In the summer of 2014 my colleague Ed Gotham wrote a very good introduction to cohort analysis and why it matters in ecommerce. As he explained, there are many ways you can use cohort analysis to give you better visibility into your ecommerce performance.
To build on Ed’s piece, I want to talk about a number of cohort reports that I believe are especially important and actionable - some of which are far removed from the classic cohort chart.
The perceived complexity of cohort analysis can sometimes be off-putting, but these four concrete examples can be actioned by ecommerce managers without the need for complicated analysis - and the insights you’ll get from them could make a material difference to how successfully you grow your business.
Defining a cohort
Let’s first remind ourselves what a ‘cohort’ actually is.
Originally it described a unit within a Roman legion - six centuries, or 600 men, roughly equivalent to the modern army battalion.
Today, it describes a group of people who share a common characteristic - whose behaviour can be analysed over time. In ecommerce, these will be a group of customers, and this kind of analysis is used to:
a) assess how any marketing or other initiatives you are undertaking are influencing customer behaviour (to better be able to adjust and optimise in order to influence them more effectively)
b) determine how each cohort behaves differently to related cohorts, which can show you the most significant characteristics that contribute to performance - to better be able to optimise for those characteristics.
As I will demonstrate with my four examples, key to effective cohort analysis is being able to answer three key questions:
What specific business question am I trying to answer? (In most cases it will be an answer that will help you improve your performance).
How will I define my cohorts? What is the behaviour common to them that I want to analyse?
- Which metrics will I need to track that will answer my question?
Which are my best marketing channels at driving long-term value?
How to analyse it: first visit source
Key metrics to track: total repeat %, total revenue and average lifetime value.
It’s surprising how many marketers are still using last click campaign analysis to measure the success of their marketing efforts, as if acquiring a new customer as a direct result of a specific campaign is the ‘be all and end all’ of marketing.
Of course, there will be multiple factors that go towards generating purchases and repeat purchases over time, so no one attribution model will give you the absolute truth on their relative value - there are too many factors in play.
But what is essential is that you take the long term view. As a rough estimate of how lifetime value builds over time, a first visit source analysis is key in helping you understand the long term value of customers acquired from specific channels.
Channels by first visit source, cumulative over all time
There are two ways to run this analysis - either as a total (example above), or broken down into discrete units of time (like below), more like a traditional cohort analysis.
Channels by first visit source, one month cohort tracked over 12 months
The key metrics to look at are:
- the percentage of the cohort who have become repeat customers
- the average total lifetime value of customers acquired in this channel
How does seasonality affect long-term customer value?
How to analyse it: Month of first purchase cohorts
Key metrics to track: revenue, customer repeat rate, and average order value.
This is the classic cohort chart, looking at customers acquired by month, and seeing how their monthly spend, repeat rates, and average order values play out over time.
This will not only give you a clear picture of the value of customers acquired at certain periods of the year, but also how subsequent retention activity - and even website changes, may have affected their behaviour.
You should also cross-match against campaigns you were running at those times - and indeed one way for you to strip out some of these influences would be to filter these cohorts by channel.
How to analyse it: Category of first purchase cohorts
Key metrics to track: customer repeat rate, average order values, lifetime value and order gaps.
These cohorts are about behaviour, and identifying patterns based on first purchase. In this case, we want to analyse categories or products first bought to see whether this gives a clue to a customer’s future value and purchasing behaviour.
A key metric to add here, in additional to AOV, repeat rate and CLV, is average order gap or repeat speed. This is particularly relevant when we consider categories - for example a drinks company might be selling beer, wines, spirits and soft drinks, and a homewares company might be selling furniture as well as smaller homewares.
The category first purchased is often a strong signal as to which type of customer they are, and can define future behaviour - for example how often they will reorder. It’s also crucial to understand, especially for businesses with a varied range of products, which products or categories to concentrate on to give them the best chance of growing long term value.
How successful is my retention marketing?
How to analyse it: customer retention rate.
Key metric to track: customer retention rate.
The last report is one of my favourites, because it’s the most simple. It’s not so actionable as the three above, but it’s more a set of flashing lights on the console of your business, and a very good indicator as to how your business is growing.
Essentially we are taking a group of customers who first shopped with you in a defined period and calculate whether, at certain points in the future, they could still be considered active customers.
The first stage of this is setting a definition for ‘active’ customers - this will depend on your business model, but let’s say you consider active to be people who have shopped within the last six months.
So in this case we take a slice of customers who first shopped, say, in January 2013, and see what % were active customers on 1st January 2014, and what % were still active on 1st January 2015 (i.e. had made a purchase between July 1st and December 31st 2014).
Of course the latter percentage will invariably be lower than the former, but the point here is to compare two different trends - how consistent the rate of decline is between the two, and how consistent the %s over time are for each cohort.
I hope I have shown the value of analysing your customers from the angle of cohorts. Many ecommerce managers will be running some, if not all, of these reports, either on a regular or periodic basis. But if one is missing from your arsenal, I highly recommend adding it. It may well give you some key insight that could make a big impact on your future performance.