Guest feedback & NPS
Restaurant NPS: why one score hides the problems that matter
Net Promoter Score was introduced as a simple company-level loyalty measure. A multi-section, multi-site venue is exactly where one blended number can stop being operationally useful. A strong kitchen can hide a slow bar, while a cold delivery order can drag down a dining room that did nothing wrong. The useful move is to split the signal, show the response count and give each part to someone who can act on it.
Updated 2026-07-31 7 min read
The short version
- NPS was introduced by Fred Reichheld as a simple company-level loyalty measure (Harvard Business Review, December 2003). A venue with distinct sections and channels is where one blended score can stop helping.
- A blended score of +40 can average together a +70 kitchen and a +5 bar. The average hides the exact problem you would act on if you could see it.
- Delivery belongs in its own column. Mixing Deliveroo and Uber Eats scores into your dining-room NPS punishes your floor team for a cold order they never touched.
- There is no honest single "good restaurant NPS" to chase. The comparisons that mean something are you against last month, and section against section.
- A section score with no owner is trivia. Attribute NPS to the team on that section and it turns from a scoreboard into a to-do list.
What is NPS, and where did the single number come from?
Net Promoter Score asks guests one question, how likely they are to recommend you on a scale of nought to ten, and turns the answers into a single figure. Scores of nine and ten are promoters, seven and eight are passives, nought to six are detractors. You subtract the percentage of detractors from the percentage of promoters, which leaves a number between minus 100 and plus 100.
The idea comes from Fred Reichheld’s article "The One Number You Need to Grow" (Harvard Business Review, December 2003). In the companies and industries studied, the recommend question linked more closely with customer behaviour and growth than the more complex satisfaction measures tested. That is a more careful claim than saying NPS predicts growth everywhere. Its simplicity is a genuine strength, and it is why the measure spread so widely.
The trouble starts when you take a metric built to describe one company and point it at a restaurant. A restaurant is not one experience. It is a kitchen, a bar, a floor, a booking process and, increasingly, a delivery operation, each capable of a very different night. Collapse all of that into one figure and you get a number that is easy to report and almost impossible to act on.
What is a good restaurant NPS?
The honest answer is that there is no single benchmark worth chasing, and hunting for one is the first symptom of the trap. NPS runs from minus 100 to plus 100, and anything above nought means you have more promoters than detractors. Beyond that, the cross-industry "good NPS" tables you find online blend industries with completely different rating cultures, so a number that looks strong for a bank tells you nothing about a Friday night in a gastropub.
The comparisons that actually mean something are internal. Is this month better than last month? Is your kitchen ahead of your bar? Is your Deliveroo channel dragging behind your dining room, and by how much? Those questions have answers you can act on this week. "Are we above the industry average" has an answer you can do nothing with, and it is often built on a benchmark nobody can source. Chase your own trend and your own spread, not a figure off a blog.
The moment you ask "what is a good NPS" instead of "which part of my venue is dragging", you have swapped a useful question for a vanity one.
Why does one blended score hide your real problems?
Because an average is a machine for hiding variation, and variation is the whole story. A single venue NPS is the mean of every part of the experience, so a strong area and a weak one cancel out into a number that describes neither.
Use a simple illustration. Say a venue posts a blended NPS of +40 and the manager files it as "fine". Split the same guests by what they were rating and the picture changes completely: the kitchen is at +70, the floor service at +45, and the bar at +5, dragged down by slow drinks on a busy pass. The blended +40 is technically accurate and operationally useless. It points at nothing. The three section scores point straight at the bar, which is the one conversation worth having on Monday. This is an illustration, not a claimed venue result.
This is the quiet cost of the one-number habit. Every month the single figure holds roughly steady, everyone nods, and the slow bar never gets fixed because it never gets seen. The number was not wrong. It was just averaged past the point of being any use.
Why does delivery need its own NPS column?
Because delivery is a different restaurant, and scoring it as if it were your dining room misleads you twice. A guest rating a delivery order is rating food after a journey, packaging, an app you do not control and often a courier you do not employ. None of that is your floor team, yet blend the channels and their score wears the damage.
The misdirection runs both ways. A strong dining-room month can be flattered by hiding a delivery problem inside the same average, so a channel that is bleeding repeat orders looks healthy. Or a rocky delivery week can pull down a dining room that ran beautifully, and the floor team gets a worse number for a courier problem they could not touch. Either way, the fix goes to the wrong place, because the measurement pointed at the wrong place.
Measured on its own, delivery becomes a channel you can actually manage: is it the food that does not travel, the packaging, the wait, or the handover. Measured inside the blend, it is just noise dragging on a number that already told you too little.
How do you make a section score something people act on?
You give it an owner. A section NPS that belongs to nobody is a fact on a dashboard; a section NPS that belongs to the team working that section is a target they can move. The difference is attribution, and it is where most feedback tools stop.
Set a score for the bar and the bar team can see it, chase it and own the recovery when it slips. Always show the number of responses beside the score and use a rolling window when section volumes are small. An NPS of +70 from ten responses is not as stable as +70 from two hundred, and a one-week swing can be noise rather than a service failure. With that guardrail in place, a smaller, owned number turns a monthly report into a practical target.
Where Revvlit fits
Revvlit measures NPS the way a venue actually runs, by section and by channel rather than as one blended figure. One QR per venue opens a short guest form with per-section questions, so kitchen, bar, floor and venue each carry their own score, and delivery is captured as its own channel, separate from dine-in, so a cold courier order never drags down your dining room.
Each response is also credited to the named team member the guest selects, and each section can carry its own NPS target that feeds the team’s points. So the score stops being a number in a monthly report and becomes something owned: the bar team sees the bar score, the floor sees theirs, and a live leaderboard in a native iOS and Android app shows who is moving it. When a guest flags a problem they are invited to leave an honest public review like everyone else, and they also get a private path to a manager, so the issue can be put right before it becomes a public one-star.
It will not tell you a magic benchmark to hit, because there is not one. What it does is stop your one useful signal being averaged into a number that points at nothing.
Start by splitting one number this week
You do not need software to see the trap for yourself. Take your next month of guest feedback and split it into three: kitchen, service and bar, with delivery in a fourth column of its own. Look at the spread rather than the average. The gap between your best section and your worst is the work, and it is the thing the single number was hiding. When you want that split captured automatically, credited to the team who move it and tracked live across every site, that is what we do.
Sources
- Harvard Business Review, The One Number You Need to Grow Fred Reichheld’s original December 2003 article introducing the recommend question.
- Bain and Company, Measuring Your Net Promoter Score NPS question, respondent categories and calculation.
See NPS by section, not one number that hides the problem
Revvlit measures NPS per section and per channel, credits it to the team who earn it, and tracks it live across every site. From £99 per site per month founding (50% off the £198 base), 14-day free trial.