Statistically insignificant, financially enormous: How we stopped guessing about Germany

At Tuinmaximaal, we sell outdoor living structures: verandas, glass sliding doors, sun shading, etc.  These are considered purchased with very long conversion cycles and order values ranging anywhere from €800 to €25,000. That combination is catastrophic for marketing measurement. A single sale gets spread across many touchpoints over months, spend made this week takes months to surface in sales, if at all, and getting a significant result might take months to accumulate the necessary data.

So we did what companies in this position do. We used the data we had and filled the rest with experience and calculated guesses. Given the measurement problem, that was defensible.

We wanted to change this. Marketing mix modelling was the obvious route for us, but this was waved off long before we even suggested it. We opened Germany in three waves, years apart. Spend rose sharply in regions with almost no brand awareness while mature regions lost budget. A national model read that as money going in and nothing coming out. Unusable.

So we stopped modelling Germany as one market, and built a regional model where each region gets its own attribution. It worked better than expected, but what we did not yet know is that this one decision changed our entire way of working. A new decision loop focused entirely on regional attribution, regional strategies and most importantly: regional incrementality testing.