Chocolate Factory in the Rhythm of the Season | Module 2 of 9

The Shelf Space:
Delisting arrives quietly

Nobody delists overnight. Rotation per shelf metre falls, shelf facings shrink, a competitor puts a new item next to yours. Sales hears about it in the annual meeting. An early warning model sees it weeks earlier in the sell-out data.

Model calculation - fictitious sample company

This module is part of a nine-part model calculation on a fictitious mid-sized chocolate manufacturer (€95 million net revenue, 140 items, 4 lines). All euro amounts are derived from the documented assumptions of the sample company. They are not results of real customers and not a performance promise. Which values are achievable on your data is shown only by a pilot.

01 The problem - the erosion of a listing is visible in the data, nobody is looking

Marginal listings vanish gradually, and re-entry takes months

The sample company supplies 12 retail customers with 140 items. Not every item-chain pair is a core product; many are marginal listings that fall first in a range review. That is exactly where erosion starts: sell-out drops relative to the category while the market grows.

The signals lie in data you already buy or receive: sell-out per item and chain, category index, shelf audits. A model reads them as an early warning instead of a retrospective.

One endangered marginal listing in the sample company: €300,000 annual revenue

With 4 endangered marginal listings per year, €1,200,000 of revenue is at stake - an assumption of the sample company.

Sell-out per chain
Category index
Anomaly detection
Early warning
Conversation before the decision

02 The model - relative rotation, not absolute quantity

Anomaly detection on the ratio of item sell-out to category, trend over 8 and 13 weeks, facing change

▸ Example values - illustration of the method, not an analysis carried out
Item-chain pairs with warning: 9 of 640
False alarm rate on the pre-period: 4.7 %
One item in one chain loses against the category - warning in week 12 (example values)
↳ The threshold is fixed beforehand

An early warning model is only as good as its false alarm rate. That is why the warning threshold is fixed before deployment and measured on a pre-period in which nothing happened. Only then may the model warn in live operation. A warning is not a verdict on the chain but an occasion for a conversation with category management while there is still time.

03 Business impact - retained listings at contribution margin

The lever: a share of the endangered marginal listings stays on the shelf

€108,000
Retained contribution margin / year
€360,000
Retained revenue / year
30 %
Retained listings
From endangered revenue to retained contribution margin
Model calculation · How the amount arises - derived in the open

Every assumption comes from the sample company and is stored centrally. With your figures the input changes, not the method.

PositionValue
Endangered marginal listings / year × annual revenue per listing4 × €300,000
Base: Endangered revenue / year€1,200,000
Lever: Share retained through early warning (assumption)30 % → €360,000
Contribution margin on net revenue30 %
Result: savings / year€108,000

Assumptions of a sample company - in a real project your data replaces these values.

↳ Defence is cheaper than reconquest

A retained shelf space costs a conversation and perhaps a promotion. A reconquered one costs listing fees, secondary placements and half a year. The value of 30 % retained listings is an assumption of the sample company; how many warnings lead to a conversation at your company and how many conversations lead to a retained listing is shown only by the pilot. Promotions to secure a listing are not counted in module 03.

04 Next steps in your factory

From the warning list to the meeting

① Bundle sell-out data

Sell-out per item and chain, category index, shelf audits - three sources that today sit in three departments.

② Pilot with pre-period

Fix the threshold, measure the false alarm rate on a quiet pre-period, then check the last two years in hindsight: which delisting would the model have seen?

③ Weekly list for sales

Every Monday a short list: "Item X in chain Y has been losing against the category for eight weeks - conversation recommended."

All 9 modules: AI in the chocolate factory