Singular’s 365-day cohort reporting: better data science equals better marketing
Cohort reporting is not limited to a month, three months, or even six months any more. Singular now supports a full year: 365-day cohort reporting periods.
In other words, if you’re doing cohort tracking or cohort analyses in verticals that need more data than a D30 retention rate report, you’re in luck. And if your marketing campaign time period is three to nine months, you can now get extra margin and extra insight in your cohort statistics.
Web-based marketers recognize cohort reporting from Google Analytics, where you can see your retention rate and the impact of your marketing efforts on the web. Mobile marketers need the same — in fact more detail — analytics in their mobile marketing reports.
More data makes you smarter. More data means your marketing campaigns bring in more revenue.
In mobile marketing, more data tells you invaluable information such as your customer life cycle. Your average revenue per group of acquired users. Your average sessions per user, by cohort. This goes far beyond vanity metrics and gets to the most important behavioral analytics that define user lifetime value (LTV).
That’s why cohort analysis tools are so vital, and how marketers looking at a cohort table can see trends and create opportunities that, like product improvements, can have cumulative benefits.
I recently took some time to talk about the change with Singular VP of Product Alon Nafta.
John Koetsier: Singular recently updated its cohort lengths to 365 days. But before we talk about why, let’s talk about cohorts. What are the primary reasons to do cohort analysis, and what do marketers learn from them?
Nafta: When we say cohorts it’s important to define first what these are in the context of marketing and user acquisition.
By definition a cohort is a group with shared characteristics. In the context of user acquisition, a cohort often refers to users acquired in similar manner. At the most basic level this could be the date, but it can extend to the marketing channel, the campaign, and even the creative. (Imagine what knowing which creative a cohort responded to could tell you about that group of people.)
A cohort report takes these groups of users, and looks at how they behave over time, say after one day, one week, one month and so on.
By “behave” we often refer to retention or any other important KPI you want to measure your product by. That gives you a very clean view since for example different dates can correspond to different marketing activities or product releases. And different channels, campaigns, or publishers may result in an acquired user profile that can be dramatically different from each other.
So a cohort analysis also helps me establish my baseline — for example, how my organic users are behaving over time — and benchmark acquired users for different campaigns to these profiles.
As a marketer, that can ultimately teach me if my paid acquisition is exhibiting the right results, and point me on what should I focus on when trying to improve. Equally important, it also gives me insight into how fast I’m earning back my acquisition costs, which ultimately is one of the most important things for effective paid marketing.
There’s just so much you can do with cohort analysis. It really is a fundamental tool for the mobile marketer.
John Koetsier: What are the primary ways to define a cohort, and when would you use each? Time of acquisition is of course one … what else is interesting?
Nafta: Interestingly enough, even time of acquisition is not a fully strict definition since acquisition is not just one singular point in time.
For mobile marketing, a common way is to look at the time (or rather, date) of install. But you can also define the starting point of a cohort by the timestamp of the attributed click or impression (AKA critical touchpoint), which tends to be the case for web marketers.
Some marketers, especially in the digital commerce space, may want to look at a different event such as registration or first purchase. They may consider that a much more meaningful and significant starting point in defining a cohort.
Once you define how the cohort is calculated, a good cohort analysis tool or report should give you as many breakdowns as possible to differentiate between important characteristics of these groups of users.
That includes data in a table or in visualizations around:
- How they were acquired
- Key properties of the acquisition campaign … the customer journey
- Whether they were new users or retargeted former users
- Acquisition costs
- What types of post-acquisition activities are they engaging in, including the active user rate
- Conversion rate to purchase or other value-creation activity
- And more, depending on your app, your vertical, and your KPIs
Lastly, you also define how many time units — commonly days — you’re looking at after the defined start time of the cohort, and if you’re measuring accumulatively.
Note, this may differ between different types of activities. For example a 30-day retention cohort would commonly mean how many users came back on the thirtieth day after install (or re-install). But a 30-day purchase sum cohort can either refer to the total number of purchases made in the first thirty days (which is more common), or just the sum on the thirtieth day.
Both are applicable. Which you’re using needs to be determined to understand what are you looking at, and avoid confusion with coworkers.
John Koetsier: Does Singular make cohorts available based on re-engaged and re-attributed users?
Looking at cohorts for re-engaged or re-attributed users just as important as it is for newly acquired users. In fact for some verticals, acquired users in the sense of users who have just installed the app is almost not interesting at all for paid marketing, since almost everything focuses on re-engagement.
John Koetsier: OK, let’s get to the big story. Singular updated cohort periods to enable 365-day cohorts. Why?
Nafta: Well, as explained earlier, you’re trying to look at how different groups of users behave over time and draw conclusions accordingly.
In some cases you can draw conclusions or make some predictions based on a relatively short timeframe. For example, you might be able to predict the life-time value (LTV) of a user in a mobile game based on the first seven days of in-app purchases.
However, for some companies and products, the window of interesting activity that is important for prediction may take a much longer time.
For example, if I do a monthly subscription for my fitness app, I’ll probably need to review at least three to six months to understand how users are engaging, retaining, and upgrading their subscriptions. Or, if it’s a digital commerce product where customers are buying more expensive items that you typically don’t buy daily or weekly, marketers likely need to look at several months, half a year, or even a full year of data to be able to produce high-quality conclusions and predictions.
These numbers are extremely important for data science teams, who are often tasked with modeling LTV as well as LTV prediction. More time allows them to improve their models, test them better against reality, and iterate accordingly.
John Koetsier: What business types or app verticals typically benefit most from longer cohort periods?
Nafta: It really varies but I think it ultimately comes down to the expected activity profile beyond retention for your product. If users are taking actions on a monthly or longer basis, one-year cohorts — and even longer — are extremely important.
We see this for subscription services, digital commerce, fintech, and even gaming (with varying impact from hyper-casual to mid-core to hard-core games). It also depends on the level of sophistication and effort companies can invest.
Longer cohort periods produce more data and can allow better models, if you have the resources in place to take advantage of it.
John Koetsier: Often we look at cohorts individually over time to see return on ad spend (ROAS) for a group of acquired users. What else can you learn by tracking individual cohorts?
Nafta: ROAS is only one metric. It’s very meaningful of course to marketers who are working on paid sources. But cohorts are also meaningful to product managers, since by looking at retention against product release dates I can learn quite a bit about how my releases affect retention and adoption.
It’s also important for understanding seasonality, and many more insights.
A different creative asset, for example, which shows a different item to be purchased, uses a different coupon or offer, or just uses different design — such as more straightforward versus more artful — can say a lot about the types of users you’ve just acquired.
This is important data collection for mobile marketers, and an analysis report with insights here often leads straight to increased revenue.
John Koetsier: As you’ve said, cohort analysis is pretty important for LTV analyses. Does Singular automatically surface LTV and ROI for cohorts?
Nafta: Yes. By default we surface three important KPIs in our cohort report: LTV, ROI, and CPE (cost per event) for every event a marketer has defined as interesting.
Of course, a lot can be customized to meet individual needs. And specifically for retention we have a designated retention report which shows the same cohorts.
John Koetsier: Anything else?
Nafta: At Singular, cohort reporting is at the core. Our philosophy is to ensure that everything can be reported in a cohorted manner.
While reporting by date — which we sometimes refer to as actuals — can give you insight especially in real time, reporting against cohorts is what truly uncovers the outcome of your marketing. This includes being able to attach the cost of acquisition, the type of campaign, ad set and ad, the bid, the strategy, and the type of campaign.
And of course … what all of these are generating in terms of business results.
John Koetsier: Thank you for your time!
Cohort reporting: next steps
Interested in learning more?