Context & objectives

A growing Belgian food retailer had accumulated years of CRM data, but poor data quality was holding back their B2B expansion. Duplicate records, missing contact details, and inconsistent transaction histories made it difficult to understand who their best customers were or where to find more like them.

The retailer needed to:

• Enrich their prospect dataset with reliable external data

• Clean and deduplicate their CRM

• Segment prospects based on transaction behavior and firmographic characteristics

The goal was to increase average customer spend and strengthen loyalty for current customers, while acquiring new high-value customers.

Approach

1. Data cleaning and validation

We started by cleaning the CRM to improve the reliability of every analysis that followed. This mainly meant removing duplicates based on cleaned VAT numbers.

We then validated the project scope, which excluded customers and transactions that fell outside the campaigns our client planned to run.

2. Data enrichment

We extracted data from two Belgian public sources (BCE and BNB) to fill gaps in customer profiles:

• BCE: sector classification, company start date, potential email and phone numbers

• BNB: financial reports including FTE count, gross operating margin, and profit/loss figures

This data enrichment gave the retailer a much fuller picture of each prospect's business profile.

3. B2B segmentation and lookalike extraction

Using transaction data and the enriched firmographic information, we identified features and tags for each customer. This allowed us to categorize customers by yearly spend and map the clusters where high-value customers were most concentrated (by sector, company age, and company size).

From there, we extracted lookalike prospects from the BCE dataset: companies matching the profile of the retailer's best existing customers but not yet approached.

4. Consolidation and handover

All outputs (features, tags, and segments for both existing and prospective customers) were consolidated into files ready for CRM import. We also delivered full documentation covering the approach, the scripts used, and the procedure for applying the B2B segmentation to new data.

Results

After completing the project within less than two weeks, we delivered:

• An enriched and consolidated dataset of existing customers

• B2B segmentation assigned to each customer

• An enriched dataset of prospective customers based on lookalike modeling

• Clear descriptions of each identified segment

• Documentation on the approach and procedure to reproduce the outputs

The client's sales partner was able to use the segments immediately to inform future campaigns. With a clear view of which prospect profiles carried the highest commercial potential, the retailer could allocate acquisition and retention resources where they would have the most impact.