Table of Content
The enormous volume of user behaviour data generated in different domains typically provides data analysts with new uncharted opportunities to mine valuable insights into user behaviour. Likewise, mobile marketing professionals can tap into the same opportunities.
In an age where most mobile user activities are electronically recorded, mobile marketers can adopt unique techniques like cohort analysis to gain critical insight into user behaviour that can be leveraged. Today’s blog piece aims to expand on this interesting technique and its possible application in mobile marketing. So, let’s dig in!
Image Credit: deep6.ai
What Is a Cohort?
Generally speaking, a cohort is a collection or group of people who tend to share statistical characteristics, like graduation date, birth month, experiences, hiring date, and so forth.
An illustration of a Cohort: image credit: webengage.com
What Is Cohort Analysis?
In essence, cohort analysis is a subset of behavioural analytics that basically aims to group customers into ‘cohorts’, instead of viewing them as a single unit. Fundamentally, as a data analysis technique, cohort analysis seeks to discover user behavioural patterns and trends hidden in time series, operating over a certain period of time, or in a certain location.
Cohort analysis originated from social science and fits suitably into the analysis of user behaviour by collectively exploring the impact of two factors: social change and age, which are highly considered to be the major source affecting user behaviour.
Image Credit: Pinterest.com
So, basically, the “analysis” is how cohorts are tracked, and actions are taken based on their subsequent activity. However, it is noteworthy that this type of analysis is at times seen as part of a more general statistical technique referred to as a cohort study. For better context, cohort analysis is employed in business intelligence, technology and Big Data, while cohort studies are utilised in medicine, epidemiology, psychology, and sociology.
In practice, cohort analysis allows for the comparison of one cohort to another in order to observe actionable differences. For example, if new customers in an October 2020 cohort had a higher average order value than the November 2020 Cohort. Alternatively, a single cohort can be observed over time (separate from all subjects) utilising a particular metric (such as sales volume to customers in the cohort, the number of visitors in the cohort) to gain better insights, for instance, new monthly subscription customers in the February 2020 cohort cancelled their subscriptions after six months at a higher rate than in the first six months.
So, for the most part, cohort analysis enables organisations to monitor patterns at disparate stages of the customer life-cycle, instead of parsing them across all customers blindly, without regard to the natural cycle a customer is in. Consequently, by critically observing behavioural patterns across cohorts, organisations can better customise their businesses to those specific customer groups.
How to run a cohort analysis. Image Credit: improvado.com
Types of Cohort Analysis
Generally, there are two main types of cohort analysis:
- Acquisition Cohorts: This type of cohort analysis segments users based on when they were attained or signed up for a product. So, depending on one’s product, user acquisition could be effectively tracked daily, weekly, or monthly. For instance, a mobile app for daily productivity can track its acquisition cohorts on a daily basis. However, a B2B mobile app with a more focused user group would primarily focus on monthly acquisition.
- Behavioural Cohorts: Behavioural cohort analysis groups users based on the activities they undertake within the mobile app during a specific period of time.
For instance, mobile users who share photos employing a Google Photo link on a given day. The time period, again, differs from app to app. For example, for a photo-sharing mobile app, a day could be a good timeframe. However, for an investment platform mobile app, months would be better for observing user behaviour.
How Cohort Analysis Works
Here is a more straightforward illustration of how cohort analysis works. Envision you employed two sales teams, one at the start of 2020 and another at the start of 2021. How would you compare the performance of your 2020 team against that of your 2021 team? You can consider performing a direct comparison of their current performance, however, it would not be highly accurate or fair, since the veterans 2020 team would be at a clear advantage since they were there longer.
However, if you split the two teams into two cohorts based on their date of hire, then tracked their first-year performances, you would be able to achieve an accurate comparison of how each team stacks up, by contrasting their respective first years. As such, you would be able to ascertain which team closed deals faster, or found more leads etc. But, more importantly, you’d be able to clearly zero in on each team’s performance spikes, and thus gain actionable insights into the factors that may have helped or hurt their respective performances.
Image Credit: clevertap.com
So, in practice, cohort analysis can be used to split mobile users into groups based on their birth dates (typically, the first time they signed up). For example,
- Cohort A – Mobile users who first signed up on November 11
- Cohort B – Mobile users who first signed up on December 11
Also, cohort analysis tends to reflect distinctions between user groups based on their relational progress. For instance, you can learn from a cohort analysis:
- The average conversion rate for November 11 users versus December 11 users.
- The amount of time it averagely took November 11 users to convert versus December 11 users.
So, to produce a cohort analysis, one should always record the life-cycle change events of customers and perform the analysis based on that.
Why Cohort Analysis?
Cohort Analysis can be employed in e-commerce to understand customer behaviour, specifically, customer engagement. Some of the clear benefits of running a cohort analysis are:
- It helps identify users who are ready to become customers
- It helps identify users who can defect or those that can be retained, i.e., chalking or predicting future behaviour.
- It helps identify users who could be a target group for cross-selling or up-selling initiatives.
- Helps identify activities, features, or changes that work for customers
- It allows marketers to exploit present data as a benchmark for future actions
- Assists with the planning of user onboarding activities
- Makes sure that marketing activities are data-driven
A Use Case of Cohort Analysis
Cohort analysis is relied on at Amazon to:
- Curating accessories sales: cohort analysis helps Amazon determine the number of customers who buy an accessory/warranty at the time of purchase and how long they take to come back to purchase one/another accessory, especially if they didn’t act at the point of sale. As such, it enables businesses to estimate the number of accessories sold within the device’s life cycle.
- Understand returns and refunds on a specific item: Amazon uses cohort analysis to forecast the expected returns and the timeline of the returns.
- Understand device upgrades: Amazon employs cohort analysis to understand how long customers own a particular device before choosing to upgrade to a newer version, which helps them to forecast the sales of newer device versions.
An illustration of how Amazon might use cohort analysis: Image Credit:chartio.com
Critical Steps Involved in Cohort Analysis
- Defining metrics: Cohort analysis aims to select meaningful indicators to optimise performance, like increasing revenue or reducing customer churn. This essentially necessitates defining metrics to be evaluated, such as customer churn rate, number of purchases, lifetime customer value, etc. Generally, there are two types of metrics that can be distinguished in cohort analysis. Firstly, actionable metrics that link repeatable actions to observed outcomes (for example, registering a user with a subsequent purchase). Such metrics help one to understand the real situation, make decisions, and improve business outcomes. Secondly, vanity metrics that add to a business’s positive image, but don’t help to understand the big picture, such as the number of likes on social networks.
- Forming cohorts: After determining the metrics, you can now group or form the cohorts. One of the key options for grouping into cohorts is the first action a customer takes when contacting a company, like registering on a website, buying or downloading an app.
- Comparing cohorts by metrics: This step of cohort analysis involves discovering distinctions between cohorts, and explaining the patterns of customer behaviour specific to a certain cohort.
How to Use Cohort Analysis for User Retention
Fundamentally, cohort analysis aims to address three basic questions:
- First, who is using the app?
- When are they looking to churn?
- Why are they losing interest?
Generally, cohort analysis seeks to connect all the dots from user behaviour to boost customer retention. Employing cohort analysis for user retention helps one to understand the number of customers who continue to be active users in the days/weeks/months that follow. They can then take the metrics associated with their purchases; then combine them to help infer future data and trends.
Furthermore, cohort analysis also enables marketers to see the effectiveness of when strategies were initiated. They can follow cohorts of customers and their respective life cycles while tracking against programs undertaken at various times for that particular cohort.
For instance, if a specific cohort was a collection of customers who made purchases in November, a marketer can measure their life-cycle, and track it against another marketing program that was initiated for a cohort that began in July.
Using cohort analysis to improve retention and churn rate. Image Credit:baremetrics.com
Based on whatever metric they want to track (for example, Repeat Time to Purchase), they can view the effectiveness of their strategies based on groups of cohorts. As such, a specific cohort can become a baseline for whichever strategy a marketer wants to measure, track improvements or inefficiencies.
Overall, cohort analysis helps organisations to comprehensively understand why, when, and how customers buy things, and why they keep coming back.
Strategies to Try After Being Empowered With Data From Cohort Analysis:
- Cohort analysis can be employed to determine exact junctures in the user journey when the users are skipping out. Thus, the user journey can be tweaked and better streamlined to make users stay longer.
- You can plan reactivation emails by employing cohort analysis for metrics like time intervals between two purchases to help properly plan the reactivation drip campaigns to keep customers in the loop. Thus, gently nudging customers when they are ready for the next purchase.
- You can plan targeted offers, coupons, and free shipping offers from cohort analysis to understand the kinds of customers who buy the most and what they buy the most.
- Curate loyalty programs, points rewards and similar gamification systems with data from cohort analysis that can help to narrow down on the exact customer sets who can be retained longer with loyalty programs.
Conclusion
All things considered, rather than always assuming facts about customers, having decisions backed by data from A/B testing cohorts can help you understand what works, and assist in reducing churn. So, with cohort analytics, mobile marketers can understand where to hit exactly with their customer retention strategies.
It’s wise to remember that the application of cohort analysis isn’t limited to a single industry or vertical. For instance, while eCommerce companies can utilise cohort analysis to spot products with more potential for sales growth, digital marketers can exploit it to identify web pages that perform well based on time spent on websites, conversions or sign-ups.