WZ Analytics

Cafes, restaurants and salons across Warsaw and Kraków had access to POS exports, spreadsheets and receipts — but no time, and no analyst, to make sense of any of it.

Owners were making Tuesday-lunch decisions on Monday-night gut feel. The real signal — which product was underperforming, which day was quietly bleeding cash — sat buried in CSV files that nobody opened.

Instead of building yet another SaaS analytics platform, we built the opposite: an AI pipeline that ingests raw POS data overnight and delivers one plain-English email at 8:00 every morning.

No app, no login, no learning curve. If the owner can read an email, they can run their business by data.

We chose Next.js on Vercel for the marketing surface, a Python worker on a single VPS for the AI report engine, and Postmark for delivery. That’s it. No Kubernetes, no serverless spaghetti.

Every morning a scheduled worker pulls each client’s data, runs it through GPT-4o with a tightly-scoped prompt library, and ships an HTML email — with fallbacks and idempotency guarantees.

Within four months WZ Analytics had 50+ paying businesses across two countries and an average +18% revenue lift among clients who acted on the recommendations.

The founder now runs a real, growing SMB SaaS — not a hobby project. The product has been featured in Polish F&B industry newsletters and has a waitlist of 200+ businesses.