Despite clearly remembering learning about it my first week on the job at Omniture in 2006, I realized recently that I did not have a lot of confidence in what participation and linear allocation would do in certain situations in Adobe Analytics. So I put a good amount of effort into testing it to confirm my theories, and I figured I’d pass along what I discovered.
First, the Basics: eVar Allocation
You may already know this part, so feel free to skip this section if you do. Allocation is a setting for Conversion Variables (eVars) in Adobe Analytics, with three options:
Let’s take a simple example to show what how this effects things. Let’s say a user visits my site with this flow:
|Page A||Page B||Page C||Page D||Form Submit- Signup|
|s.eVar5=”Page A”||s.eVar5=”Page B”||s.eVar5=”Page C”||s.eVar5=”Page D”||s.events=”event1″|
Most Recent (Last)
Most eVars have the “defaultiest” allocation of “Most Recent (Last)”, meaning in an event1 report broken down by eVar5, “Page D” would get full credit for the event1 that happened, since it was the last value we saw before event1. So far, pretty simple.
Original Value (First)
But maybe I want to know which LANDING page contributed the most to my event1s (there are other ways of doing this, but for the sake of my example, I’m gonna stick with using allocation). In that case, I might have the allocation for that eVar set to “Original Value (First)” so then “Page A” would get full credit for this event1, since it was the first value we saw for that variable. If my eVar is set to expire on visit, then it’s still nice and straightforward. If it’s set to never expire, then the first value we ever saw for that user will always get credit for any of that user’s metrics. If it’s set to expire in two weeks, then we’ll see the first value that was passed within the last two weeks.
This setting is frequently used for Marketing Campaigns (it’s not uncommon to see s.campaign be used for “Most Recent Campaign in the last 30 days” and then another eVar capture the exact same values, but be set to “Original Campaign in the last 30 days”).
If I’m feeling a bit more egalitarian, and want to know ALL the values for an eVar that contributed to success events, I would choose linear allocation. In this scenario, all four values would split the credit for the one event, so they’d each get one fourth of the metric:
(Though it may not actually look like this in the report- by default it would round down to 0. But I’ll talk about decimals later on).
So, that’s allocation.
Then what is participation?
Participation is a setting you can apply to a prop, so that if you bring a Participation-enabled metric into the prop’s report, you can see which values were set at some point before that event took place. Repeat: to see participation you must have a prop that is set to “Display Participation Metrics”:
And the metric you want to see needs to have participation enabled (without this, in the older Reports and Analytics interface, that event won’t be able to be brought into the prop report):
Unlike linear allocation for an eVar, participation for a prop means all the values for that prop get full credit for an event that happened. So, given this flow:
|Page A||Page B||Page C||Page D||Form Submit- Signup|
|s.prop1=”Page A”||s.prop1=”Page B”||s.prop1=”Page C”||s.prop1=”Page D”||s.events=”event1″|
You would see a report like this, because each value participated in the single instance of that event:
New Learnings (for me): Content Velocity
One thing these settings can be used for is measuring content velocity: that is, how much a certain value contributed to more content views later on. For instance, if I have a content site, and I want to know how much one piece of content tends to lead to the reading of MORE content, I might use a participation-metric-enabled prop with a participation-enabled Page View custom event, or I might use an eVar with linear allocation against a Page View custom event (whether or not the event has participation enabled doesn’t matter for the eVar). For my test, I did both:
|Page A||Page B||Page C||Page D|
The prop version of this report would show me that Page 1 contributed to 4 views (its own, and 3 more “downstream”). Whereas Page 2 contributed to 3 (its own, and two more downstream), etc…
Alternatively, the eVar would show me some thing pretty odd:
Those weird numbers don’t make sense on this small scale (how could 0 get 6.3%?), because it is rounding, and not showing me decimals. If I want to see the decimals, I can create a really simple calculated metric that brings in my custom Page View event (event1) and tells it to show decimals:
The report then makes a little more sense and show us where the rounded numbers came from (and how Page 4, with “0” Page Views, got 6.3% of the credit), but may still seem mysterious:
Those are some odd numbers, right? Here’s the math:
|Page 1||2.08||1+0.5+0.33+0.25||It got full credit for its own view, then half the credit (shared with page 2) for the event on Page 2, then a third of the credit (shared with Page 2 and Page 3) on Page 3…|
|Page 2||1.08||0.5+0.33+0.25||It only got half credit for the event that took place on its page (shared with Page 1), then a third of the credit (shared with Page 1 and Page 3) on Page 3, etc…|
|Page 3||0.58||.33+.25||It only gets a third of the credit that took place on its page, and a quarter of the credit for the fourth page.|
|Page 4||0.25||0.25||The event that happened on this page is shared with all four pages.|
Crazy, right? I’m not going to tell you which an analyst should prefer, but as always, you should ask the question: “What will you DO with this information?”
What happens when multiple values appear in the same flow?
Let’s say the user does something like this, where they hit one value a couple page views in a row (Page B in this example), or they hit a value 2 separate times (Page A in this example):
|Page A||Page B||Page B
|Page C||Page D||Page A
For the prop, it’s pretty straightforward. This will look like 6 event1s, where Page A gets value for all 6, and Page D gets credit for just 2 (itself, and the Page A that came afterwards):
For the eVar, it gets a little more complicated (I added in a calculated metric so you can see the decimals). Page A (accessed twice at separate times) got double credit for the conversion (which I might have predicted), but Page B (accessed twice in a row) ALSO gets double credit for the conversion (which I didn’t predict, probably because I’m too used to thinking in terms of the CVP plugin):
A couple things to be aware of:
- Settings for participation and allocation don’t apply retroactively- you can’t apply them to existing data. If you want to start using it, you need to change your settings and you’ll see it applied to future data. However, this can mess with existing data, so be careful.
- There appears to be a bug in Analysis Workspace for both. I’m going to follow up with Adobe, but I basically can’t get either Prop participation or Linear Allocation to work in Workspace. I’ll come back and update this post if I get more info from Adobe about this.
Both participation and linear allocation aren’t used often, but they can uniquely solve some reporting requirements and can provide a lot of insight, if you know how to read the data. I hope my experimentation and results here help make it clearer how you might be able to use and interpret data from these settings. Let me know if you have other use cases for using these settings, and how it has worked out for you!