Transparency · Chocolate Factory in the Rhythm of the Season

Where the numbers come from:
every assumption open

Complete calculation path for each of the nine modules. Every parameter has a source or is marked as an assumption. The "certainty" column says what is documented, what is estimated and what you should verify in your factory.

How we calculated, and what deliberately not

↳ Four principles

1. Every parameter has a source or is marked as an assumption. 2. No module has independent evidence for its degree of effect; the lever ratios are assumptions of the sample company, oriented at documented bands where they exist. 3. No double counting: write-offs, shelf gaps, rush costs, cocoa costs, scrap, promotion budget and direct sales are each counted exactly once. 4. The total is a gross effect before implementation costs; module 06 stands outside the total.

12
Input parameters
1
From public data
6
Estimate
5
To verify in your factory
Cautious
€704,280
Factor 0,6 on every module
Base (presented)
€1,173,800
1.24 % of net revenue
Extended
€1,643,320
Factor 1,4 on every module

The scenarios are multipliers on the module values of the base calculation, not a second calculation with different assumptions. They show the band in which the model calculation would move under different lever ratios.

Input data - what is documented, what is estimated

The sample company is fictitious. Its orders of magnitude are aligned with the industry frame so that the calculation stays sound.

ParameterValueSourceCertainty
Net revenue per year€95,000,000Freely set; €7.92 per kg against €7.59 per kg in the industry mean (source 1)Estimate
Employees380Freely set; €250,000 revenue per head, industry predominantly mid-sized (source 1)Estimate
Chocolate products per year12,000 tFrom revenue and value per kg; industry mean source 1Documented
Active items140AssumptionEstimate
Production lines4AssumptionEstimate
Material input50 % = €47,500,000Assumption without anchor; cost structure survey not verifiedTo verify
Cocoa products (liquor, butter, powder)3,600 t × €6,000 = €21,600,000Volume about 30 % of weight (milk- and filling-heavy range); blended price assumption with semi-finished mark-up, bean price source 2 is no anchorTo verify
Contribution margin on net revenue30 %AssumptionEstimate
Budget for trade promotions8 % = €7,600,000Assumption; industry estimate for consumer goods worldwide around 20 % (source 9, tier 3), deliberately belowTo verify
Revenue in direct sales3 % = €2,850,000AssumptionEstimate
Write-offs after the best-before date0.6 % = €570,000Assumption; processing carries 19 % of the food waste of the EU chain (source 3), no chocolate rate publishedTo verify
Revenue lost through shelf gaps1.2 % = €1,140,000Assumption; survey finding on shelf gaps (source 4) as a frame, not as a rateTo verify
↳ How you verify

The twelve parameters of this table are the input of the whole series. In a data workshop they are jointly replaced by your actual figures; the calculation paths stay the same. The result is a model calculation on your figures instead of the sample company.

Nine modules, nine open calculations

Click a module for formula, calculation and anchor. All values come from the same configuration as the module pages.

M01The Seasonal Volume
€331,500
Savings = write-offs × reduction + shelf-gap revenue × recovery × contribution margin + rush costs × reduction
ParameterCalculation
Fewer write-offs€570,000 × 30 % = €171,000
Recovered contribution margin from shelf gaps€1,140,000 × 25 % × 30 % = €85,500
Fewer rush costs€250,000 × 30 % = €75,000
Module result€331,500
↳ Anchor and limit

Accuracy: the M5 competition (source 5) documents on retail data that the best learning method forecasts 22.4 % more accurately than the best statistical reference - accuracy, not a euro effect. A consulting study (source 6) cites 20 to 50 % fewer forecast errors; it stands here only next to the documented anchor. The 30 % reduction of write-offs and the 25 % recovery are assumptions. The survey finding on shelf gaps (source 4) provides shopper reactions, not a loss rate.

M02The Shelf Space
€108,000
Retained contribution margin = endangered marginal listings × revenue per listing × retained share × contribution margin
ParameterCalculation
Endangered revenue4 × €300,000 = €1,200,000
Retained revenue30 % → €360,000
Retained contribution margin30 % → €108,000
Module result€108,000
↳ Anchor and limit

An association survey (source 7) reports that a third of brand manufacturers were affected or threatened by delisting - an industry finding without independent review. Number of marginal listings, revenue per listing and retained share are assumptions. Promotions to secure a listing are not part of module 03.

M03The Promotion
€228,000
Better contribution margin = budget × shifted share × (return of target promotions − return of shifted promotions)
ParameterCalculation
Shifted budget€7,600,000 × 25 % = €1,900,000
Contribution margin per budget euro, shifted against target0,9 → 1,02 = +12 percentage points
Return difference on the shifted budget€1,900,000 × 12 % = €228,000
Module result€228,000
↳ Anchor and limit

That a substantial share of trade promotions is unprofitable is peer-reviewed (source 8); a market research study (source 9) puts the share at 59 % and stands here as an industry estimate next to it. Share, base return and target return are assumptions of the sample company; the module is the second-largest item with the weakest derivation.

M04The Changeover Sequence
€162,000
Savings = gained line hours × value per hour + transition scrap × material value × reduction + overtime × reduction
ParameterCalculation
Changeover time and gained line hours900 × 40 min = 600 h; × 20 % = 120 h × €500 = €60,000
Less transition scrap90 t × €4,000 × 20 % = €72,000
Less seasonal overtime€150,000 × 20 % = €30,000
Module result€162,000
↳ Anchor and limit

A peer-reviewed case study from food processing (source 10) reports 34 % less changeover time and 11 % more capacity; a methods paper (source 11) describes the procedure. The 20 % lever lies below. Value per line hour (about a third of the theoretical contribution margin), scrap and overtime are assumptions; part of the transition scrap is reusable.

M05The Remaining Shelf Life
€154,800
Savings = (write-offs − savings module 01) × reduction + returns × reduction
ParameterCalculation
Remaining write-offs after module 01€570,000 − €171,000 = €399,000
Fewer residual write-offs€399,000 × 20 % = €79,800
Fewer returns€250,000 × 30 % = €75,000
Module result€154,800
↳ Anchor and limit

The retail one-third rule and the finding that over three quarters of unsold goods in retail end as waste are institutionally documented (sources 12, 13). Storage temperature as the main driver of fat bloom is peer-reviewed (source 14). The remainder is derived from module 01, never set as a literal; reductions and returns are assumptions.

M06The Cocoa Coverage
€216,000 *
Avoided dispersion = cocoa purchasing × dispersion effect (assumption); not added
ParameterCalculation
Cocoa purchasing3,600 t × €6,000 = €21,600,000
Avoided dispersion€21,600,000 × 1.0 % = €216,000
Series totalwithout this module
Module result (avoided dispersion, not added)€216,000
↳ Anchor and limit

That an optimal coverage ratio can lie below full hedging is peer-reviewed (source 15); a meta-analysis over 1,699 hedge ratios (source 16) shows the band. In-house research (sources 17, 18) is a negative result against price forecasting with an open calculation. The 1.0 % dispersion effect is an assumption without anchor and describes a risk measure, not a cost advantage.

M07The Taste
€80,000
Savings = complaints × reduction + cost of one rework × cases per year
ParameterCalculation
Fewer sensory complaints€80,000 × 25 % = €20,000
Avoided rework€120,000 × 0,5 = €60,000
Module result€80,000
↳ Anchor and limit

The standard for selection, training and monitoring of sensory assessors (source 19) documents the procedure, not the effect. All amounts are assumptions of the sample company. Sensory analysis finds drift, not foreign bodies.

M08The Origin
€52,500
Savings = full-time equivalents × full cost × share of automated review
ParameterCalculation
Review costs today1,5 × €70,000 = €105,000
Automated review€105,000 × 50 % = €52,500
Module result€52,500
↳ Anchor and limit

Obligation and dates are officially documented (source 20): application from 30.12.2026 for large and medium-sized, from 30.06.2027 for small companies. In-house research (sources 21, 22) describes procedure and limits. Review effort and automation share are assumptions; supply risks are deliberately not quantified.

M09The Direct Customer
€57,000
Additional contribution margin = direct sales revenue × effect × direct sales contribution margin
ParameterCalculation
Additional revenue€2,850,000 × 5 % = €142,500
Additional contribution margin€142,500 × 40 % = €57,000
Module result€57,000
↳ Anchor and limit

That targeting by effect beats targeting by risk is peer-reviewed (source 23). The 5 % effect and the contribution margin in direct sales are assumptions of the sample company.

* Module 06 (€216,000) is a risk measure and is not added to the series total.

Three scenarios, eight modules added up

Savings per module - cautious, base, extended (module 06 stands outside)
In the cautious scenario: €704,280 per year

That is 0.74 % of net revenue - as a gross effect before the costs of implementation, which are quantified in the conversation, not on this page.

↳ What this calculation deliberately does not contain

Capital tied up in raw material and seasonal stock, energy for conching, tempering and cooling, sugar, milk powder and packaging are not calculated. The total is a gross effect before implementation costs. No module has independent evidence for its degree of effect; every lever ratio is an assumption that is replaced with your data.

Source list with evidence class

Evidence classes: peer-reviewed (tier 1), institutional (tier 2), industry and consulting studies (tier 3, never stand alone), in-house research with open calculation (tier 4). Retrieved on 27 August 2026.

  • Bundesverband der Deutschen Süßwarenindustrie (BDSI): Confectionery industry in figures 2025 - 275 companies, 78.5 % with up to 249 employees; chocolate products 1.103 million t worth €8.373 billion. Tier 2
  • International Cocoa Organization (ICCO): daily cocoa price of 26 August 2026, 5,896.83 US dollars per tonne; monthly reports 2025. Tier 2
  • Eurostat: Food waste and food waste prevention, reference year 2022 - processing carries 19 % of food waste, 25 kg per head. Tier 2
  • Gruen, Corsten, Bharadwaj (2002): Retail Out-of-Stocks, study for the Grocery Manufacturers of America - shelf gaps worldwide 8.3 %; reactions: 26 % brand switch, 9 % no purchase, 31 % other store, 15 % postponement, 19 % same brand. Tier 3
  • Makridakis, Spiliotis, Assimakopoulos (2022): M5 accuracy competition, International Journal of Forecasting - best learning method 22.4 % more accurate than the best statistical reference on retail data. Tier 1
  • McKinsey & Company: AI-driven operations forecasting in data-light environments - 20 to 50 % fewer forecast errors; stands only next to source 5. Tier 3
  • Markenverband / European brand manufacturer survey (2022): a third of surveyed manufacturers affected or threatened by delisting (cited from business press). Tier 3
  • Ailawadi, Harlam, César, Trounce (2006): Promotion Profitability for a Retailer, Journal of Marketing Research 43 - quantification of the net profit of promotions. Tier 1
  • Nielsen (2015): Trade Promotion Performance - 59 % of promotions without break-even; consumer goods manufacturers invest around 20 % of revenue in trade promotions. Industry estimate next to source 8. Tier 3
  • Maalouf, Zaduminska (2019): A case study of VSM and SMED in the food processing industry, Management and Production Engineering Review - changeover time −34 %, capacity +11 %. Tier 1
  • Lozano, Saenz-Díez, Martínez et al. (2017): Methodology to improve machine changeover performance on food industry based on SMED, International Journal of Advanced Manufacturing Technology 90. Tier 1
  • Verbraucherzentrale: Grocery retail and best-before date - one-third rule of delivery. Tier 2
  • Thünen-Institut (2024): How food waste can arise - over three quarters of unsold goods in retail end as waste. Tier 2
  • Zhao, James (2018): Fat bloom formation on model chocolate stored under steady and cycling temperatures, Journal of Food Engineering. Tier 1
  • Rolfo (1980): Optimal Hedging under Price and Quantity Uncertainty: The Case of a Cocoa Producer, Journal of Political Economy 88. Tier 1
  • Białkowski, Bohl, Perera (2023): Commodity futures hedge ratios: A meta-analysis, Journal of Commodity Markets 30, article 100276 - 1,699 hedge ratios. Tier 1
  • myBytes Research: The Single-GARCH Limit on Soft Commodities - volatility model on cocoa, coffee, sugar, cotton; no early warning; open calculation, fixed seed. Tier 4
  • myBytes Research: The Second Layer: regime model and lead time - cocoa detection on 16 May 2023 as a whisper, not an alarm; open calculation, fixed seed. Tier 4
  • ISO 8586:2023: Sensory analysis - Selection and training of sensory assessors. Tier 2
  • European Commission, Access2Markets (28 January 2026): Regulation (EU) 2025/2650 amending Regulation (EU) 2023/1115 - application from 30.12.2026 and 30.06.2027 respectively. Tier 2
  • myBytes Research: Six Months before the New EUDR Deadline - placement of the amended regulation, 8 to 14 first placers per mid-sized company. Tier 4
  • myBytes Research: Reliability of individual pixels in EUDR evidence - threshold sensitivity of the cocoa probability layer. Tier 4
  • Ascarza (2018): Retention Futility: Targeting High-Risk Customers Might Be Ineffective, Journal of Marketing Research 55. Tier 1

Industry and consulting studies (tier 3) never stand alone in this series and never as potential; they complement a peer-reviewed or institutional anchor or are marked as an industry estimate.

Note on the context of this portfolio

All euro amounts are model calculations. They are derived from clearly named assumptions (assumption base → lever → result) and illustrate the order of magnitude, not a commitment. The total is a gross effect before implementation costs; no module has independent evidence for its degree of effect, the lever ratios are assumptions.

This calculation rests on publicly accessible sources and marked assumptions. It may be too high or too low; the reality in your factory decides. A data workshop puts your figures into the same formulas - then the question "where do these numbers come from" is answered, because you have seen how they arose.

All 9 modules: AI in the chocolate factory