How to Do a Cohort Analysis from Scratch

Nov 03, 2020

SaaS is full of buzzwords, abbreviations, and debated metrics. Sometimes it's quite challenging to make sure you're tracking the right stats. Chasing unimportant data and comparing useless statistical sets is time lost.

Cohort analysis sounds like it's just another one of those unnecessary and time-consuming things. While the setup and tracking might be tedious, it's anything but useless. Tracking bird-eye-view metrics over long periods can give you a pretty good idea of where your business is headed. Cohort analysis focuses the efforts on more in-depth and better-optimized results.

In this article, we explore all aspects of the magical tool called cohort analysis. We'll reveal why it should be a number one priority for your business if you already have healthy metrics-tracking habits. Let's dive in.

What Is Cohort Analysis

A cohort is a group that shares a definite characteristic. Cohort analysis is a visualization mechanism. It's a tool that lets you segment the customers and use them to investigate statistical outcomes. This study combines various SaaS metrics and analytical factors with the group's unique attribute. Each cohort's characteristic acts as a defining facet for the result.

Basic Cohorts

There are two distinct variations of cohorts: basic and segmented. Basic cohorts include the customer acquisition month, plus another metric for cross-reference. The customer retention rate, acquisition cost, and customer lifetime value are all among those data points.

For example, with cohorts, you can choose to watch a group of male customers and their purchasing habits over the first three months of 2020. This overview will give you valuable insight into a specific demographic and help you conduct a behavioral analysis for future strategic decisions.

To make it even more interesting, use other characteristics to enhance the current cohort. Add the source of customer acquisition and have a targeted analysis of the group behavior related to that specific channel audience. You can then use the segments to decide which of your origin points is performing better.

You might come across evidence that will make you change the acquisition channel strategy to make it use its full potential.

Segmented Cohorts

Another cohort type is the 'segmented' one. Here, you add more demographic characteristics to each group (e.g., users' age). This type of cohort analysis adds a sharper edge and specificity to the expected outcome. For example, the behavior of 18 to 25-year-old customers can differ quite a bit from 30 to 45-year-old ones.

During cohort analysis, the most important thing is to be consistent about the selected characteristics and view the changes over fixed periods. In most SaaS companies, monthly evaluations and reports are standard. Other companies that have enough information might want to try the weekly method.

The power of cohort analysis lies within your ability to decide which factors you should check for more precise outcomes. Also, make sure that you have enough data to manipulate and superimpose for the needed results. To do this, track the essential SaaS metrics. We have a separate article in our blog in case you need a refresher.

Why Is it Important for SaaS?

Learning as much about your customer as possible is the basis for SaaS success. User behavior is everything when it comes to new client acquisition or the retention of existing ones. It also helps to optimize your business operations, pricing, generate ideas for new features, and much more.

You can spend time scrutinizing the broader, higher-level metrics and get excellent results. However, cohort analysis gives you the exact identifier, helping you target specific company issues.

Cohorts can be a gold mine for countless meaningful insights if you target them in the right direction. Matching them with other metrics has the potential to give you precious answers to any question you ask.

For customer retention, for example, you can find patterns and predict future behavior. This data can also give you the best planning ability and help grow the company. The more you know about a specific cohort, the better questions you'll ask about it, which is the only way to reveal blind spots in business operations.

You can also 'upgrade' the analysis, comparing two similar cohort results. Doing this has the potential to reveal larger, company-wide patterns, leading you to rethink strategy or abandon some operations.

The third level of cohort analysis is gathering historical outcomes and using them to predict the future. Comparing historical data with current information reveals long-lasting company trends. It helps find patterns and repeating behaviors within cohorts over time. With this knowledge, it would be logical to assume a similar outcome for the next period (e.g., month or year).

How to Do It?

For an excellent cohort analysis, you need a plan. These projects usually have many moving parts. You should be very specific about expected outcomes and the questions you ask before starting the research.

We've devised a template for you. Of course, each company can use this as a guideline, then specify the goals and questions it has in mind.

Ask a Question

Before starting the analysis, make sure you're asking the right questions. Suppose you want to know which software tier is more popular with your users from social media, such as Facebook.

Then, identify the type of data you need to answer this question. First of all, monthly historical data is necessary to make sure your sample size is enough to make assumptions. Next, consider two more data points: the cohort of people coming from Facebook and the number of people that chose each tier.

Collect the Data

To start your cohort analysis from the best possible spot, pull the statistics we mentioned above.

In our example, you'll need to know where your customers are coming from. For that, set up a tracking mechanism and have it send information to your analytics program. This arrangement will allow you to learn more about the customers coming to your website from Facebook. Once the user is on your site, it's not difficult to know which plan they choose.

If you want to have all the stats ready for this process, be systematic about it. Track crucial SaaS metrics from the moment your business starts getting clients. Gathering data on customer acquisition cost (CAC), lifetime value (LTV), churn, retention rate, monthly recurring revenue (MRR), and other essential stats will set you on the path of success.

Set a Goal for the Experiment

Your initial question is fundamental, but it's not enough. At this point, you might have a suspicion about customer behavior you'd like to check. First, make an assumption. Then, as a result of the study, you'll prove or reject it.

Suppose you propose a hypothesis that most Facebook customers that come to your site eventually choose the most profitable, middle pricing tier. You need to prove this statement. In case it's true, setting a new Facebook advertising strategy will increase customers' numbers from the platform. As a result, the new ads will have a massive impact on your bottom line.

This position will start your adventure in the world of data observation. Here, you'll find out a few exciting things about your business.

Conduct the Experiment

Now that you know the goal, set up the spreadsheet. Combine the cohort data with the setting you want to check and put it into a time perspective. Once you do that, the charts will light up with useful information.

Suppose your data shows that 60% of users from the January Facebook cohort signed up for the middle tier. You also see a similar picture from the February, March, and April cohorts coming from the platform. This information is enough to make an extra effort towards your social media ads strategy.

Although you'll be increasing the expenses related to digital marketing, the results won't disappoint. Your efforts on social media will start to realize their full potential.

Sum up the Results

Reviewing and reporting the analysis outcomes is as crucial as asking the right questions in the beginning. No matter if your hypothesis managed to hold up or not, summarizing the result and acting upon it is a step in the right direction.

Revisit the Proposal in a Few Months

Let's suppose that the previous example worked, and you changed the strategy. However, in a few months, you notice that the percentage of users signing up from new social media cohorts is dropping. Should you continue spending as much money on the ads as you've been doing?

The decision-making process based on cohort analysis is cyclical. What worked a few months ago might not work during the current period. At this point, it would be wise to go back and repeat the experiment.

Some outcomes can also be seasonal, so comparing this year's results to the previous one is useful, too.

Build the Next Experiment

Let's assume the previous experiment went well at first. However, in the long run, the bottom line failed to behave as well as you thought it would. You suspect that customers in the existing Facebook-specific cohorts drop the service after a couple of months.

What should you do to find out the answer? Conduct another experiment, of course. You have two options:

Compare the monthly churn in the Facebook + middle-tier cohorts
Analyze the monthly churn across 'pure Facebook' cohorts without the added tier characteristics

The first choice can show you how many people stopped using the middle tier. The second option will contain a larger, more strategic view of the situation. The choice depends on your intent.

This analysis will answer the vital question: 'Is social media as central for the bottom line as we originally thought?' Opening these types of strategic conversations within the team leads to better decisions in the future.

Conclusion

Keeping an eye on all the metrics can be challenging enough, let alone conducting a few cohort analyses. This multi-faceted, meticulous operation is an exceptional decision-making tool. It equips you with actionable insights about the possible developments in your business.

As we already established, having a solid statistics foundation is a must. Tracking SaaS metrics is the first step on your journey to better analyses. However, gathering metrics can't be enough. The second most important thing is to keep data interactive. Actionable and systematic statistical points are always helpful.

Another crucial thing about cohort analysis is that you need the data points over time. Backing up the company history in numbers will go the extra mile. A well-organized spreadsheet with filters and methodical data points puts the whole story right before your eyes.

However daunting, cohort analysis is always beneficial if you approach it with a goal in mind. Also, the payback is too generous to pass on this opportunity. All the precious data can turn into better operations, more revenue, increased customer retention, and new bright ideas for the future of the business.

Is it worth spending time on cohort analysis? Let us know what you think. If want to know more about buzzwords from SaaS universe, visit the SaaSpedia.