Three keys to understanding cohort analysis

[mdsojolh43]

A cohort is always framed within a given period, during which a certain group of users navigated your site, and/or committed specific actions. The first thing to do is therefore to define this period of time to study. A bit like a race, which has a beginning and an end, what interests us about our cohort is to know what behavior it had during this "race" on your site. More concretely, you could - for example - analyze the behavior of a cohort during an event such as sales. Beyond the chosen period, and the profile of your cohort, you can of course sectorize a little more by choosing to follow, for example, only users arriving on your site from Facebook, Google, or other...

Key 2: Define latency
Now that our cohort is defined, in terms of period and persona , it is time to define the duration during which you will follow your cohort. If, for example, your objective is to understand the behavior of your users who return a month after their first visit: the duration of your analysis will therefore have to extend over one month. There is no good or bad way to define the period over which you will follow your cohort, it all depends on your activity, the events that are in progress, the industry in which you operate… We advise you to do what seems most logical to you, based on your own expertise.

Key 3: End date of your analysis
If you want to analyze a cohort of visitors who visited your site between January 1st and 7th, with a one-month latency: you will obtain your analysis results on February 7th. Once this date has passed, the usa whatsapp data users you have observed will no longer be considered as belonging to this cohort. Be precise in choosing the date on which your cohort analysis should end, otherwise you may see incomplete feedback. Let's take the example below to better understand:
example-cohort-analysis

In this example, we defined a cohort and wanted to know how much time, on average, they spent on our site. Google Analytics detected 1,374 users that fit our cohort between May 27 and June 2, and then continued to analyze the number of visits from this same group of people week after week. Finally, in this case, the cohort analysis ends on July 15.

For example, we can see that the users in our cohort who visited our site during the week of May 27 to June 2 stayed (on average) 2m27 on our site during that week, then 16 seconds the following week, and finally 6 seconds during the third week of analysis. What is surprising in this example are the last results for "Week 1" and "Week 2", respectively 4 seconds and 1 second... These figures can be explained simply because this screenshot was taken on July 9! Which means that, for the "Week 1" and "Week 2" that we have just mentioned, the information retrieved only corresponds to 2 days out of 7 of the week, which necessarily weighs down your statistics at that precise moment. Finally, still in this example, the cohort analysis ends on July 15, which means that the "Week 2" box relating to the week "July 1 - July 7" will remain empty, since the latter is supposed to analyze user behavior at a date after the end of the analysis.