Restaurant Discovery Is a Ranking Problem — And the Inputs Are Wrong

Marcin Stepien — Sun, 11 Jan 2026 13:18:53 +0000

Most restaurant discovery systems look data-driven, but their core signals are surprisingly weak.

Star ratings are treated as ground truth, even though they ignore time, context, and intent. A review from three years ago has the same weight as one from yesterday. Paid visibility, SEO authority, and review velocity quietly shape rankings, yet are presented as neutral relevance. From a systems perspective, we’re mixing monetisation signals with decision signals and pretending the output is objective.

The UX layer makes this worse. Most interfaces assume users want to browse, compare, and read. In reality, most food decisions are execution-driven: limited time, limited patience, and a strong desire to avoid regret. The cost function is speed and confidence, not optimality.

This isn’t really a data problem. It’s a ranking and interface problem. We optimise for engagement and retention, while users are trying to answer a much simpler question: what’s a safe, decent choice right now?

I’ve been exploring a different approach through a small experiment called BiteNow (https://www.bitenow.com.au) — not to replace maps or reviews, but to act as a real-time decision layer that prioritises immediacy, context, and low cognitive load over historical popularity.

I don’t think the current model is “wrong” — just misaligned with how people actually decide under constraints.

Curious how others would rethink ranking signals or UX if speed and confidence were the primary objectives.

Forem: Marcin Stepien

Restaurant Discovery Is a Ranking Problem — And the Inputs Are Wrong