# Carrefour (B1 Paris): Optimizing the Assortment with Three Algorithms

**Authors:** Albert's Deep Dive
**Categories:** Business Deep Dives
**Tags:** Carrefour, Retail, Optimisation d’assortiment, Algorithmes, B1, Data engineering
**Last Updated:** 2025-11-05T14:13:23.390Z
**Reading Time:** 1 min read

---

## Summary

A demanding retail case at B1 Paris: heterogeneous data, operational constraints, and a winning three-algorithm approach to list the assortment, recommend removals/replacements, and validate.

---

Carrefour BDD offered a rich case: optimizing sales across different store types and regions. The datasets, among the most complex of the year, put students to the test (merging, cleaning, operational constraints).

The winning team (Madeleine Landry, Raphaebl Lazimy, Theodor Le Molgat, Gabriel Leben) presented a three-algorithm framework:
1) enter a store identifier to retrieve the full product list (quantities, revenue),
2) generate data-driven removal/replacement recommendations,
3) validate the proposed changes against operational constraints.

Other teams presented a range of ideas, with notable creativity and collaboration.

## Key Takeaways

1. Use a **three-algorithm framework**: pull the **store identifier** SKU list, generate **data-driven removal/replacement** recommendations, then enforce **operational constraints** before rollout.
2. Invest in rigorous **data merging and cleaning** to handle complex retail datasets; reliable inputs drive reliable assortment decisions.
3. Optimize at the **store and region level** to reflect local demand; avoid one-size-fits-all assortments that dilute sales and margin.
4. Pair analytics with **feasibility checks** (shelf space, pack sizes, MOQs, planograms) so recommendations are executable in real stores.
5. Drive outcomes through **cross-functional collaboration** among merchandising, supply chain, and data science to align KPIs and constraints.

## Frequently Asked Questions

### What is retail assortment optimization and why does it matter for Carrefour?

Assortment optimization ensures each store carries the right mix of products for its local shoppers, maximizing sales and margin while reducing stockouts and waste. For a multi-format, multi-region retailer like Carrefour, tailoring assortments by store type and region unlocks significant revenue and customer satisfaction gains.

### How does a three-algorithm framework improve store-level assortment decisions?

A three-step approach typically pulls the full SKU list for a store, scores items for retain/remove/replace based on KPIs (e.g., revenue, margin, velocity, substitution risk), and then validates changes against real-world constraints. This reduces guesswork, speeds decisions, and ensures recommendations are both profitable and executable.

### What data do you need to build data-driven removal and replacement recommendations?

You need SKU-store sales and margin, inventory and out-of-stock data, promo history, returns, and customer demand signals (e.g., basket affinities). Operational inputs like shelf capacity/planograms, pack sizes, supplier lead times/MOQs, and regional attributes help ensure recommended replacements fit constraints and local demand.

### How do you validate assortment changes against operational constraints?

Translate constraints into rules—shelf facings and linear meters, case-pack multiples, delivery minima, price zones, and planogram compatibility—and run each recommendation through them. Scenario testing (baseline vs. proposed) helps confirm feasibility and highlights where to adjust the mix before rollout.

### How do you measure the impact of an assortment optimization pilot?

Track KPIs like revenue, gross margin, units per transaction, substitution rate, on-shelf availability, inventory turns, and waste. Use matched store A/B tests over several weeks and control for promotions to attribute uplift confidently.

### How do you choose the best replacement when removing underperforming SKUs?

Prioritize substitutes with similar attributes and price tiers that have higher velocity or margin and proven basket affinity with the category. Check supply reliability and shelf fit to avoid introducing new stockouts or planogram issues.

### What are common pitfalls when cleaning and merging complex retail datasets?

Frequent issues include inconsistent SKU IDs, re-listings counted as new items, misaligned calendars, promo-induced outliers, and negative sales from returns. Create a master ID mapping, standardize time periods, winsorize outliers, and document rules in a data dictionary to prevent silent errors.


---

*Article from [Albert's Deep Dive](https://deepdive.albertschool.com) - Albert School's Journal*