# Crunch Analytics

> Crunch Analytics is a reviewed AI agency in Ghent, Belgium, listed in the AIAgencies.eu register and last assessed 2026-09-10.

Canonical page: https://www.aiagencies.eu/agency/crunch-analytics
Generated: 2026-09-23 · Curated by Gabor Kiss · No paid placements — inclusion, score, and rank are never sold.

## Facts
- Location: Ghent, Belgium
- Website: https://www.crunchanalytics.be
- Hourly rate: not published
- Team size: 10-49
- Founded: 2016
- Services: AI Consulting, AI Development, AI Automation
- Industries: E-commerce, Manufacturing, SaaS & B2B
- LinkedIn: https://be.linkedin.com/company/crunch-analytics

## About
Crunch Analytics is a Ghent data and AI agency working on pricing optimization, demand forecasting and process automation for retail, FMCG and manufacturing clients.

## Documented clients
Documented engagements: Torfs, Ardo, Brico (Maxeda), Unilin, and Decospan. All 5 carry a source URL on file.

## Curator assessment
Reviewed 2026-09-10 by Gabor Kiss against published evidence.

The published cases are retail and manufacturing decisions handed to a model, each with the client named: end-of-season markdown pricing at the shoe retailer Torfs, promotional pricing for Brico under Maxeda, assortment analytics at Unilin that took an analysis cycle from weeks to days, demand forecasting at the frozen-food group Ardo.

The Torfs work is the one to read, because the result was measured against the old method rather than asserted—an A/B comparison with the retailer's traditional markdown process, reported at up to 30% more revenue and up to 8% better margin in the AI group.

Running against that, the write-ups stay at business level: no model families, no accuracy methodology, no monitoring or retraining design is published, so the engineering has to be assessed in conversation.

The weaker fit is buyers outside retail, FMCG and manufacturing, or teams wanting a language-model product rather than forecasting and pricing decisions.

### Key strengths
- An A/B test against the client's existing process is the strongest form of evidence in this category, and the Torfs markdown case publishes one
- The markdown assistant was scaled to further product categories after the first deployment, which says the work survived contact with operations
- Case studies are honest about the boring part: the Ardo write-up argues that model quality follows from product lines, customer concentration and logistics context rather than algorithm choice

### Good to know
- No stack detail is published—ask which models are in production, how accuracy is measured and who retrains them once your assortment changes
- The team page did not resolve at review time and no headcount is published; ask who is assigned and whether the data engineering and the modeling are the same people
- No rate is published—anchor the conversation against the European range on our rate index before scoping

---
Source: AIAgencies.eu — the curated register of European AI agencies. Rate provenance: "agency-confirmed" = disclosed via submission or claimed profile; "directory-listed" = Clutch or two agreeing B2B directories (lone unverified sources are never written). Full method: https://www.aiagencies.eu/methodology · Rate index: https://www.aiagencies.eu/rates