Customer analysis using data

Customer analytics examines who your customers are, what they buy or use, and the value they bring. The result is a decision: what to offer each customer, where to allocate your budget and which group needs attention.

If you mainly want to see what happened, business intelligence is a better fit. Customer analytics looks for patterns at the individual customer level and is one of our AI applications.

Customer lifetime value (CLV) is the value a customer is expected to bring over the entire relationship. A CLV model has two parts: how much someone spends and how long the relationship lasts.

With a subscription, you can see when someone cancels. In an online shop, you cannot: you never know for certain whether someone is still a customer. The BG/NBD model developed by Fader, Hardie and Lee estimates that probability based on how often and how recently someone bought something. The Gamma-Gamma model estimates spending per purchase.

The simplest approach is RFM: you group customers by recency, frequency and monetary value — how recently, how often and how much they buy. If you want to find groups based on behaviour, you can use a clustering method such as k-means. It does not choose the number of groups for you; you decide that with the marketing team.

In our churn analysis, segmentation was necessary to see patterns by subscription type, length of the relationship and activity. A segment is only useful when someone takes a different action for it.

Churn is the number of customers who stop using your product or service. Churn analysis is customer analytics focused on one question: who leaves, and why? You can read how we approach this in our blog post on reducing churn.

Decide in advance who counts as having left. Many subscribers cancel temporarily to get a discount. We did not count those who returned quickly as having left, so the behaviour behind genuine departures remained visible. You can also read the DPG Media case study.

Purchases or usage that you can link to a customer: online shop orders, purchases made with a loyalty card, subscriptions, CRM data, and website and app usage. The more of these sources share the same customer identifier, the less manual work it takes to connect them.

In a shop without a loyalty card, you cannot link a purchase to an individual. Market basket analysis still works: it shows which products are often bought together. Read more in our blog post on AI in retail.

Grouping customers by behaviour or preferences is profiling. The European Data Protection Board (EDPB) describes it as the automated analysis of personal data to evaluate personal aspects, such as behaviour, preferences and interests. You need a purpose and a legal basis under the GDPR to do this.

You cannot simply use data collected for another purpose for advertising. According to the European Commission, customers can always object to direct marketing. Your data protection officer assesses the legal basis; we work with as little personal data as possible.

Preferably, it runs in the environment you already have. As a Databricks and Microsoft partner, we build on Databricks or Microsoft Fabric. For our churn analysis, we used dbt on Snowflake. Read more about choosing a platform on our data and AI platform page.

A one-off analysis produces segments. If customer lifetime value needs to be recalculated regularly, it becomes a model that needs monitoring and someone to own it. That could be your own team, or we can handle it through management and continuous improvement.

When you have a handful of major customers you know personally, when purchases cannot be linked to a customer, or when no one will use the results. Analysis without action only produces a report.

If you mainly want to track how revenue and customer numbers develop, Power BI dashboards are often enough. If you do not know where your customer data is, start with a maturity assessment. This gave Nationale Opera & Ballet a growth plan that links data to full auditoriums and a broader audience.