Closed alpha with 3 restaurant-owner friends: what broke first

In mid-December came the first moment of truth: showing the product to real restaurant owners, not friendly relatives who say it looks lovely. We picked three restaurant owners we already knew from the interview rounds, willing to get their hands dirty. We gave them access, a promise of bug fixes within hours, and a bottle of wine for the trouble.

A week later we had a notebook full of things to fix. The feeling was the one you always get with first users: you realize many things you took for granted were not obvious at all.

What broke in the first 24 hours

Things did not hold up on day one:

• The reservation time picker. Designed around 15-minute intervals. One of the venues works in exact 30-minute shifts and wanted to block the in-between gaps. The whole component had to be rebuilt.
• The confirmation emails. Landing in spam on Gmail. Diagnosis: SPF and DKIM misconfigured. Lesson learned about email infrastructure we thought was trivial.
• Importing the existing menu. One of them tried to paste their menu PDF as is. Our importer assumed structured format. We went from "paste a PDF" to assisted parsing within three days.

Three big iterations in one week

Saying "we iterate fast" sounds nice in an interview. In practice it means one day you decide something whole was wrong and you redo it. We did three large redesigns in seven days:

1. Floor plan. The first version was a free canvas with fine drag-and-drop. Too complicated. We simplified it to a grid with predefined templates.
2. Service screen. The original showed every reservation of the day in a single list. At the lunchtime venue, with 60 covers, it was unreadable. We moved to a per-shift view with large cards.
3. Onboarding. We had it as a five-step wizard. People got bored on the second step. Now it is three steps with minimal data, and everything else gets filled in when the need actually appears.

What surprised us in a good way

It was not all problems. Three things they liked without us pitching them:

• The speed. The interface responds instantly. Compared to the tools they had been using, that alone was half the sale.
• The menu editor. Once the import problem was solved, editing a dish and seeing it appear on the public site live felt like magic.
• Being able to assign different permissions to each team member without having to talk to support.

What we took into the beta

The alpha left us with 47 prioritized improvements. The first six were critical; the rest, desirable. Before moving to the next phase, the six critical ones were fixed. The private beta, with more restaurants, was the next step. That is what the next post is about.