The ABCs of Cohort Analysis

We — and the brands and nonprofits we work with — put the supporter or consumer at the center of everything we do. Designing campaigns with the user in mind will focus your content on the needs and behaviors of your supporters — not on whims or assumptions. Through personalization, customized user journeys, and unique content that’s based on someone’s relationship with the brand, we help our clients develop user-centric approaches.

But how do we extend that to all aspects of our work? Specifically, how do we practice user-centric analytics?

Cohort analysis is a general term for analyses that are technically focused, with the individual as the unit of analysis. We develop cohorts by grouping individuals in our communities based on a common characteristic. One example of cohort analysis would be looking at how an email performed with desktop users versus mobile users. Contrast this with analyzing an email campaign, where you would typically break out the performance of each individual email, or a paid media campaign with individual ads or pages. (Note: You might refer to “cohorts” as “segments” — we’re using “cohorts” to refer to this principle broadly.)

You don’t need to do a large or complicated analysis to start reaping the rewards of analyzing at the cohort level. Incorporating some of the principles of cohort analysis into the strategies you’re already doing can make them more effective, and, ultimately, make your programs better.

Here are the ABCs of building cohort-level analysis into your programs:

A: Acquire the right audience

One of the most valuable cohort analyses you can do is a lifetime value analysis for different acquisition sources. Acquisition source captures a lot of valuable information that affects your supporters’ interactions with you. For example, a user who comes to your site organically to make a donation is fundamentally different than the one you acquire through a paid channel or another campaign, or one who joins from a forwarded email.

One key output from this analysis is understanding the return on investment (ROI) for paid acquisition campaigns. Did the 1,000 email addresses you acquired through a partnership campaign end up donating more than you paid to acquire them? Are supporters from one social platform more likely to purchase your product than those from another, all else being equal?

Through our work with a large advocacy organization with both fundraising and advocacy goals, we found that supporters who joined the list via an advocacy action were both more likely to donate and more likely to give larger gifts, compared to supporters who joined via other signup methods. This discrepancy tells us that the acquisition method plays a big role in longer-term fundraising success, and shows the organization where they should be investing more heavily.

A lifetime value analysis also allows you to measure your acquisition efficiency for the actions that matter most to your organization. If your objective is advocacy, that’s easy — you can total up the number of actions taken by supporters from a source and move from a simple cost per acquisition (CPA) measurement to a cost per advocacy action (CPAA) metric and measure whether you’re acquiring the right kind of supporters to effect legislative change.

B: Bias-proof your user personas and surveys

User personas and journeys can be effective ways to represent your supporters as individuals with distinct needs, behaviors, and attributes. Incorporating cohort-level quantitative data can be a way to make your user personas more resistant to bias from the people creating them. By defining cohorts based on acquisition method, giving history, or level of engagement, you can gain a better understanding of the size and importance of different segments, and apply a quantitative lens to an otherwise qualitative practice.

And in those cases where you have a lot of data to sort through and don’t know where to start, you can use machine learning methods to create cohorts using a clustering analysis. (Note: If you want to learn more about how we’ve used clustering analysis to help organizations communicate with their digital supporters, get in touch.)

If you’re surveying your supporter or consumer base, I’m guessing you’re not getting a fully representative sample. Combine your survey with a cohort analysis, and look at responses based on the concrete data you have about respondents. Do your donors make up 10% of your email list but 20% of survey respondents? Weight your responses — and how you interpret them — accordingly. With well-defined cohorts, you can analyze how responses differ between groups and develop testable theories about what drives different supporters.

C: Content and cadence are crucial

When’s the right time to ask for a second gift? Are you sending too much email? Are you sending enough email? Should you be sending more coupon codes to lapsed customers?

These questions have plagued non-profits and brands for years, and although there’s no single right answer, we can get closer to the truth with some straightforward cohort analyses.

Split your community into cohorts, e.g. count how many of your end-of-year donors last year gave again this year. Is your donor retention rate lower than 50%? This could be a sign that you need to rethink your communications strategy. (Shameless plug: Donor acknowledgment and other personalization tactics are easy as pie in the BSD Tools).

Are certain acquisition sources more likely to unsubscribe within the first month of joining your email list? Supporters who join from different sources or different methods (like social, or gated coupon codes, or paid media ads) have different built-in levels of awareness with your organization or brand. There’s always an opportunity to refresh your welcome series to be more engaging, informative, and, well, welcoming.

This is just a taste of how you can start practicing user-centric analytics right now, in the programs and analysis you’re already conducting, and begin to realize the benefits of it.

Obviously, the possibilities are endless. If you want help thinking through more ways analytics can make your programs run even better, get in touch with us.