Anchoring Bias: Why the First Number You Hear Distorts Every Number After It

A recruiter sends you a candidate’s salary expectation: 95,000. You hadn’t fixed a number yet. You were going to benchmark the role, talk to HR, look at what the rest of the team is paid. But 95,000 is in your head now. By the time you land on an offer, you’re at 88,000, and you feel good about it, because you came in under their ask. What you can’t see is that if their number had never reached you, your own analysis would have settled around 78,000. Their opening figure pulled your final one up by ten grand, and the whole time it felt like rigorous work.

You probably know about anchoring already. You might even have written 95,000 down and told yourself “that’s just their anchor, I’ll set it aside.” And you still landed at 88,000. Naming the number isn’t the protection people think it is, and what follows is why, and what actually moves the figure back.

The evidence

Start with how hard this thing is to shake. It’s among the most reliably reproduced findings in decision research, and the reason it’s famous is that it works even when the anchor is plainly garbage.

In the original demonstration, people watched a wheel spin to a random number, then estimated what share of countries in the United Nations were African. The wheel had nothing to do with the question, and they knew it. People who saw a high number still guessed far higher than people who saw a low one, by more than twenty points on average. A number they’d watched come up at random moved their answer.

It gets worse with money. In one study, people wrote down the last two digits of a personal ID number, then bid on wine, chocolate, and gadgets. The ones with higher digits bid sixty to a hundred and twenty percent more than the ones with lower digits, for the identical items. The anchor was obviously meaningless, and it still set the price they were willing to pay.

The instinct now is to say: fine, but that’s untrained people guessing at trivia and trinkets. Expertise must help. It doesn’t. Experienced judges were asked to set a sentence after rolling dice that were openly described as random. The ones who rolled higher handed down longer sentences. Real estate agents walked through an actual house with identical information packets, except for the listed price. The listed price moved their independent valuations, and when asked afterward, eighty-one percent of them flatly denied it had. These were professionals whose living depends on getting the number right, and they couldn’t feel it happening.

How it works

So why doesn’t knowing help? Because the bias doesn’t live in the moment you accept a number. It lives in how you look for your answer.

When you face an uncertain figure, you don’t compute it from nothing. You start from whatever value is already in front of you and you adjust away from it, in small steps, checking each step against what you know. The trouble is where you stop. You stop at the first value that feels plausible. And the first plausible value you reach is almost always still on the anchor’s side of the truth, because you’re walking toward your real estimate, not past it. You quit the moment the number stops feeling wrong, which is well before it starts being right.

That’s what naming the anchor can’t fix. The anchor isn’t just a starting line you can step off. It sets the direction you walk and the neighbourhood you search. A high anchor sends you looking among high-but-plausible numbers and you settle on one. A low anchor sends you the other way. Your final figure isn’t anchored because you believed the opening number. It’s anchored because you went hunting for your answer in the streets right next to it.

That’s also why naming it falls short. Labelling the number “anchor” tells you where you started. It does nothing about where you’re searching, and the search is the whole problem.

The anchor sets the neighbourhood you search in, and you stop the moment the number stops feeling wrong.

How to use it

The clean version is prevention: form your own estimate before anyone else’s number reaches you. If you’re pricing a deal, sizing a timeline, or setting a budget, do the work first, off your own benchmarks, and write the figure down. Once it’s on paper it’s harder to drift, because now you have a fixed point to drift from. This is the strongest move there is.

But you won’t always get to go first. The recruiter emails the number before you’ve thought about it. A supplier quotes you on the call. The board floats a target before you’ve run your own. So the harder move is what to do once the anchor is already in your head and you can’t un-hear it.

Don’t try to adjust away from their number, because you’ve seen that adjustment stops short. Build a fresh estimate from a different starting point entirely. Go to the raw inputs and re-derive the figure as if their number didn’t exist: for the salary, start from the market band and the team’s actual pay and work up, rather than starting from 95,000 and working down. You’re not correcting their anchor. You’re refusing to search in its neighbourhood at all, and giving yourself a competing one built from real data.

When you can’t fully escape it, at least widen the search on purpose. Force yourself to argue the number is far too high, then far too low, and find the real evidence for each before you settle. You’re deliberately dragging your search out of the narrow band the anchor wants to keep it in. It won’t make you immune, but it pulls the final figure further from the opening one than a single muttered “that’s just their anchor” ever will.

And in a negotiation, treat order as a decision, not an accident. The first real number on the table shapes the range everyone argues inside. If your own analysis is solid, there’s a case for putting your figure down first and setting that range yourself. If it isn’t solid yet, the move is to stall the number, do your own derivation, and come back with it, rather than reacting to theirs in the room.

Why it matters

The thing that makes anchoring dangerous isn’t its size, though the size is real. It’s that it wears your own handwriting. Most thinking traps leave you with a faint suspicion you’re being worked on. This one doesn’t. You experience a pulled number as a reasoned one, which is exactly why the agents denied it and why you’ll trust your 88,000.

So the useful question to carry isn’t “what do I think the number is.” You’ll always have an answer to that, and it’ll feel earned. The question is “where did this number come from, and would I have landed here if I’d started somewhere else.” That’s a harder question, and it’s the only one that catches the figure wearing your reasoning as a disguise. Most of the time, when you actually trace it back, the honest answer is that someone handed you a starting point and you furnished the rest yourself, which is what good judgement is supposed to prevent and usually doesn’t.