What is a customer relationship worth, today and in the years still to come?
For HSE (Home Shopping Europe), this was long a question without a reliable answer. How many new customers a campaign brings was always measurable. What those customers would still be worth in future could only be estimated.
HSE brought this question to foobar Agency. The data platform was already in place. What had to be settled was what needed to be built within it, so that the question would turn into a figure solid enough to base budget decisions on.
HSE (Home Shopping Europe) is one of Europe's largest live commerce companies. Its business model combines TV commerce, social commerce and eCommerce across numerous divisions and brands.
The broader this portfolio, the more weight falls on a question that sounds simple: which customer relationships deserve the most attention in the long run, and which marketing spend never pays off?
For the marketing and CRM leads at HSE, this is the basis of every budget and campaign decision.
The marketing team at HSE could say at any time how many new customers had been won. How valuable and how sustainable those relationships were remained open. Until then, no customer lifetime value existed.
That left an entire class of decisions without a foundation: which campaigns bring customers who stay for the long term? Which brands attract particularly valuable customer groups? And where does marketing budget disappear into relationships that never pay off?
For a budget decision, what matters most is what a customer relationship will still carry in future. Looking back at purchases already made does not answer that question, and exactly this forward-looking part was missing.
The task was therefore not one of implementation. Before anything could be calculated, it had to be decided what should be calculated.
Customer value sounds like a number you simply calculate. For a retail group of this breadth it is a decision with consequences: for budgets, for target groups and for the question of which brand contributes what.
In joint working sessions, HSE and foobar Agency answered three questions before the first model was built.
What counts towards value? Which revenues feed in, how returns and marketing costs are treated, which time horizon applies.
At what level is the figure reliable? At the level of individual people, customer value always remains blurred, because single outliers dominate the result. The prediction becomes reliable at the level of customer groups, such as the highest-revenue customers of a brand or all first-time buyers from a campaign.
How do we know it is right? Which data sources hold up, which checks the figure has to pass before budgets depend on it.
The second decision in particular spared the project a great deal. It established what a model of this kind can promise, before expectations of individual customer values arose that no model can meet.
The data platform was a given. The design freedom lay in what would be built within it and how the parts fit together. Three design decisions shaped the result.
History and prediction kept separate. First came the historical customer value as an independent, verifiable figure. On top of it came the actual product: the prediction of future value. This keeps the base usable on its own, and the prediction model can be replaced without touching it.
Operability as a selection criterion. Because the model was to run where the data lives, prediction quality alone was not the measure. The approach established in academic literature was implemented and evaluated, then set aside in favour of a simpler method. That method is easier to operate in this environment, delivers more reliable results and also covers one-time buyers. As in all predictive analytics projects at foobar Agency, development was iterative: first a simple, transparent variant as the benchmark, then targeted refinement measured against exactly that benchmark.
Three routes for three audiences. The figure has to reach different people: the business reporting for ongoing steering, a conversational route for short-notice questions from the leadership circle, and a dedicated monitor for the model itself.
At HSE, marketing budgets can now be steered by the actual value of individual customer groups, down to division and brand level.
One-time buyers are covered for the first time as well. The standard approach delivers no prediction for this group by design, and it carries the greatest uncertainty, because after a single purchase there is little data.
One finding from the analyses is changing the conversation in marketing for good: for a considerable share of customers, future value after deducting marketing costs is negative, partly due to product economics, partly because of the marketing costs alone. A basis for putting budgets where they pay off across entire customer groups.
The real gain has shown itself in operation since. HSE has updated the model several times and improved prediction quality demonstrably, with its own team and on the jointly created foundation. A topic that could not be measured before has become a capability in-house.
What tipped the scales was that both came from one team: the ability to define, together with the business, what should be measured in the first place, and the experience to then anchor and operate the result in an existing platform.
Added to that is the instinct for which approach holds up in practice and which only looks convincing on paper. An instinct that develops across many comparable projects.
Equally decisive: a project team with undivided focus. A question that had been in the air became a model running in production within weeks.
The CLV prediction is a foundation that can be built upon.
As part of the long-standing collaboration with HSE, the model is already being extended with acquisition-specific metrics, to assess the impact of new-customer campaigns more precisely.
Every further stage builds on a data platform designed from the outset for exactly this purpose: to grow along with the company's questions.
A single percentage point of deviation ties up millions in capital. HSE makes returns and cancellations predictable today, with a learning forecast model on Snowflake that has replaced the rulebook grown over years.
Since 2019, foobar Agency has accompanied porta Möbel on its path to digital market leadership: from legacy analysis via incremental headless commerce to a full composable commerce architecture on MACH principles.
We look forward to your enquiry.
Please accept marketing cookies to load the registration form.