From Inventory Data to Dynamic Store Ranking

Simply comparing absolute shrinkage would not be very meaningful. A large, high-revenue store will naturally show a higher absolute shrinkage than a small location.

Conversely, a small store may be far more conspicuous in relation to its sales or sales area, despite a lower absolute loss.

The goal was therefore an evaluation logic that combines several perspectives and thereby enables a more differentiated assessment of the stores.

A special feature of the model was the dynamic calculation of the national average.

At the beginning of the financial year, the comparison group consisted only of the stores whose inventory had already been completed.

Whenever another inventory was added, the average values were recalculated automatically. As a result, the benchmark itself changed over the course of the year.

The ranking was therefore not a static snapshot, but evolved with the actual inventory progress of the financial year.

For each of the four categories, the model determined how strongly a store deviated from the current national average.

The stores were then scored within each category according to their results. The maximum achievable score per category corresponded to the number of inventories completed at that point in time.

For example: with 18 inventories completed, the best store in a category could receive a maximum of 18 points. Once 51 inventories had been completed, the maximum rose to 51 points. The weakest store received one point.

The points from the four categories were then added up to a total score, which automatically produced the overall ranking — the higher the total, the better the placement.

Alongside the current overall rank, the previous year's placement was also shown.

This made it possible to see how a store had developed compared to the previous financial year.

The ranking could therefore not only highlight current anomalies, but also make positive or negative developments of individual locations visible.

The goal was not simply to produce a list of good and bad stores.

Rather, the model was intended to make anomalies visible and to identify stores where a more in-depth root cause analysis or operational support could be worthwhile.

In particular, the combination of percentage shrinkage, absolute shrinkage, sales area and revenue also brought smaller locations into view — stores that might not have stood out in a purely absolute comparison.

This created a data-driven basis for prioritising stores and for deriving further analyses and actions in a targeted way.

The evaluation was built so that new controlling and quarterly data could be pasted into a separate worksheet.

The relevant KPIs, reference values, points and placements then updated automatically.

As a result, the same logic could be reused throughout the entire financial year without rebuilding the ranking manually after every new inventory.

I initially developed the model as an approach for a nationwide ranking. Parts of the underlying ranking and evaluation logic were used in my own region and occasionally in further regions.

This case study illustrates the approach as an example of the way I work and of my analytical and development skills.