When was the last time you were in a maze, trying to figure your way out by taking lots of wrong turns on the way? Looking at things from the ground level makes it impossible to properly assess where you are and the best route from A to B, right?
But what if you could zoom out in order to zoom in? What if you could step back and have a birdseye view of the bigger picture in order to make sense of it all?
When it comes to data analysis, there’s a common misconception among marketers that the more reports they look at — the more extensive their insights would be.
In reality, though, it’s the opposite that’s true. Data can be deceiving when it’s too broad.
On the flip side — clearing out the noise, setting up the right dimensions, aligning the data correctly and slicing-and-dicing it accordingly — will help you gain a clear and accurate view of what’s working and what’s not.
Most importantly, though, looking at the data in the right context is what matters most when it comes to your marketing activities.
And this is where cohort analysis comes in.
It allows you to find the most efficient route from A to B. How?
By looking at your marketing activities as a whole, while cross-referencing performance metrics that enable meaningful trends to rise to the surface. Trends that would otherwise go unnoticed, because up till now you’ve been looking at your data out of context.
So what is cohort analysis anyway, and why should you care?
When we say cohort analysis we refer to the process of taking a large group of users (or customers) and breaking these groups down into smaller segments that have certain characteristics in common over a set period of time.
These characteristics could include a common region or language, preferred items, acquisition date, etc.
In the realm of mobile apps, cohort analysis is a highly effective way for gaining deeper insights into how specific groups of users engage with your app over time, and even refining KPIs when measuring campaign performance.
Prime use cases of cohort analysis include:
- Optimizing user acquisition (UA) campaigns – pinpointing which segments are underperforming and where corrective action is needed.
- Improving user retention and lifetime value (LTV) – reengaging high quality users and scaling their base.
A huge advantage of cohort analysis is the fact it enables you to compare “apples to apples” when it comes to identifying trends, making it a reliable way to monitor changes over time, without having to perform behavioral analysis on an individual user basis.
It allows you to answer pivotal questions such as:
- Who’s engaging with your app?
- When do users usually churn — and why?
- How much of your revenue comes from new vs long-time users?
- When is the best time to re-engage your users?
- Which media source drove the most ROI-positive users?
Double whammy – The two main models of cohorts
UA and behavioral cohorts are the most commonly used cohort types. What do each actually mean and how can they apply to your day-to-day as a marketer? Let’s explore further:
User acquisition (UA) cohorts
Acquisition cohorts analysis is usually the bread and butter of advertisers and UA managers.
This is where users are segmented based on acquisition date and source, which — depending on your product and analysis needs — you can run on a daily, weekly, or monthly basis.
For example, you can create several cohorts based on different regions, and compare the average number of sessions per user for each cohort, over the first 30 days of each user’s activity.
How might you leverage this type of cohort analysis?
Let’s say you’re the proud UA manager at an eCommerce app. You can use cohort analysis to track the style and size preferences of different demographics of users and build the inventory accordingly to maximize sales.
Similarly, it can also help you create targeted recommendations and personalized promotional offers in an effort to improve customer trust and build loyalty.
Unlike acquisition analysis, behavioral cohorts are usually the focus of product managers.
This type of cohort analysis focuses on users’ actions while using in your app. Putting special event triggers to use, you can gauge the behavior of demographically different users, and optimize and personalize your campaigns accordingly.
For example, let’s say you have a food delivery app. A good example for behavioral triggers would be the choice of cuisine and ordering frequency.
Or, if you have a social media platform — behavioral cohorts could help you gauge the most followed pages by your Spanish users, or the most liked posts by your Indian user base.
But still.. Why should you care?
If we boil it down to a key win, cohorted data allows you to connect acquisition dates and remarketing conversions with your performance metrics.
In other words, it helps you to drown out the noise, compare apples to apples, produce actionable insights in the battle against churn, and lock sight on constantly moving targets like campaign performance, app engagement, and even feature adoption.
And if you’re still not entirely convinced, let’s explore the curious case of the average revenue per user, aka ARPU.
It’s not enough to just compare one day’s ARPU to the next, as we need to account for all the noise and filter it out: compare ARPU on similar days of the week, compare the number of users, activity, and whether or not there was a promotion that could have driven more in-app purchases.
So, when it comes to holding KPIs up side-by-side and assessing strategic success, cohort analysis is your go-to tool. But as we’ve already discovered, this is not just about analyzing KPIs.
By uncovering hidden trends, cohort analysis provides the insights you need to optimize live campaigns and introduce real-time changes. It’s a proven method to boost engagement and revenue, helping you identify the right path to scale adoption and solidify retention.
How can cohort analysis help you up your marketing game?
Great question. Here are 4 practical instances where cohort analysis could come in extremely handy:
1 – It helps you maximize your ad spend by accessing actionable insights
Cohort analysis allows you to figure out which channels, campaigns, or even ad sets within a campaign — are yielding the best and most loyal users.
For example, looking at the cross-regional performance of a specific UA campaign, you can easily identify that a certain region is under-performing.
You can then try out a few tactics to improve conversion in that region, like personalized emails that include a special offer or a coupon code in honor of a regional holiday — to encourage the completion of a purchase.
While running cohort analysis on the same UA campaign, you could find that in another region, although there are a decent amount of app installs, there are practically no app launches and no purchases made at all. This might indicate fraud, which will require running a deeper analysis using your MMP.
2 – It allows you to get insight into the best time to re-engage
Leveraging cohort analysis, you can determine that user spending plateaus around day 4 in a certain region, which can then prompt you to schedule a remarketing campaign around that time.
For example, seasonality and timing. Being able to acquire more valuable users by answering questions such as “At which month, day or hour of the day is it best for me to engage my users?”.
There are cases in which hourly granularity matters, and cohort analysis can help derive extremely valuable insights when it comes to superbly-timed campaigns.
3 – It enables you to identify and eliminate sticking points in your user journey to boost retention
Put cohort analysis to use in order to uncover pain points in your onboarding, during first steps, on paywalls, or challenging levels — by cross-referencing these churn-prone areas of your app with monthly user retention.
To learn more about the science and art behind a bullet-proof app engagement and user retention strategy – see here.
4 – Understand how product changes affect your users
Figuring out how to fix an issue or improve a feature can often be just as difficult as diagnosing it in the first place.
If, for example, you know that user engagement depends heavily on using a core feature, trying to force-feed your customers into engagement with relentless emails and push notifications — is very likely to lead to churn.
So, instead of jumping the gun on big product changes, A/B test modifications on your cohorts to get an idea of what works and what doesn’t, which will allow you to safely introduce data-driven changes to your product.
This is how we do it – 3 everyday use cases of cohort analysis
Example #1 – Assessing the success of a multi-regional remarketing campaign
Campaigns can be localized to a specific region to provide the most personalized experience for the user. For example, a weekend-themed shopping campaign may run in the US on a Friday afternoon, but on a Thursday in Egypt (where weekends start on a Friday).
When it comes to shopping, remarketing campaigns usually include KPIs such as number of users who made an in-app purchase, purchase revenue, or percentage of re-engaged users. But, to gain a full view of user behavior, KPIs, and success metrics — cohort reports are there to provide you with a multi-dimensional analysis.
Let’s break down the following scenario:
A shopping app ran a remarketing campaign across multiple English-speaking countries, and the UA manager wants to find out how many of those users actually completed a purchase.
He or she would then group results by country in order to assess which region the campaign was most successful in.
Let’s say that roughly 80% of the 14,000 users who re-engaged with the app after the campaign were in the US, and these users also made the most purchases.
However, a closer look at the cohort report brought up an interesting trend. While users from Canada showed half the number of engagements that users in Australia did, Canadian users spent far more money than their Australian mates.
So, by looking at the same data set from a couple of different angles, the UA manager was able to identify that while Australia delivered in quantity, it didn’t deliver much in revenue. Knowing that will enable him or her to double down on the Canadian campaign with confidence.
Example #2 – Assessing the ROI of media sources for ad revenue
Providing insights on all revenue streams, cohort analysis can answer questions such as “Which media source drove the most in-app purchases?” or “Which campaign produced the highest revenue per user in Mexico?” and so on.
Let’s say you’re the PPC manager of a hyper-casual game that relies on ads as their main revenue stream, and need to gather metrics on which media sources are driving the most ad revenue.
The ad revenue report will provide you with a full view of the ROI of acquiring new users (all numbers are hypothetical):
- $88,594 was spent on advertising on Media Source 1, acquiring over 586k users, showing a slow and steady incline.
- Media Source 2, on the other hand, was significantly cheaper in terms of UA while bringing in more revenue from ads than Media Source 1. In fact, by Day 7 post-install, Media Source 2 is much closer to breaking even than Media Source 1.
- Switching the view, you’ll see that Media Source 2’s break-even point occurs around Day 12, whereas Media Source 1 is still far from it.
In this case, cohort analysis will show you that no matter how much you zoom out, Media Source 1 will still perform at a loss, and help you better allocate your precious budget accordingly.
Example #3 – Assessing retention according to an in-app event
Apps that provide an on-demand service, such as ride-hailing or food delivery, often have success KPIs that relate to in-app events. Downloads are a critical component, of course, but actually using the service is far more indicative of successful conversion and retention.
To provide marketers with insight on their app’s ability to maintain an active user-base, cohort’s “sister” report, retention, also gives an overarching view from engagement to use, indicating which media sources drive more engaged users over time.
For example, say you’re the marketing manager at a taxi booking app and want to assess the success of a certain campaign based on how many new users booked a ride after downloading the app.
If you only analyze the number of downloads, it will appear that Media Source 1 is the clear winner.
But, by looking at the data all the way up to Day 10 post-install, it’ll be evident that in terms of absolute numbers, Media Source 2 brings in higher-value users and raised retention rates.
So, analyzing user behavior on install day and overtime, will provide a clearer image of the retention success of that campaign per media source.
4 steps to ensure your next cohort analysis is a raging success
Step 1: Determine the right set of queries
Defining your KPIs and success metrics will steer you in the right direction, so before anything else — let’s establish what it is you’re looking to find out.
Are you after measuring campaigns side-by-side to compare media sources? Or measuring the success of the same campaign across different regions? How do you define success, anyway?
Be sure to nail these questions down before diving into the next step.
Step 2: Define your metrics
As soon as you have solid ideas around the questions you need answered and the metrics you’ll need in order to answer them, you’re in fact halfway done.
Here’s a handy formula for cohort analysis questions: Group users with similar characteristics to compare their behavior and metrics over a specific time-frame.
|Country||Sessions per day||Last month|
|Media source||Revenue per user||3 weeks ago|
|Ad campaign||In-app events per user||5 days in June|
The characteristics are the dimensions by which you will be measuring your outcome, the KPIs are the actual metric you’ll be analyzing, and the time-frame will set the window for your measurement.
To learn more about app marketing KPIs – see here.
Step 3: Define the specific cohorts
This is where we get down to business.
Now’s the time to:
- Set up your granularity, date range, and attribution period
- Select the cohort type (e.g. UA, remarketing), and trend type (e.g. LTV)
- Select your grouping dimension (e.g. media source), and minimum cohort size — to avoids cluttering your report with insignificant data
- Select any additional filters that might be of relevance (e.g. campaign, attribution touch type, country, keywords, etc.)
Step 4: Run a cohort analysis & evaluate results
Go, go, go!
Keep in mind that looking at siloed reports cannot give you the full picture or allow you to draw actionable insights, so don’t forget to slice and dice your data to unearth interesting trends, like app downloads vs. ROAS across various regions.
To dive a little deeper into how to set up a cohort report in your AppsFlyer dashboard – see here.
- When we say cohort analysis we refer to the process of taking a large group of users (or customers) and breaking these groups down into smaller segments that have certain characteristics in common over a set period of time.
- Cohort analysis is a highly effective way for gaining deeper insights into how specific groups of users engage with your app over time, and even refine KPIs when measuring campaign performance, without having to perform behavioral analysis on an individual user basis.
- Prime use cases of cohort analysis include:
- Optimizing UA campaigns – pinpointing which segments are underperforming and where corrective action is needed.
- Improving user retention and LTV – re-engaging high quality users and scaling their base.
- No matter which vertical you’re in, you can use cohort analysis to answer your UA and retention questions. Knowing what to look for and how to look at it, are key to digesting the extensive data displayed on these reports.
- Cohort analysis can help you determine the KPIs of your future campaigns, as well as set your company’s benchmark for campaign success.