A cohort is defined as a group of individuals all sharing a common characteristic over time. As such, Cohort Analysis refers to the analytical pattern of comparing different cohorts to understand how they differ.
Cohort Analysis can help inform decisions in marketing optimization, customer success outreach, user experience design, and so much more.
With so many keen minds within the Looker community, we often hear about some pretty creative applications of cohort analysis. We thought we’d share a few of our favorites:
Simply Business, a UK based online insurance broker, used location and time-based cohort analysis to determine if radio advertising was delivering on its promise to grow business.
Even as opportunities for online marketing seem to multiply overnight, offline marketing channels continue to be excellent drivers of new customer engagement. Radio advertising, for example, remains an affordable choice for both brand-building and direct response campaigns—even for the most innovative Internet businesses. But how does a modern business, whose marketing efforts are built on ROI, apply this online marketing mindset to traditional offline tools?
Simply Business ran an experiment to see if they could attribute ROI to their radio efforts.
They designed a regionalized campaign using a sophisticated data pipeline and data mining tools that would enable them to identify which website visitors had probably arrived at based on a recent radio ad.
The company then could establish a cohort of those visitors and track them, all the way through the purchase funnel, with tremendous clarity. In turn, calculating a reasonable attribution of new revenue to the radio initiative.
Unlike other channels, such as paid search, which typically acquire customers much further along in their purchase consideration, the company expected radio to be more about driving awareness and therefore to have a longer payback period. However, by observing the behavior of the radio audience segments over time, they were able to validate that brand nurturing does eventually drive positive ROI directly—an insight which spells out the business case for further investment in radio advertising.
Harry’s, a subscription-based razor delivery service, was trying to decide which product or product bundle they should promote to first time subscribers. To do this, they dug into their data to see which products or product bundles were purchased by their highest-valued customers.
While this seems fairly run of the mill, Harry’s user experience results in a more complex user lifecycle normal grouping of longest subscribed users or other method of finding the “best” customers just wouldn’t be enough. So they got creative and implemented Kaplan-Meier analysis into their deep dive.
The Kaplan-Meier estimator is a statistic used to calculate the survival function from lifetime data. In medical research, it is often used to measure the fraction of patients living for a certain amount of time after treatment. In marketing, it can be used to analyze the amount of customers still purchasing after a defined period of time. Using this method of analysis, Harry’s could compare cohorts of users throughout their customer journey to see what the overall experience is for any one cohort at any one stage in the journey, regardless of whether the customer is still subscribed today.
To research which product or bundle they should promote to first time subscribers, the looked at their current subscribers. First they looked at cohort current and past subscribers, over time, and then lumped them into groups based on the items they purchased first to compare the groups.
Looking at the results, it suddenly became clear that users who bought only the razors - as opposed to those who purchased razor and shave cream bundles - on their first purchase were, in the long run, more valuable as customers, as they were more likely to “survive” over time.
Bonus: Try Looker on your own data today! Check out this Discourse post which provides details on how to implement this into your model right away.
They shared this information with the Product Team to investigate whether this increase was this just a coincidence or whether was there actually a problem with the latest release.
Digging into their user behavior data the Product Team set out to determine if the increase in tickets was a result of the update, but they soon faced a problem: they didn’t have data on user intention. They only had data on user activity.
When requesting assistance, one user had shared the screen shots of how they remedied their accidental charge by immediately requesting double that amount from their friend, both repaying what they sent and getting the money they had originally requested.
This seemed like an intuitive action that other users might take to get back the money they accidentally sent and was something they tracked in the app.
So, in Looker, they made a new cohort of users who paid a friend and then immediately requested 2x that amount back. The update in question had only rolled out in iOS, so they compared the instances of this happening on iPhone users compared to Android and Web.
Visualizing this data made the answer to their question very clear:
No, this support ticket increase was not a coincidence, and they had a hypothesis about what was causing this issue: the new update changed the default action from “pay or charge” to simply “pay” with a secondary option to “charge.” The intention of this change was to make it even easier to pay their friends, but perhaps they’d made it too easy.
Cohort Analysis is a simple yet powerful tool that every analyst can employ. Understanding user groups and their behavior isn’t limited to a single group or person within an organization, so with access to data and some creative thinking, anyone can derive value from comparing different user cohorts.
If you want to try this on your own data and already have Looker, check out the following Looker Blocks® to help you get started:
Oh, and request a demo today, so you can see how Looker makes Cohort Analysis quick, easy, and repeatable.