Discover how cohorts allow you to segment and better understand your customers, increasing value and loyalty in B2B e-commerce and SaaS.
Cohorts are a key concept that has its origin in epidemiological research, and they refer to a group of individuals who share a common characteristic during a specific period. For example, a cohort may consist of people born in the same year, residents of the same geographic area, or individuals who began medical treatment at the same time. This approach allows researcherss observe and analyze how different factors, both environmental and lifestyle, influence the health and behavior of these groups over time.
In the context of e-commerce or B2B SaaS businesses, cohort analysis is used to study customer behavior over time and understand how their actions change from their first interaction with our business. A cohort in this context is defined, for example, as a group of customers who made their first purchase in the same month. By following the behavior of this cohort, analysts can observe buying patterns, retention and loyalty rates, and the effectiveness of marketing strategies over time.
Cohort analysis in business allows companies identify crucial trends and areas for improvement. For example, by looking at how a specific cohort behaves in terms of purchase frequency, average spend, and abandonment rate, a company can detect if certain changes to its website, promotions, or marketing campaigns have had the desired impact. If a cohort shows a high abandonment rate after the first month, this could indicate the need to improve the post-purchase experience or implement more effective retention strategies, such as loyalty programs or re-engagement campaigns.
In addition, this type of analysis can be critical to personalizing the customer experience. By understanding how different cohorts respond to different marketing stimuli or product changes, companies can better segment your audience and offer more relevant promotions and recommendations. For example, if a cohort that entered the site through a social media promotion shows greater interest in specific products, the company may direct similar marketing campaigns to future customers with similar profiles and behaviors.
It follows from the above that a cohort analysis does not necessarily have to be carried out at a general level, but rather that customers can be segmented to create and analyze cohorts of different types. Thus, for example, cohort segmentation may involve dividing customers into specific groups based on certain characteristics based on a temporal pattern that is usually months. Here are some common ways to segment cohorts.
1. Acquisition Channel
Segmenting cohorts according to the channel through which customers were acquired (such as social networks, organic searches, email campaigns, paid advertising, etc.) helps evaluate the effectiveness of each channel. This allows you to adjust marketing investments and optimize strategies for each channel. For example, if customers acquired through paid advertising show a higher retention rate, you can consider increasing the budget on that channel.
2. Type of Product Purchased
Segmenting cohorts according to the type of products purchased allows us to understand customer preferences and behaviors in relation to different product categories. For example, customers who buy luxury products may have different buying behaviors and needs than those who buy everyday products. This segmentation helps personalize the shopping experience and marketing strategies for each group.
3. Geography
The geographical segmentation of the cohorts allows us to analyze differences in customer behavior depending on their location. Purchasing preferences and patterns can vary significantly between different regions, so understanding these differences is crucial for effectively personalizing offers and marketing campaigns.
4. Value of the First Purchase
Dividing cohorts based on the value of the first purchase can provide insights into how the initial value spent relates to retention and LTV. For example, customers who make a high-value purchase initially may have different buying behavior compared to those who start with lower-value purchases. This analysis helps identify more valuable customer segments and develop strategies to retain them.
5. Browsing Behavior
Segmenting cohorts based on browsing behavior on the website (for example, time on site, number of pages visited) can provide information about initial engagement. Customers who show a high level of interaction with the site early on may be more likely to become repeat customers.
Like these, other relevant segmentations may exist... By using these and other forms of segmentation, companies can gain a deeper view of their customers, improve retention, optimize marketing strategies, and ultimately increase sales and customer satisfaction.
How a cohort analysis is constructed.
To make the exercise easier, we're going to use a standard template that you can download hither.
- The first thing will be Popular the template with your business details. As you can see in the template, it starts with cell “B4”, which must be filled in with the number of new customers who made their first purchase that month. In our example, we are talking about the month of January 2022 and the number of new customers that month was 43.
- Then we're going to fill in cell “B5” with the number of new customers who placed their first orders in the month of February 2022 and so on.
- Then we completed column “C” starting with cell “C4” with the number of orders placed by new customers who purchased for the first time in January 2022. Note that the number of orders this month must be at least equal to the number of new customers.
- The “D” column we will complete it with the number of orders placed during the month of February 2022 by those who became customers for the first time in January 2022.
- The “E” column we will complete it with the number of orders placed during March 2022 by customers who purchased for the first time in January 2022. And so on.
In this way, row 4 will represent the orders placed by customers who purchased for the first time in January 2022 over the months up to the month of the study. And since we want to estimate the number of times a customer will buy over their lifetime as such, we will extend the analysis a number of additional months in order to be able to make projections. How many months? We'll explain it later.
Once we have “populated” the template with the order data as explained, we will review the table below, which will be automatically filled in based on the data we have used.
- In this way, the column “C” will show the number of times that customers who placed their first order in the month of January 2022 have purchased that same month. In this case, 1.05 means that only 5% of customers who bought for the first time that month made more than one purchase that same month. And the data in cell “G31”, i.e. 0.15, means that only 15% of customers who made their first purchase during the month of June 2022, have placed an order again during the month of November 2022 (M+5). In this way, if we average all the values in column “C” we get 1.04 which means that out of every 100 new customers we can expect 104 orders in the first month.
- AND the “D” column” means that out of every 100 new ones, we can expect them to place 67 orders the following month. Thus, if we add row 47 using the information we have (that is, without adding any projection), that sum gives us 6.49, which represents the number of orders placed on average by each new customer.
But what happens from there? Well, to know this we have to make a “projection” and for this we are going to resort to representing the data we know in a graph, adding a logarithmic trend line, which is the mathematical function that best represents the evolution of customer behavior over time, and from there we project until the function becomes “0”, that is, until customers stop buying. In our case, this will happen in month 45. If we add up all these values, i.e. more projected reais, the sum will be 8.45, meaning that for each new customer we can expect them to place that number of orders over their useful life with our company, which on average will be 45 months.
To complete the analysis with more valuable data, we multiplied the average order value by this number and we would obtain the LTV or “Lifetime Value” of a customer.
And voila! This is done and this is what a cohort analysis is for.
