of optimism.
As I'd mentioned at the start of this post, each question above could have a slightly different answer. I would love to have you jump in and help the folks who asked the above questions benefit from your experience and wisdom. Please share your insights via comments below.
Thank you. Merci. Arigato.
By Avinash Kaushik
Every once in a while I take a pause and answer your questions, your burning questions (!), about digital marketing and analytics.
I'd requested you to submit questions on my Facebook or Google+ pages and am always delighted at the wide range of challenges you share.
For some answers, What is Avinash's typical day like? Hour-by-hour report please. :)?, you'll have to wait for my biography (thanks for asking Simo Ahava!). I'll answer a selection of other questions in this post.
We will cover questions in four areas: business/strategy challenges, analytics/technical challenges, career/self-development questions and rampant speculation.
Many of these questions have multiple possible answers, I invite you to participate in the discussion by adding your own answers via comments.
Let's do this!
^ Business/Strategy Challenges
Glenn Walker
Hi, I have recently started working with more enterprise clients, its been fun but there are a lot of new challenges. I am having issues prioritizing 1) recommending fixing on site issues affecting real traffic levels versus 2) correcting significant configuration issues in Analytics measuring current site traffic. Both are large scale issues requiring buy in from execs and multiple departments.Strategy one…
I need to pick my spots and decide where to assign resources first. Grow traffic first, with even with bad measurement I can find positive ROI areas for growth or invest time getting Analytics in order first for more objective decision making? How do you think about making reconsiderations for a scenario like this??
Prioritize by where you will make money for your client quickly. Even the worst analytics configuration in the world will most likely allow you to measure cart and checkout abandonment rate. If you can fix that, more revenue will immediately flow into your client's bottom-line. They will show affection towards you. Bank it. Next, it should be easy to measure bounce rates for landing pages (you would have to have zero code on the site not to be able to do this). Find campaigns where they are spending most money, lower the bounce rate and reduce acquisition cost. Earn more affection.
When you feel you have enough, use it to buy time/money to go fix the configuration problems.
The mistake we make is that we obsess about every big, small and insignificant analytics implementation challenge and try to fix it because we want 99.95% comfort with data quality. Six years go by. Nothing changes for the business. We wonder why data people are not loved. :)
Don't make that silly mistake.
Strategy two…
Book two hours with the senior most company leaders who will talk to you, and create the Digital Marketing and Measurement Model .
If there is anything you can measure, even with your broken analytics implementation, do that first. Add value to the business. Then prioritize fixes to the analytics implementation based on what your DMMM indicates is important to measure.
Strategy three…
Leverage the Digital Analytics Ladder for Magnificent Success to help you prioritize where to focus next.
Three different strategies to help you figure out what you should do next, even with a horrible analytics implementation. What they have in common is they encourage you to extract whatever value from the data you can first, prove your worth to the business, and then focus on analytics code fixing.
Stephen Cornelius
How do you come up with compelling analytics KPIs if there isn't a simple relationship between online activity and profit, for example when you sell online content via a traditional offline annual subscription sales process??I'm afraid there isn't enough context in your question to answer it specifically. There can be so many different answers based on your specific scenario.
But if you would like a quick collection of tips: Multichannel Analytics- Tracking Online Impact Of Offline Campaigns The enabler of tying offline activity to online is ensuring you have a weak or strong primary key. The post provides more detail.
I believe these two posts with a collection of some of my favorite metrics will inspire you: 1.Best Metrics For Digital Marketing: Rock Your Own And Rent Strategies 2.Best Web Metrics / KPIs for a Small, Medium or Large Sized Business.
Joseph Boisseaux
Why is it so incredibly difficult to make people understand that last click attribution model is just idiot? That seems so trivial common sense that I am wondering if France is not in a parallel dimension.This does drive me bananas! In this day and age using last-click attribution to measure digital success is spectacularly dumb. Genuinely awful.
Who is to blame?
First, I blame the analytics vendors. Vast majority of Adobe Analytics / Google Analytics remain last-click based. (Yes, yes, yes, with GA you can dive into Multi-Channel Funnels reports to move beyond last-click.) Many other tools remain 100% last-click based. If they won't take this seriously, how will they users ever see the light?
Second, I blame you and myself and all other analysts. Even when we have free solutions like MCF in GA with free attribution modeling tools, we don't really use them. Yes it does take a small mental shift, but if, the smart ones (!!), won't make the shift how can we blame anyone else? Are all your reports and presentations beyond last click?
Finally, and only lastly, I blame the management teams. They still tend to think of digital as a fulfillment channel. They have still not embraced the strategy for optimizing for marketing portfolios and still obsess about optimizing silos (they learned this from their TV, Print etc. strategies). They are actively losing money and actively creating upset customers. But they don't realize the cost. I blame their entrenched thinking.
Oh, and nothing is weird about France. Pretty much every company here is using last-click (at least until I visit them :)).
If you would like to move beyond the stupidity, sorry, of last-click:
Strategic advice: Multi-Channel Attribution: Definitions, Models and a Reality Check
Tactical advice: Multi-Channel Attribution Modeling: The Good, Bad and Ugly Models.
Justin DuxTactical advice: Multi-Channel Attribution Modeling: The Good, Bad and Ugly Models.
Tag Management. What you need to know before you choose third party tool.First, if you are going to touch the code on your site make sure you get a tag management tool right now. Analytics implementations are getting numerous (tools) and more complicated with every passing data. Get a tag management tool. It will speed up code changes, it will improve the quality of your tagging, angels will sing songs in your praise.
How should you choose one? I'll share the same advice with you I'd shared about choosing a web analytics tool in Sept 2006… Get the nicest free tag management tool you can find. The Google Tag Manager is a good one, you don't need to use Google Analytics to use it. Deploy it. Enjoy it. Revel in its glory. At some point you'll bump into a small issue. Note down the limitation. If it is not a deal-breaker, keep using the tool, keep benefiting from it. Then you'll find something else. Make a note of that one.
At some point, three years from now, because you'll evolve your sophistication, you'll have five limitations and now it has reached a big problem point for your company. You now have your "what do you need to know before your choose a third party tool" list. You will make the smartest possible decision for your company because your selection will be based on your experience with a free tool that you actually used rather than reading competitive FUD literature, and you found actual problems you could not live without.
You're welcome.
^ Technical Analytics Challenges
Mehdi Oudjida
What would be the future of Adwords marketers with not provided searched query ?There might be some confusion here.
- Their reflex would be to expand the number of bought keywords as work around (manual long train work) to try to understand the performance using directly bought keywords?
- It would drive a part of ads to cheaper bids (for the beginning) of long train keywords ?
- Adwords min bulk quotas to be displayed need to be reduced by Google to follow this responsive behavior ??
What has actually happened is that Google team has announced that they are removing the query from the referrer on ad clicks by users who use secure (SSL) search on Google.com. So analytics packages et. al. won't have access to this data.
But you as the advertiser will still see the data in the Search Terms report inside your Adwords account. You will be able to measure performance of your bids just as you did in the past.
For automated reporting you can also use the AdWords API Search Query Performance Report or the AdWords Scripts Report service.
There is a small bummer here for sure. I like to analyze my AdWords keyword performance using custom reports , especially using dimensions like Matched Search Query, inside Google Analytics and in context of other campaigns I'm running.
Isak Easa
What is a way to analyze Not provided data in GA, its increasing day by, can you suggest how to analyze brand vs non brand out of it?You are in luck!
Ok, only partially. But, here's an extremely detailed posts that looks at five different data sources to help you make the best of keyword data that is available in other places to optimize your SEO (or even PPC) strategies: Search: Not Provided: What Remains, Keyword Data Options, the Future.
The post covers what is still there in your analytics tools, competitive intelligence tools, Google Webmaster tools, AdWords Keyword tool, and SEO tools.
Alexander Velinov
My question may be a trivial, but… Do I need campaign tagging with utm parameters for Google Analytics in order to receive more valuable information in multi channel funnel reporting and what exactly valuable information may I achieve ? Let use the question in general way.Let's unpack what is going on here.
I know that it depends of business, goals, measurement plan, resources and so on… But i talk in general. Btw our business is lead generation website and we have a lot of campaigns in different channels. Till now (I am in company from month) we use internal link tracking system which works only in user session, and do not use cookies so i think that we don't have exact information for real business decisions. Thanks in advice.
Most of the time the way Google Analytics (or WebTrends or whomever) knows where someone came from is by parsing the information in the URL. If someone comes from a link, that information gets provided to Analytics, you can see where the visitor came from. If there is nothing in the referring string, that visit is marked as Direct.
If you are deliberately sending traffic, say via a campaign or an activity you are undertaking, it is best to pass that in the referral string. That way Analytics knows it was your handiwork to send that traffic. It will put that data in Campaigns section of the Acquisition report.
Here's an example. I post on twitter, http://goo.gl/W6P01k, the link brings you back to my website, and you'll see this url:
http://www.kaushik.net/avinash/best-web-metrics-digital-marketing-own-rent-strategies/?utm_source=social-media&utm_medium=twitterfbgp&utm_campaign=aktw
The URL parameters help GA put the data in the right place and classify it as a campaign. Like so…
So a very long way of saying that if you do anything to generate traffic, always use campaign tracking parameters. Always. Typically this will apply to Paid Search, Affiliates, Email Marketing, Social Media, and Display campaigns.
If you have a lot of them, aggregate them up. In my case above I can see individual campaigns or just create and advanced segment for social-media.
Fruition Internet Marketing
Do you have any metrics to measure the effectiveness of offline campaigns (Print/TV/Radio)??You have three options at your disposal, depending on how hard you want to work / how accurate of an answer you want (and remember, you don't always need the most accurate answer – it is very smart to do, even back of the napkin, cost benefit analysis).
I'd outlined the simplest possible option in a post on how I measured the impact of one of our radio campaigns on our digital existence and profits. Here it is: Excellent Analytics Tip #12: Unsuspected Correlations Are Sweet!
Your second option is to ensure that you invest in various techniques that allow you to create a primary key to tie your offline campaign data with online behavior and outcomes. More details are in my post on tracking online impact of offline campaigns .
Finally, the hardest option, and the one that is most rewarding and perhaps even the most accurate, is to measure effectiveness of offline campaigns by leveraging controlled experiments .
Measuring multi-channel campaigns and outcomes takes some effort, but if you are willing you can totally do it.
Martin Penner
Why is Google Analytics telling me that the average time-on-page for my homepage is 16 minutes? It can't possibly be true?Two things to remember.
All web analytics tools by default don't measure time on page for a bounced visit. So if many people come to your site and leave instantly from your home page then their time in the system is N/A (not available). Of the sessions where time is measured (because a click was made on a link that goes into your site), it is entirely possible that for a good percentage of people they land on your site, go do something else, for whatever reason, later see the tab open and make a click and go deeper into your site causing a higher time period to be recorded for your site. They can't leave the tab/page open for a lot time, after 29 minutes of inactivity the visitor session is terminated. There could be other such reasons causing your high home page time on page.
Check if you have a high bounce rate, if so you don't need a lot of people to exhibit weird behavior for your time on page metric to get messed up.
Bonus reading: Standard Metrics Revisited: #4 : Time on Page & Time on Site
Denis Pinsky
In recent years an increasing percentage of traffic is being labeled as 'Direct', for the most part I know why this is happening, but is analytics industry working on something that will provide more accurate 'Channel' attribution??Here is a comprehensive guide to look over: Excellent Analytics Tip #18: Make Love To Your Direct Traffic The post shares six reasons why traffic is imprecisely classified as Direct.
I do think the analytics industry is all it can to classify your traffic as cleanly as possible. There are other shadier ways to solve this problem, they will break privacy laws and breach user trust and so I'm glad no legitimate analytics solution is doing anything like that.
Recently the single biggest reason for a spike in Direct traffic is the massive increase in use of mobile applications by all of us. A huge chunk of social media consumption is via dedicated mobile apps. And I don't think you need me to share with you the number of mobile apps, and numbers on mobile app adoption. Mobile apps don't pass a referrer, the visitor gets classified as Direct.
So, for every single campaign you execute, very link you share via social media, and every single action you might undertake on mobile, make sure you are using campaign tracking parameters .
If you want to go one more step further and really ensure all things at your end of the responsibility spectrum are covered, check that you have your analytics code implemented completely and correctly.
Bernardo Contopoulos
My current challenge: I want to measure how much the increase in usage of my subscription-based online content is caused by an increase in new subscriptions, and how much is caused by efforts we make to stimulate older subscribers to use more our content (ongoing training, phone calls…)?You are right, you cannot rely on new and returning visitors/users.
As far as I’m concerned, I’m dead on the unique visitors metric (or news vs. returning), as it seems more and more people/companies clean or block third-party cookies.
Short version: how can I measure the results of our efforts in client acquisition and retention distinctively, if I cannot totally rely on unique/new vs. returning visitor data?
This is a little bit of a complex problem, so you are indeed better off working with an authorized consultant who can evaluate your unique circumstances are help you implement the right solution very quickly. Here's a list: www.bit.ly/gaac You can also try to "figure out out" :), but I'm afraid if you don't have the technical chops (and that is ok) it will simply take you too long.
All that said there are two solutions that might work.
You can use custom variables, with scope set to Visitor or User, to anonymously identify people who have received your new subscriptions and measure increased content consumption by those people. You can also of course use this strategy to differentiate between new and old subscriptions. Oh, and if you want to analyze behavior of new subscribers from a specific time period, say everyone in Jan 2016, you can use the spiffy cohort analysis option in advanced segmentation . Truly sexy stuff.
Another more advanced strategy might be to leverage the User ID option with the new Universal Analytics roll-out by the team at Google. This will allow you to do some pretty spiffy things related to tracking people and do so across devices (which my above recommendation will not do).
Bonus: User ID implementation guide .
Dan Chow
What is the best way to avoid sample data in segments and views in GA? (without upgrading to GA premium :))Use standard reports. They are not sampled (unless you apply filters of some kind or advanced segmentation on top of the report and other such things).
Sampling kicks in most frequently when you are looking at the data across a very large time period and use my favorite Google Analytics features like Custom Reporting and Advanced Segmentation.
Also please remember that while the default sampling is applied at 250k, you can change this (look at the top right of any given report) to anywhere from 1k to 500k.
For a more detailed and specific answer: How sampling works in Google Analytics
Sometimes that is simply not sufficient. Either our peers don't know how to use sampled data, or have a psychological barrier to overcome. In those cases, please use the techniques outlined above or pay for the Premium version.
Bonus: Web Analytics Data Sampling 411
^ Career / Self-Development Questions.
Kaja Sousek
How to not get frustrated if you are responsible in corporation for digital marketing, but you are the almost only one there, because digital dept. is nonexistent and you get to your Hippo two times per year to discuss digital things?It does not seem like your company takes digital marketing seriously. If they did, you would see the HiPPO more than twice a year. Even if you were not the most important person on the digital side of your business.
On the surface it may looks like that everything is running smoothly as you care about website, run campaigns, do reports & analyses, but you know all the time that your company did not buy "big picture" yet.
So with that as a background, what do you do?
If you have enough influence (and you can have that even without a big title), then try to take charge of as much of the digital effort as you can and prove to them that by being serious you can win big. Pick the area with the most amount of revenue or cost, use data and digital savvy to improve revenue even more or reduce cost a lot. That will attract attention.
If you have very little influence, try to pick a small area. Say, email marketing. Rock it. Prove how well it can work. Perhaps the right light will shine on your effort and your management team will take you seriously, and then the digital business.
If you have no influence, keep doing the best you can but get your resume ready and find another job. This is not always an option, you might be in a geographic location where this in not an option at all. But if it is an option, in this type of a scenario without any influence for the sake of your personal passion and ambition you are better off some place else where you can add value and achieve professional success.
Webbing Yourway
Do you believe that a person who focused on the technical in and out of the analytical tools, had a job that did nothing but implementation and training users how to use the tool, is at a disadvantage to those those that only use the tools to drive insight / reporting but cannot tell you how the tools work?It is one of those cases were we have to define what disadvantage really means.
If yes, why do you think this is the case and do you think it is fair??
If you consider disadvantage to be having a limit on how high your salary can be and how high your influence on the business side can be then yes, I do believe that having a job that is only focused on implementation is a disadvantage.
But if you are at your happiest doing a job that is technically challenging and allows you to solve difficult data collection and data processing challenges, then it is not a disadvantage. You are doing what makes you happy. Is there anything more important?
As to why I consider it to be a disadvantage (with the above mentioned definition)… Analysis is an incredibly difficult challenge not because it is hard to use the tools, it is hard because you have to be comfortable with ambiguity, you have to deeply understand business strategy, you can't just stop at data puking rather you have to identify actions to take (which means big network of people relationships and business savvy) and compute impact and then recommend things that will work (or you are out of a job). These jobs also mean, for better and for worse, more interfacing with senior management and influencing them (in your technical job you won't as much, even as your job is important), and that does matter a lot.
So, those jobs will pay more, will allow you to drive more change than a job that is simply implementation and tools training.
For more on this, and salary structures and job promotion options, please see this post: Analytics Career Advice: Job Titles, Salaries, Technical & Business Roles.
Kara Martens
What is the best way to start really learning Google Analytics, beyond the basics? Certification? Specific reading materials? or just old-fashioned hands-on training??If you simply want to learn how to use Google Analytics, your very first stop is the Google Analytics Academy, learn all the material in the Digital Analytics Fundamentals course and proceed to take your Google Analytics Individual Qualification (IQ) test.
I do believe that tools training can take your career forward, but less far than you might desire. You want to actually get good at analytics. The business of analysis. Transforming data into insights. And all that good stuff.
In that case seek books, blog posts, certifications that teach you how to think about analysis. This blog is a good start, :), but there are others. I link to some in the right navigation. You are welcome to consider my book Web Analytics 2.0 (which is not tool centric).
This blog post shares other practical tips, books and certification options: Web Analytics Career Guide: From Zero To Hero In Five Steps!
Josh Thomas
Thanks for everything you're doing for the community. As a B2B marketer looking to get more heavily involved in web analytics, I'm looking for a place to start – specifically your books.Hello Josh. I recommend skipping Web Analytics: An Hour A Day and just jumping to Web Analytics 2.0 (chapter 5 specifically provides advice on B2B and non-ecommerce websites).
As a beginner, should I plan to start with the slightly older An Hour a Day, or is that information already in or updated within your second book Web Analytics 2.0? ?
Please also see the post above titled Zero to Hero, I believe you'll find it to be of value. And please see the Unmissable Articles listed on the bottom right of this post.
^ Rampant Speculation
Suzanne van Tienen
To what extent do you personally believe unique user (cross-device) and persona based analytics will succeed – and stick??Let's get this out of the way: The world already lives in a multi-device, multi-channel world. It is silly, even today, to pretend otherwise. Your current, today, right this very moment, digital analysis should be based on person-based analysis.
Not people-based, a euphemism I use to refer to small groups of "persons" where you can't identify any one person. Person-based, where you can track a person and their behavior across devices and channels. Digital first. Digital and real-world in the near future.
How likely is this?
See my reply above to Joseph Boisseaux where he, rightly, complains about all of still being stuck with last-click attribution. And switching away from that is actually really easy, and businesses still refuse.
So person-based analysis will take a long time. Initially it will just be technical challenges (it is really hard to implement a unique user_id tied to one person, no matter easy analytics tools say it is). Then there will be challenges related to privacy and government rules (unclear at the moment, and if they become clear what their impact might be).
Does this mean you should not try?
No. You are making wrong decisions already by not focusing on person-based analysis. Every little step you take away from visit-based analysis makes you less wrong every day. And that is totally worth shooting for!
Let's end on that note of optimism.
As I'd mentioned at the start of this post, each question above could have a slightly different answer. I would love to have you jump in and help the folks who asked the above questions benefit from your experience and wisdom. Please share your insights via comments below.
Thank you. Merci. Arigato.
Subscribe to:
Posts (Atom)