In my original series on visitor segmentation, I promised to provide some examples from Unica’s NetInsight as well as Omniture’s Discover and Data Warehouse. The previous installments covered numerous Omniture examples, so I decided to use NetInsight for this topic – building descriptive segment profiles.
Not every analysis or use of segments requires building descriptive profiles. If you're analyzing a conversion funnel and build a segment of 1st time users, you may only need to see the funnel performance for that segment. The segment itself (1st time users) is self-explanatory and additional descriptive information isn’t necessary.
But for the majority of segmentation uses, it’s important not just to build a segment but to show how it differs from the average site visitor and from other visitor segments.
What are some examples? If you code your site visitors with offline personas, you want to be able to see how each persona actually behaves and compare them easily. If you want to understand the before vs. after behavior for visitors around a key event (like Registration) you need to be able to profile those before and after time-periods to reveal what changed. If you want to understand how a group of visitors in a specific use-case (like catalog searchers) behave differently from other groups (like non-catalog searchers), you need to be able to easily identify and describe those differences.
The ability to create a rich and interesting description that shows how the visitors in the segment actually behave is an essential skill for an analyst. You can’t use the same profile for every problem and nearly every business has its own unique variables and requirements when it comes to segment descriptive. But there are some obvious and common techniques that apply to nearly any segmentation.
In today’s post, I’m only going to cover some of the basic elements in NetInsight for creating these profiles. In following post(s), I’m going to show more complete examples using some of the approaches described below.
I decided to work with an example based on offline categorizations of visitors. Many, many companies track important visitor types, personas, and visitor characteristics for logged in and registration-based sites. Moving these categorizations into your web analytics solution is tremendously useful. These “natural” segmentations profoundly improve reporting and analysis.
One of the first things you’ll probably want to do when you integrate these categorizations is understand how each visitor group uses the web site and how they differ. So that will be the basis of our profiling exercise.
In this case, the company had a half-dozen basic categorizations of their users – these represented fundamentally different types of business client. They passed these categorizations (along with a user-id) to NetInsight. One of the really nice things about NetInsight is that profiles (think Report Suites in Omniture though there are some differences) can be keyed on multiple identifiers. This allows them to join pages and sessions based on both the cookie id and the visitor id. You can’t really do this in Omniture – and the lack of this feature makes tracking visitor behavior over time and across secure and public site sections rather painful.
In NetInsight, nearly everything custom is passed as a parameter. Out-of-the-box, in the interface, you get a simple report that shows the basic stats for any given parameter. Here’s the report for the user parameter (I’ve blurred the data throughout):
If you haven’t worked with NetInsight, one pleasant surprise is how customizable reports in the basic interface are. You have a full report-builder, but any report can be fully customized by adding or removing dimensions (row keys), metrics (columns) or filters (segments). You see just a sample of the available metrics for this report in the right hand panel (Available Metrics).
For this next report, I dropped the Last Visit field (nearly always useless) and added a content categorization dimension:
This is the beginnings of a useful report. It shows how much of each key type of content was consumed by each visitor category. That’s a fundamental segment characteristic – but this report is still thin and difficult to consume.
Here’s a more robust version with some deeper metrics:
A report like this provides a snapshot of total behavior (views, visits, views/visit, visits/visitor, avg. visit frequency and avg. time) at the aggregate segment level and for each of the major content categories.
I might also choose to configure my report for Content Categories and eliminate the User Type dimension entirely. I can still filter the entire report by any specific user type:
One painful aspect of both these reports is that it’s hard to compare segments side by side. NetInsight provides several days to make that easier. One way is to build Custom Metrics that are segment specific (a Metrics of Views by a Category for example). With this technique, you can show multiple segments side-by-side in columns for one or more dimensions.
Here’s an example where I’ve profiled several segments by campaign sourcing:
This report compares the Views/Visit for three different segments (and all visitors) across different campaign channels. Since the segments are side-by-side, comparison is much easier.
Filtered metrics like these are great for profile comparison. The only drawback is that they take some work to create and populate in the interface. It’s a several step process and if you have a good number of segments and a good number of metrics, it can be quite a bit of work.
Another technique that I’ll make extensive use of in the actual profile reports is indexing. When I look at segment usage of content categories, for example, the most important measures are indexical. A given segment may consume more of category X than anything else. But if everyone consumes more of category X, this doesn’t really distinguish the segment visitors.
In the following report, I created two metrics – one is segment specific metric of page views (Views / Visit for Indiv) and the other is an index that compares this rate to the rate for all visitors. I wouldn’t need it, but in the report below I added the View/Visit rate for all visitors. In an actual profile report, I’d probably just use the indexical:
I haven’t really covered much new ground in this post – but I wanted to give a sense of how NetInsight covers the basics of segment reporting (basic segmentation, multi-dimensionality, filtered reports, filtered metrics and indexicals.
Not every analysis or use of segments requires building descriptive profiles. If you're analyzing a conversion funnel and build a segment of 1st time users, you may only need to see the funnel performance for that segment. The segment itself (1st time users) is self-explanatory and additional descriptive information isn’t necessary.
But for the majority of segmentation uses, it’s important not just to build a segment but to show how it differs from the average site visitor and from other visitor segments.
What are some examples? If you code your site visitors with offline personas, you want to be able to see how each persona actually behaves and compare them easily. If you want to understand the before vs. after behavior for visitors around a key event (like Registration) you need to be able to profile those before and after time-periods to reveal what changed. If you want to understand how a group of visitors in a specific use-case (like catalog searchers) behave differently from other groups (like non-catalog searchers), you need to be able to easily identify and describe those differences.
The ability to create a rich and interesting description that shows how the visitors in the segment actually behave is an essential skill for an analyst. You can’t use the same profile for every problem and nearly every business has its own unique variables and requirements when it comes to segment descriptive. But there are some obvious and common techniques that apply to nearly any segmentation.
In today’s post, I’m only going to cover some of the basic elements in NetInsight for creating these profiles. In following post(s), I’m going to show more complete examples using some of the approaches described below.
I decided to work with an example based on offline categorizations of visitors. Many, many companies track important visitor types, personas, and visitor characteristics for logged in and registration-based sites. Moving these categorizations into your web analytics solution is tremendously useful. These “natural” segmentations profoundly improve reporting and analysis.
One of the first things you’ll probably want to do when you integrate these categorizations is understand how each visitor group uses the web site and how they differ. So that will be the basis of our profiling exercise.
In this case, the company had a half-dozen basic categorizations of their users – these represented fundamentally different types of business client. They passed these categorizations (along with a user-id) to NetInsight. One of the really nice things about NetInsight is that profiles (think Report Suites in Omniture though there are some differences) can be keyed on multiple identifiers. This allows them to join pages and sessions based on both the cookie id and the visitor id. You can’t really do this in Omniture – and the lack of this feature makes tracking visitor behavior over time and across secure and public site sections rather painful.
In NetInsight, nearly everything custom is passed as a parameter. Out-of-the-box, in the interface, you get a simple report that shows the basic stats for any given parameter. Here’s the report for the user parameter (I’ve blurred the data throughout):
If you haven’t worked with NetInsight, one pleasant surprise is how customizable reports in the basic interface are. You have a full report-builder, but any report can be fully customized by adding or removing dimensions (row keys), metrics (columns) or filters (segments). You see just a sample of the available metrics for this report in the right hand panel (Available Metrics).
For this next report, I dropped the Last Visit field (nearly always useless) and added a content categorization dimension:
This is the beginnings of a useful report. It shows how much of each key type of content was consumed by each visitor category. That’s a fundamental segment characteristic – but this report is still thin and difficult to consume.
Here’s a more robust version with some deeper metrics:
A report like this provides a snapshot of total behavior (views, visits, views/visit, visits/visitor, avg. visit frequency and avg. time) at the aggregate segment level and for each of the major content categories.
I might also choose to configure my report for Content Categories and eliminate the User Type dimension entirely. I can still filter the entire report by any specific user type:
One painful aspect of both these reports is that it’s hard to compare segments side by side. NetInsight provides several days to make that easier. One way is to build Custom Metrics that are segment specific (a Metrics of Views by a Category for example). With this technique, you can show multiple segments side-by-side in columns for one or more dimensions.
Here’s an example where I’ve profiled several segments by campaign sourcing:
This report compares the Views/Visit for three different segments (and all visitors) across different campaign channels. Since the segments are side-by-side, comparison is much easier.
Filtered metrics like these are great for profile comparison. The only drawback is that they take some work to create and populate in the interface. It’s a several step process and if you have a good number of segments and a good number of metrics, it can be quite a bit of work.
Another technique that I’ll make extensive use of in the actual profile reports is indexing. When I look at segment usage of content categories, for example, the most important measures are indexical. A given segment may consume more of category X than anything else. But if everyone consumes more of category X, this doesn’t really distinguish the segment visitors.
In the following report, I created two metrics – one is segment specific metric of page views (Views / Visit for Indiv) and the other is an index that compares this rate to the rate for all visitors. I wouldn’t need it, but in the report below I added the View/Visit rate for all visitors. In an actual profile report, I’d probably just use the indexical:
I haven’t really covered much new ground in this post – but I wanted to give a sense of how NetInsight covers the basics of segment reporting (basic segmentation, multi-dimensionality, filtered reports, filtered metrics and indexicals.
In the next post, I’ll cover the basics of what needs to be in a good segment profile.
And don't forget, I'll be doing a full class on Segmentation at our MIS 2010 Think Tank Training for Unica NetInsight. If you're attending MIS 2010 next month, do check it out!
Comments