One of the buzzwords of the year in the small word of Analytics is undoubtedly that of cohorts. But what about its actual use? Isn’t it just another “buzzword”, or is it a major contribution benefiting web analysts?

What is a cohort?

Used for a long-time in medicine, and particularly in demographics, a cohort is a sample of the population which has experienced the same major event over the same period of time.

How does this open up new perspectives? Isn’t this just simply segmentation?

It is of course a technique used in segmentation, but what makes it different is that it explores a dimension which up until now has not been studied: time (the reference period).
The cohort will fix the 3 part visitor/action/period into a segment, making it possible to study the behaviour of this population over time.

In what cases will cohorts solve a problem that classic segmentation can’t?

In all cases where it is necessary to associate an event with a reference period. We can then multiply the examples which show the analytical value of the cohort.

The example of an e-Commerce site

Let’s take the example of an e-commerce site which has two sales’ peaks in the year: the summer and winter sales. The marketing manager would like to study the population of those who make a purchase during the summer sales: where they come from, the campaigns they were exposed to, their behavioural profile and also the share of the population who made another purchase (or who didn’t) during the winter sales etc.

In classic segmentation, the marketing manager would create a segment “conversion” (population having made a purchase).

  • By selecting the period as the summer sales, the marketing manager obtains 1,000 conversions.
  • By selecting the period as the winter sales, the marketing manager obtains 2,000 conversions.
  • No information on new purchases, on abandoned purchases or on customer acquisition.

However, if the marketing manager creates a cohort, for example a population having made a purchase in the summer:

– By selecting the period as the summer, the marketing manager will obtain the same number of conversions: 1,000.
– By selecting the period as the winter, the marketing manager will obtain 600 conversions. What conclusions can be drawn from this?

Out of the 1,000 purchasers in the summer, 600 came back to make a purchase in winter

  • = 60% of customers retained
  • = 40% of customers lost

This means that out of the 2,000 purchasers in winter, 1,400 are new customers

  • = 30% of customers retained
  • = 70% are new customers

We can clearly see in this simple example the quality of the information received: the manager can now work on more accurate segmentation for a better description of (and therefore better target):

  • Internet users who have made two purchases (what do they have in common, were they exposed to a particular offer, were they attracted by specific products etc.?)
  • Internet users who did not make a second purchase
  • New customers at the end of the year


The classic segment lets us study two segments of customers:

  • those who purchased in summer
  • those who purchased in winter

The cohort makes it possible to study three additional segments of customers:

  • those who purchased in summer but not in winter
  • those who purchased both in summer and in winter
  • those who purchased in winter but not in summer

The web analyst must then look for similarities and differences to obtain customer profiles and then draw-up efficient and effective action plans aimed at these segments. We can therefore confirm that the use of cohorts makes a major contribution to the work of an analyst in their search for efficiency. This does not mean having to choose between the classic segment and cohort, but rather using a combination of the two.

Segment attributes

Generally, it is then necessary to add an additional step to requalify the segments obtained for new criteria: the attributes (chart below) of each segment obtained must also be studied in detail to extract the different elements from them.

For example: 

An example of the attribute analysis performed on the segment: customers who made a purchase both in the summer and in the winter:

  • 70% women
  • 85% between the ages of 35 and 50
  • 68% upper socio-professional category
  • 57% arrived on the site via the newsletter
  • 73% read a comment
  • 59% left a comment
  • 89% watched the product video
  • 93% read the shipping costs
  • 81% placed an order with free delivery
  • Etc.

According to this example, the next campaign will be announced in the newsletter, will contain a video and a link to the different comments made. The next campaign will also mention the minimum total spend necessary for free delivery, and will present middle if not, top of the range products which are geared towards working-age women etc.


The cohort is an essential contribution to the segmentation process by introducing the referential of time which is associated with a type of action or behaviour. The cohort can be used alone, but for the best performance possible it is best to use cohorts with classic segmentation, allowing analysts to go that much further in their work. In all cases, the remaining final step involves requalifying the population obtained for the different social-demographic, behavioural criteria etc., which themselves will form increasingly relevant messages for improved targeting.

At this point a quote from Pangloss, from Voltaire’s Candid, springs to mind and makes us think: “All is for the best in the best of all possible worlds.” However, we need to remain objective: what has been mentioned earlier has been proven, but has restrictions when applied in real life:

  • Firstly, the use of socio-demographic criteria assumes the use of identified visitors and that a qualified database already exists (for example a CRM), which is far from being the most common case.
  • Even if we eliminate these criteria, the problem of cookies, their life span and privacy policies continues etc.

Of course, a subject as important as this deserves to have an entire article devoted to it, which is why I will not start this long task here. Contrary to what a “guru” in the field of web analytics has said (who only deals with a subject using theory and avoiding certain constraints), it is our responsibility as a major player in the field of Digital Analytics to remind everyone to what extent we are aware and attentive of these major challenges. In making this necessary reminder, we will be able to uphold our conclusion on the use of working with cohorts and the progress that cohorts represent in the segmentation process, by respecting a reasonable cookie lifespan and a coherent policy of confidentiality.


Knowledge Manager After On-the-job and Off-the-job training in purchasing and management at Carrefour, and sales training at Procter & Gamble, JM evolved in the mass retail sector in top management positions for large hypermarket, central purchasing and logistics groups, with an expatriate experience in Africa as a Central Director. In late 1995, JM created an Internet start-up company and after three years (late 1998) he joined Alain Llorens and the AT Internet team where he took up his position in sales, and was also at the heart of the pioneering adventure in Web analytics. At 55 and after almost 13 years seniority in the company, JM has been Knowledge Manager since 2009.

Comments are closed.