We all know ecommerce, for many, is a very competitive space. So we’re constantly looking for that edge to differentiate our offering from the competition and drive forward perpetual growth for our company.
In the age of information one of those edges is superior, data driven, decision making. This encompasses a lot of different techniques, however I can tell you now without a shadow of doubt that cohort analysis is one of the most important.
So without further ado…what is cohort analysis?
Cohort analysis definition
A cohort is a group of people who share a common characteristic over a certain period of time.
Cohort analysis involves examining how specific characteristics change over time, either for a specific cohort or, more commonly, between similar groups of cohorts.
Now you may think this is very similar to user segmentation (for example, you may currently segment customers to target those with a high LTV in your email campaigns). The key difference here is that a cohort is based on specific time frames. This allows us to track how users behave over time and, more interestingly, how that same behavior differs for different cohorts.
The importance of cohort analysis
Now you might be saying, why is this useful to my ecommerce business?
Well, many people fall into the trap of analysing key metrics for a specific time frame based on all their visitors and customers. They then make judgements on the performance of their store based on this information. The problem with this is that the experience is not the same for everyone. Here are the main reasons why:
1) Lifecycle stage – Every prospect and customer is at a different stage in their lifecycle. They may have only just made their first purchase or are a loyal customer making their fourth purchase.
2) Website changes – It is highly likely that your website is going through multiple site changes as you try and optimise the user experience and run new marketing campaigns. Again, this means that the site experience can be dramatically different throughout the year.
You can now see how, because the experience is not the same for everyone, taking averages of your data for say a specific month can hide important information about how certain metrics are actually trending. Are they staying the same or actually have recent site improvements to the user experience had a positive effect on the spending habits of new cohorts?
Here’s a simple example using customer retention. Over the last three months you have implemented a new email strategy designed to engage users more frequently, incentivise repeat purchase and stop churn. In order to determine if this campaign is working we need to use cohort analysis to analyse how average revenue per customer is declining on a month by month basis.
We can see that the cohort analysis shows a very good trend as a result of the new programme which is improving each month for each new cohort. If we look at the averages, which are mixing data on both new and returning customers, we can see that it doesn’t do the campaign justice. The figures are no where near as good.
“Without cohort analysis you cannot understand the interaction between short term acquisition volumes and delivering revenue into the business medium term”
Cian Weeresinghe – CMO Secretescapes.com
So hopefully you can now see that cohort analysis is a neat way to analyse if changes intended to improve your business in user experience, strategy or marketing are having the intended effect. It enables you to answer the question – are things actually improving?
Here are six more examples of how you can use cohort analysis to make better, data-driven decisions for you store.
Examples of how to use cohort analysis
1) To predict lifetime value – There is a relatively straightforward way to calculate a reasonable prediction of lifetime value using cohorts. This can be incredibly useful so you can determine how much you can spend to acquire and retain a customer. If you’re interested check out a great walkthrough by lightspeed ventures here.
2) To assess conversion funnel improvements – You have the ability to see if and to what extent a change to the customer experience has improved the conversion rate for new customers. For example, if you are trying to assess whether or not you are converting more customers on their first visit to your site.
3) Channel acquisition analysis – If you break down your average revenue per channel for different cohorts on a month by month basis you can see if your optimisation efforts for particular channels are acquiring better customers. An event better metric to use if you have the ability is to compare predictive CLV by channel.
4) Frequency analysis – Determine if over time you are making new customers purchase more frequently following their first purchase.
5) To gauge whether UX changes are improving user engagement over time or if metrics just appear to be improving due to growth – Growth can sometimes mask engagement problems. For example, if you are acquiring a large amount of new users and just focus on how revenue is increasing then things can look great however under the hood you may find that as the months go by engagement drops off very sharply. Your good performance is only sustained by your rapid growth. Don’t fall into this trap.
6) To determine how long it takes customer to perform a specific action you’re trying to encourage – This is perfect for those of you that already know what actions improve the chances of a new customer becoming a great customer. If you don’t then try doing some customer analysis on your best customers, cluster together the data they have in common. It may simply be to get them to visit the site x times every month. You can use cohort analysis to show how different cohorts behave over time and if the number of visits per month degrades too quickly for example.
Here’s a great exmaple from launchbit about the reduction in email open rates. A key metric they were trying to improve.