The Confirmation Bias: 7 Restaurant Beliefs You've Never Actually Tested

Confirmation bias is the tendency to seek out, interpret and remember evidence in ways that fit what you already believe — and to systematically pay less attention to evidence that would prove you wrong. It isn't a matter of dishonesty or stupidity: it's the default strategy every human brain uses to test a belief, and a restaurant, where almost nothing is cleanly measurable, gives that strategy free rein.

In 1960, psychologist Peter Wason gave subjects a deceptively simple task. He had a rule in mind that generates sequences of three numbers — the sequence 2-4-6 fit it. The subject could propose their own three-number sequences, was told each time whether it fit the rule, and eventually had to guess what the rule was. Almost everyone formed a hypothesis immediately — usually 'even numbers increasing by two' — and then went on to propose sequences that fit exactly that hypothesis: 8-10-12, 20-22-24, 100-102-104. Every time, the subject was told 'yes,' and every time, confidence in the hypothesis grew. What almost nobody did was propose a sequence specifically designed to break their own hypothesis — for instance 3-4-5, ascending but not even, not spaced by two. The real rule was simply 'three ascending numbers,' something almost every participant could have discovered with a single well-chosen test. Instead, many subjects left the experiment fully confident in a rule that was wrong, built on a stack of confirmations that could never have produced anything but 'yes.'

This tendency later got a name and a thorough scientific synthesis. Psychologist Raymond Nickerson described confirmation bias in his widely cited 1998 review as the systematic tendency to search for, interpret and remember information in ways that favor what you already believe — not out of bad faith, but as the outcome of what efficient reasoning actually looks like in practice. The crucial point of Nickerson's work is that it doesn't feel like bias from the inside: someone showing confirmation bias is genuinely weighing evidence, just systematically asymmetrically — evidence for the existing belief gets an easy pass, evidence against it gets a hard one.

Researchers Joshua Klayman and Young-Won Ha gave that asymmetry a more precise name in 1987: the 'positive test strategy.' Someone checking a belief almost instinctively looks for cases where that belief would predict 'yes,' rather than actively hunting for the case most likely to disprove it. In everyday life that strategy is often harmless, even efficient — most beliefs a person holds are roughly true, so finding confirmation costs little effort and rarely produces a surprise. It only becomes a trap the moment a belief happens to be wrong: a positive test strategy applied to a false belief just keeps generating more apparent evidence, without ever finding the one case that would expose the mismatch.

This guide walks through seven concrete beliefs most owners form at some point and never test again — from the menu to the books, in the order a restaurant actually runs into them. At the bottom is a falsification check: for each of the seven, it asks when you last checked it with evidence that could have proven you wrong, and names the belief that's least tested at your own restaurant. Everything runs in your own browser; nothing is sent or stored.

Why a restaurant offers close to ideal conditions for confirmation bias

Feedback in a restaurant is slow, noisy and rarely repeatable. A quiet Tuesday could be the weather, a local event, a new price, a competitor opening down the street, or plain chance — there's no controlled experiment, just one evening that could be explained a dozen different ways. That exact ambiguity is the breeding ground for confirmation bias: ambiguous evidence has to be interpreted, and researchers Charles Lord, Lee Ross and Mark Lepper showed in 1979 that people systematically interpret the same ambiguous set of data in whichever direction matches what they already believed before they saw it. Two owners can look at the exact same numbers from the same month after a price increase and each grow more confident in their own, opposite belief — one that guests didn't mind, the other that guests are quietly staying away.

On top of that sits a second factor few other industries face so sharply: in a restaurant, there's usually no one who plays the role of critic separately. An accountant checks the books, a health inspector checks the kitchen — but whether the signature dish still sells, or whether that one staff member really isn't a team player, is rarely formally tested by anyone other than the owner themselves. The same person who wants 'our lasagna is a hit' to be true is also the person who decides whether last month's numbers count as proof of it. Without that separation between who forms the belief and who tests it, the positive test strategy runs unchecked.

And the stakes are rarely neutral. Psychologist Ziva Kunda described in her influential 1990 review of motivated reasoning that people arrive faster at the conclusion they want to reach, bounded only by their ability to justify it — and that higher personal stakes don't reduce that tendency, they strengthen it. A restaurant is rarely 'just a business' for the person who opened it: money, identity and years of unpaid overtime hang on specific decisions turning out to have been right. Per Kunda's research, the brain then works harder, not less hard, to arrive at the conclusion it already needed to reach. The seven mechanisms below aren't a character flaw — they're exactly what an efficiently reasoning brain does under noisy, high-stakes, self-graded conditions, which describes nearly every decision an independent owner makes.

The Ultimate Guide Staff & Operations: the complete system behind a team that keeps running Hiring, onboarding, scheduling, delegating and the psychology that colors every decision about it — everything a restaurant needs to run without funneling through the owner. Open the guide

The 7 beliefs you've probably never actually tested

They're listed in the order a restaurant runs into them — from the menu you build to the books you do (or don't) open at the end of the month.

1. "This is our signature dish" — the menu item nobody actually rechecks

A dish usually earns its reputation early: a strong opening month, a compliment from a regular, a mention somewhere online. From that point on, every subsequent signal gets read through that lens — a slow week for that dish gets blamed on the weather or a generally quiet stretch, never on the dish itself, while a stray compliment gets filed straight away as fresh confirmation. That's Klayman and Ha's positive test strategy in action: you keep looking for confirmation the dish still works — a happy comment, an empty plate back — instead of the evidence that would actually settle the question: where the dish genuinely stands in this month's sales numbers.

The falsifying test here is mechanical and doesn't care how well the dish is remembered: a Stars/Workhorses/Puzzles/Dogs quadrant built from margin and sales volume together. Menu engineering is built exactly on this, and the recipe costing tool makes the real margin per plate visible instead of an estimate that happens to match what you were already hoping for.

Confirmation bias lives exactly in the gap between these two columns: the left one is easy to find, the right one is rarely actively sought.

What feels like proof

• Nobody complained about the new price
• The regulars who still show up love it
• It worked the last time we tried it

What actually tests it

• Compare covers to the same period last year, adjusted for season
• Ask the regulars who stopped coming, not the ones still here
• Deliberately look for the one time it didn't work, and find out why

Neither column is dishonest on purpose — confirming evidence is simply easier to find than falsifying evidence. The only question is which of the two you actually consult before deciding.

2. "A price increase will scare guests away" — the test that never actually got run

Owners afraid of a price increase almost never run that test cleanly. Instead they wait for a sign their fear was right — one comment about things 'getting expensive,' one regular ordering slightly less — and treat that as proof, without ever systematically comparing covers to the same period a year earlier. Lord, Ross and Lepper's research on biased interpretation explains exactly why that works: the same ambiguous month of data after a price increase gets read by a fearful owner as 'see, I was right,' and would be read with equal confidence by a confident owner as 'no effect, we're fine' — the numbers alone don't decide anything; the belief that came first does.

The falsifying test is simple but rarely done: compare covers to the same period last year, adjusted for season, instead of counting conversations at the table. The article on raising menu prices goes deeper into how to structure that, and the restaurant benchmark puts your own numbers next to a real external band instead of your own memory of 'normal.'

3. "I can spot a good hire from the interview" — an interview that rarely tests what it should

This isn't about the first impression itself — that's the territory of the halo effect, where one striking signal colors everything else. Confirmation bias takes over afterward: once convinced within the first few minutes, an interviewer unconsciously starts asking questions and interpreting ambiguous answers in ways that keep confirming that first read instead of testing it — Lord, Ross and Lepper's biased interpretation, applied live, to a real person, in the same conversation. An interviewer sold on a candidate within three minutes unconsciously asks warmer, easier follow-up questions; faced with the same doubt about someone else, the questions get subtly sharper. So the 'evidence' gathered by the end of the interview wasn't even collected under the same conditions.

The fix is a structured interview with fixed, job-specific questions that don't bend with the first impression, tracked through the hiring board, and a separate, scored check of job-specific traits via the role-based personality test — decoupled from the interviewer's gut feeling. A job ad that already prompts the right, job-specific questions means there's something to test from minute one, instead of only something to confirm.

4. "Our regulars are happy, so we're doing fine" — a sample of whoever's still standing there

The regulars who give compliments are, by definition, the ones who are still coming back — a sample that has already filtered itself down to whoever is satisfied enough to return. An owner who takes that warm feeling as proof that 'things are going well' misses the counterpart: the guests who quietly stopped booking are rarely tracked down to find out why. Nickerson's definition of confirmation bias explicitly covers not just how evidence is interpreted, but which evidence gets actively sought in the first place — and at most restaurants, the disconfirming evidence (the guests who stopped coming) is never even collected, let alone weighed.

The regulars & VIP guest book makes that gap visible instead of invisible — a guest who hasn't booked in three months stands out instead of fading from memory. The review response generator and its companion reputation check bring the wider signal together, beyond the handful of loudest voices who happened to say something.

Not spread evenly across the seven beliefs, but clustered around four kinds of decisions.

Decisions about people — hiring and managing (32%) Decisions about money — pricing and the books (30%) Decisions about the product — the menu and marketing (24%) Decisions about guests — how you read your regulars (14%)

Illustrative ratio based on the seven mechanisms in this article, not a measured percentage of any specific restaurant. The check below runs on your own answers.

5. "This channel is what brings us guests" — the channel that never got an honest test

An owner remembers a handful of vivid wins — someone mentioning they found the restaurant on Instagram — and from then on silently credits every new guest to that channel, without ever tracking what share of actual bookings genuinely comes from it. Nickerson names this pattern explicitly: people ask more questions and run more checks for hypotheses they already favor, so a channel the owner already expects little from never gets a fair test — its real potential stays invisible by design, not by evidence.

The fix is simple but rarely done: ask every guest at booking how they found the restaurant, and track it systematically instead of relying on memory. The guide to AI search visibility and the article on email marketing describe exactly the channels most owners write off without ever having measured them.

6. "That person just isn't a team player" — one early judgment colors months of behavior

The moment a manager decides early that someone 'isn't a team player,' every subsequent ambiguous action — turning down an extra shift, a quiet evening, a joke that lands wrong — gets read as more proof, while the exact same ambiguous action from a colleague already in good standing gets read charitably or simply goes unnoticed. This is literally Lord, Ross and Lepper's 1979 experiment — identical, ambiguous behavior, judged in opposite directions depending on the belief that came first — transplanted from a laboratory to the pass of a kitchen.

The skills matrix replaces the cumulative impression with concrete, per-skill scored criteria, so a judgment rests on what someone can actually do rather than on a story that keeps confirming itself. The training plan and the staff handbook give every new hire the same, fair starting point instead of the reputation of whoever held that job before.

7. "The books are probably fine this month" — the ledger that only gets opened on good nights

An owner glances at the till on a busy Friday and feels immediately reassured, but quietly skips the numbers from a slow Tuesday — or only opens the books once the month already felt good. That's confirmation bias applied to your own financial data: the timing of checking is set by mood, so the sample of 'months I actually looked at closely' ends up stacked with months that already looked fine, while exactly the months most worth investigating get skipped most often. This isn't full avoidance — that would be the ostrich effect — it's selective looking, in a way that keeps the belief 'we're basically fine' intact.

A fixed nightly ritual removes the choice of when to check: the daily close & cash-up sheet turns checking the numbers into a habit instead of a mood, the cash-flow planner adds monthly discipline, and the restaurant benchmark compares your own numbers to a real external band instead of your own memory of 'normal.'

Falsification check: which belief have you actually tested?

For each of the seven beliefs above, answer honestly: have you ever tested it with evidence that could have proven you wrong, or does it mainly rest on signals that happened to fit what you already thought?

The check adds up to a single score from confirmation-only to stress-tested, and names your least-tested belief — not to hand out a grade, but to show you where the cheapest first test is. Everything runs in your browser; nothing is sent or stored.

Two things to keep in mind about your own result. A low score isn't a failure — it's exactly what Wason saw in almost every subject in 1960, and what Nickerson described in 1998 as the default strategy of an efficiently reasoning brain. The point of this check isn't to assign blame, but to show where the cheapest first falsifying test sits.

And remember that a high score at one moment isn't a permanent pass. A belief that's thoroughly tested today can slide back to confirmation-only in six months, the moment things get busy again and there's no time left for an honest check. So repeat this check every quarter rather than once.

What to do this week

Testing seven beliefs at once is impossible in a week. This order works, because each step makes the next one concrete and cheap.

First: run the falsification check honestly on yourself

• Answer the seven questions above with an honest read of when you last tested something, not of how confident it feels.
• For your least-tested belief, write down exactly what evidence could prove you wrong.
• Pick one belief to actually test this week, instead of tackling all seven at once.

Then: replace memory with a fixed measurement

• Put the margin and sales of every menu item into the menu engineering quadrant, instead of trusting how good the dish sounds.
• Ask every new booking how they found the restaurant, and track it systematically instead of relying on memory for your marketing channels.
• Look at the daily close just as closely on a quiet night as on a busy one, with the daily close tool.

Then: build structures that take the testing out of your memory

• Structure interviews with fixed, job-specific questions via the hiring board and the role-based personality test.
• Score staff on concrete criteria in the skills matrix instead of on a story you've built up.
• Retake the falsification check from this article in a quarter, and compare whether your least-tested belief has shifted.

Confirmation bias isn't a character flaw — it's how a busy brain tests beliefs when nobody pushes back

Wason's subjects in 1960 weren't foolish and weren't lying to themselves — they were doing the one thing an efficiently reasoning brain does under time pressure: seeking confirmation instead of the one case that would break their own hypothesis. Nickerson, Klayman and Ha, and Lord, Ross and Lepper each showed, in their own way, that the same pattern applies to nearly every belief anyone ever forms — including the seven this article walks through.

The fix isn't 'be less confident' — forming beliefs is unavoidable and usually useful. The fix is: know which seven places in a restaurant are most exposed to confirmation-only evidence, and build a fixed, mechanical checkpoint at exactly those places that doesn't depend on whatever happened to stand out most. A number from a quadrant, a comparison with last year, a fixed question at every booking — they all do the same thing: they make the falsifying test cheap enough to actually run.

Retake the falsification check above once you've tackled one of the seven beliefs — not to hit a score, but to see whether the next weakest spot has shifted. That's the one moment confirmation bias works entirely in your favor: when you're the one choosing which evidence to go looking for.