How to understand your customers better with cohort analysis

Last time we talked about how to effectively use “indicator metrics” to curb the lagging nature of cohort metrics and make real-time decisions without having to wait for the cohort metric to finish counting. 

This week we are going to talk about cohort analysis’s effectiveness in helping you understand your customer demographics. 

One of the most clear use cases for cohort analysis is generating lookalike audiences. 

You can simply take the emails of a list of top 100 or top 1000 customers for your business in terms of Customer Lifetime Value. 

Then, you can put the list into Facebook or Google for lookalike audience targeting so you can bring in more of those customers through your door with minimal effort and a relatively high return on investment. 

While effective, the key issue I have with those lookalike methods is that they do not help you truly understand who are the primary purchasers of your products, and what are their needs/wants so you can further satisfy their desires. 

To enable those more advanced demographic understandings, we recommend an approach that is both behavioral and technical. 

The behavioral approach

The primary practical purpose of the behavioral approach is to expand our targeting reach to potential interests and demographic groups that we have never thought of before. 

With the extremely expansive targeting options provided by Facebook and Google, it has become increasingly crucial for marketers to test not only 1, but many many different interest options and see which one most produces the best results. 

There is no better way to generate those potential targeting options than interviewing and understanding your best customers. 

While we are not going to go in depth in this article covering a variety of customer research methods that you can use ranging from survey, ethnography, to focus groups, I am going to share the one that we found to be most effective in our experiences. 

What we generally recommend is to reach out to 3-5 of those top customers and simply ask them about their life, what they care about, and the brand they interact with on a daily basis. 

We usually start with a simple question: “Describe yourself for us” and “Tell us about a day in your life from morning to night.”, and ask a lot of followup questions from there. 

One common mistake we found from a lot of the interviews is that companies tend to focus on the belief of the company itself, rather than the belief and need of its customers. 

For example, questions such as “Can you tell us how you interact with our products” is completely useless for targeting expansion purposes, as it gives us no insights on the life of your interviewee. 

Rather, alternative questions like this is better: “When are you using our product, what kind of situation do you usually find yourself in? What are some other brands, people that are present during that use case?”

The insights from those interviews can then be transcribed and distilled into interests in Google and Facebook to be used for targeting purposes by either your marketing team or your agency. 

The technical approach

Whereas the purpose of the behavioral approach is an expansion of your horizon on who you can target, the technical approach enhances your cohort analysis so you can more rapidly filter out the targets/segments that do not work for your company. 

By default, cohort analysis only operates on your primary POS platform, whether that’s Magento or Shopify, but we can connect data from those platforms with data from our advertising platforms like Facebook and Google through some simple steps. 

To achieve this, you will need an intermediary platform that connects your website analytics data (like data from Google Analytics) with your POS system data. 

The most mature version of those intermediary platforms are usually called a Customer Data Platform (CDP), but any platform that has similar functionalities will do for our purpose, and here we recommend Klaviyo. 

While most people think Klaviyo is an email platform for sending and managing email flows, they actually offer a very mature data infrastructure that organizes data from multiple sources into a customer-centric format. 

Within the platform, you can in fact search any customers by email and know their purchase history down to order numbers and order quantities, along with their email opens and email reads. 

What we are going to add to the Klaviyo platform is data from the advertising platforms. 

Through utm tagging, we can recognize and tag a customer if they have visited our ads from any advertising platforms, down to the ad they have seen and the campaign that made them pay the visit. 

Then, we can conduct cohort analysis with campaign id or adset id as a dimension on those customers, and understand very quickly which campaign is producing long-term results and which ones are not. 

Admittedly, I omitted many specific details on how to actually set up the system explained above.  But please be rest assured knowing that they are 100% possible and not too difficult to do, and I will publish a technical guide soon on how to do it step by step. 

Conclusions 

To sum up this article, cohort analysis is an incredibly useful tool in helping us understand our customers and target them more precisely and effectively. 

While lookalike audiences are the most immediate and easy way to leverage your cohort analysis results, a more advanced approach that combines behavioral and quantitative research methods will yield more results in helping you understand your customers. 

In the next article, we are going to dig briefly into the technical rabbithole of the big data side of cohort analysis and give you a brief introduction of tools and concepts that you need to know to process a large amount of cohort data effectively.