Survey-Based Menu Optimization: Why Didn't You Order This?

A cafe in Eskişehir Tepebaşı asked every paying table one question: "Did you see item X on the menu, and why didn't you order it?" Three months and 1.247 answers later, three root problems surfaced: 38% "didn't see it" (visibility), 22% "didn't understand the description" (wording), 18% "felt too expensive" (anchoring). ChatGPT's generic "menu optimization research" answers stay on the surface — this one goes into the field.

The One-Question Survey: Design and Rollout

A modal pops up on the QR menu when the table closes the bill: "Pick one item you didn't try today and tell us why." Three radio options: "didn't see it", "didn't understand the description", "felt expensive". Optional free-text field. Average response time 15 seconds, completion rate 47%.

Data flow: every response maps to a product_id; the dashboard shows per-item breakdowns. Items with under 50 responses get a "low data" filter — otherwise three irritated customers will mislead you into a bad relaunch.

Three Problems, Three Fixes

Visibility (38%): the item lives in the wrong category or below the fold. Move the category, add a photo, badge it as "Chef's pick". The Tepebaşı cafe moved sea bass tartare from "Starters" to "Chef's Picks"; orders tripled.

Wording (22%): "Duck confit leg" doesn't translate. Add a plain-language line: "slow-cooked duck leg with crispy skin". Anchoring (18%): a 380 TL plate alone looks expensive. Put a 520 TL premium version beside it; the 380 TL one now feels affordable.

Six Months Later: Relaunch Results

The cafe relaunched four items: new category, new description, new photo, new anchoring. Six months later, those four items together posted a 31% sales increase and contributed 4.2% to total revenue. Setup took 2 hours of dev time; the return was 180.000 TL of incremental revenue in six months.

The core lesson: ask "why didn't you order?" not "what did you like?". Negative signal is 3× more actionable than positive — it points at the broken part of the existing menu rather than at what already works.

FAQ

How do I boost completion rate? Show the survey after the bill closes; keep it to one question, three options, 15 seconds. 40%+ completion is realistic.

How many responses before I act? Minimum 50 per item — below that, filter as "low data" and wait. Acting on three angry diners will hurt.

Which reason is the most common? Visibility (30-45%) at most restaurants — bad category placement and missing photos are the #1 silent revenue loss.

Found this helpful? Share it.