A Guide to Cohort Analysis: Why is It Important for Marketing

Immediate return marketing campaigns are a dream of marketers. However, they are possible for simple products that are available for one-click purchase. As for complex and expensive products, their promotional costs pay off longer. How to evaluate the effectiveness of such campaigns? First things first, consider the time factor and detail your customers according to it. If you are wondering how to do it, we are here to tell you about cohort analysis.

What is cohort analysis

Cohort analysis is an analytical method that involves dividing customers into groups (cohorts) based on certain characteristics and tracking their behaviour within a defined time frame. This method allows one to watch user activity in real time.

A cohort is a group of people who share one or more characteristics:

• an action (purchase, registration, click) that they performed;
• a period within which it happened.

A time reference distinguishes a cohort from a segment, a broader and more general concept.

For example, 2012 and 2018 Harvard graduates are two different cohorts, but they all belong to the «Harvard graduates» segment.

By actions, cohorts are divided into two types:

• Engagement — application installation, first click, registration.
• Monetization — purchase, payment, etc.

The cohort method considers the following features for analysis:

• A shared action for a group of users: subscription, registration, purchase, etc.
• A period within which this action happened: a day, a week, a month, etc.
• A study interval during which the cohort is observed.
• A metric that affects the business: ROI, customer retention, conversion, LTV, etc.

Cohort studies help to understand how key metrics differ for different segments, to see more details on an advertising campaign and other marketing activities (rebranding, testing a new website, etc.).

How to apply cohort analysis

Cohort study is not a universal method, it requires a sufficient number of users. It is advisable to have at least 1000 users in the database to conduct this analysis. This method is suitable for volume B2C and B2B businesses with a long sales cycle.

Cohort analysis helps to evaluate:

• Effectiveness of attraction channels

Cohort analysis will show which channels bring you the most loyal users. Then the business will allocate a larger budget for effective channels and work with them more actively. A question may arise: Why is cohort analysis necessary if the company can estimate how many clients it has received immediately after the campaign? Well, it is not that simple.

For example, 2000 users signed up for the service right after they saw Facebook ads. The marketer was happy because there was a result. However, 90% of users stopped logging into the service after a month. At the same time, we attracted users through newsletters. 1000 people came, and only 15% of them stopped using the service after a month. If we had analysed the result immediately after the campaign, we would have decided that Facebook is the most effective channel. But in fact, there was almost no target audience there.

• ROI

For a long sales cycle, the return on investment in advertising is not a quick thing. When it comes to large B2B transactions, real estate, and electronic services that can transform the entire business, it is impossible to make a purchase decision immediately after the first advertisement. So it is necessary to be patient and check the results after a while.

For example, user K first learned about Altcraft Platform and visited the website back in January when we had our advertising campaign. It takes time to explore opportunities and make decisions in the company where user K works. Only 4 months later, the user requested a demo from the service team, and 5 months later, the companies signed the contract. If we had calculated the ROI for the month after the campaign, we would have decided that it was a failure. Cohort analysis showed that it was not the case.

• LTV tracking and prediction

LTV (customer lifetime value) calculates the income from the customer for the entire period they use our products or services. This metric shows whether the costs of attracting new customers are justified. When we know how long the user stays with us and how much spends on us, we can calculate this data for similar cohorts.

• Test results

Cohort analysis will show how the conversion rate will change in the long term after updates for A/B tests. It may turn out that a successful element attracted more users but they are not quite target ones: accidentally clicked, signed up, but did not use the service.

• User activity

Cohort analysis will help you find out after what time the client stops actively using the product or leaves altogether. It is better to know such details because you can work on them with the user in advance.

For example, the analysis showed that 70% of users lose interest in the service after 3 months. In this case, the company needs to pay attention to this period: create a reactivation newsletter, offer a bonus, etc.

How to conduct cohort analysis

1. Define the goal and the metric associated with it that you will track during the analysis. Metrics are the basis for cohort analysis.

For example:

The goal is to determine the most successful sales channel for a mobile application. Let's consider conversion (registration) as a metric. It is important to consider how the Retention Rate changed in order to understand how many of the registered users remained in the application.

2. Determine the cohorts that you will study.

People who signed up for the service in June after advertising on Instagram, Facebook, Yandex and Google are 4 different cohorts.

3. Analyse different cohorts for a selected time interval.

Let's consider the result of all four cohorts for 3 months after the month of registration and estimate how many users remained active after registration during each of the months.

Cohort analysis is performed in Google Sheets or Microsoft Excel. However, you will have to figure out how to create the formula for calculating this or that metric in the sheets yourself. In Google Analytics, cohort study is automated but there is a limit for dividing users into cohorts: it is possible to track the first user action within a certain period.

Cohort analysis in Google Analytics

Marketing automation platforms can conduct cohort analysis in a more detailed and clear way. In the Altcraft Platform, cohort analysis is available for cohorts of users who have performed an action within a week or a month. You can view user activity in the report by:

• the number of unique clicks;
• the number of unique openings;
• the ratio of unique clicks to openings;
• the ratio of unique openings to sent messages;
• the number of unique clicks to sent messages.

The data is visualised as a graph and a table.

Cohort analysis in Altcraft Platform

Examples of cohort analysis

Let's analyse some examples of cohort analysis for different metrics.

Testing the effectiveness of channels

The goal is to determine which channel is the most effective one in terms of attracting new subscribers for your newsletter. We will analyse 3 cohorts by channels of attraction: a pop-up window on the website, Facebook ads, VKontakte affiliate posts. Each cohort is calculated from March 15-30 — a period for the campaign. 3000 users subscribed to the newsletter from all channels. Most of the users (1600) subscribed after seeing Facebook ads. 5 months later, 782 of all subscribers remained active. Let's have a look at the table to analyse all unsubscriptions during this period.

In the table, you can see the number of users as a percentage by month who continue to open emails.

Considering only the first results, we could conclude that Facebook advertising was the most effective in attracting subscribers. But it turned out that these users were not interested in the subscription or subscribed by accident. 5 months later, only 6% of these users still opened emails. VKontakte affiliate posts attracted the highest quality audience of all, 58% of subscribers from this channel continued to read the newsletter.

Calculation of LTV

The goal is to determine the LTV of users who started to use a food delivery app in 2020. Let's consider 3 cohorts - customers who made their first order in January, February or March 2020. We need to analyse their behaviour within 6 months. Then we will calculate ARPU (average revenue per user) for each customer. Calculations are performed in rubles.

LTV is calculated for each cohort separately or for all users at once.

To calculate LTV, keep in mind the following rule: Lifetime is the period from the beginning of your partnership with the client up to its end.

LTV for a cohort is calculated in the simplest way: add up all the ARPU values over the Lifetime of this cohort. LTV for the January cohort — 14100, February — 12600, March — 9300.

Besides, when calculating LTV using cohort analysis, it is important to consider not only numbers but also the situation in which the campaign was held. The table shows that customers spent more money in March 2020. They started to make fewer orders after 4-5 months. On the one hand, we can assume that advertising campaigns in March were the most successful ones because even those people who had used the app before started to make more orders. On the other hand, there was a lockdown in many regions of Russia, so food delivery became a necessity for many people. Then we can explain the decline in activity by the last months with the same advertising campaigns. When the lockdown was lifted, users began to go shopping themselves again. It is not correct to use this data to forecast LTV for other periods because the situation may change.

Conclusion: when calculating LTV, it is important to consider not only numbers but also the situation in which the campaign was held.

Testing

It is necessary to analyse the results obtained after testing the design of a new section of additional orders in the online store. There are two new designs A and B, and also the old one — Old. They form three cohorts. We are going to analyse CTR (click-through rate) of each of the designs within the period from June 5 to July 11. Next, we are going to analyse conversion from each design over the next 3 months.

As a result, design A received more clicks in the first week. However, its conversion rates were lower than those of designs B and Old. Besides, new designs did not show a significant increase in conversion compared to the old one. So we conclude that the concept of both designs is not the most successful one.

Conclusion

Cohort analysis is a tool that requires preparation. Marketers need to collect a lot of data and understand which metric to investigate now in order to improve business performance in the future. But the costs are worth the result — a deep and detailed understanding of the company's marketing, proper budget allocation, and effective data-driven strategies.