Most explanations of anchoring start with a number. The more interesting part is what happens after we see that number.

Amos Tversky and Daniel Kahneman named the effect in 1974. They described a two-step mental move they called “anchoring and adjustment.” You take a starting value, then nudge it up or down until you reach an answer that feels right. Simple enough, right? The problem is with step two, where the nudging tends to stop too soon. We tend to settle near the anchor and call it a day.

They demonstrated it with a wheel of fortune that was secretly rigged to stop on one of two numbers, either 10 or 65. Participants watched the wheel spin; they believed it was random, and then they estimated something completely unrelated. The pair had the participants estimate the share of African countries in the U.N. The participants who saw 65 guessed high. Those who saw 10 guessed low. A number with no connection to the question still bent their answer.

Stop and think about how strange that is. The participants thought they had just watched a random spin. The number meant absolutely nothing, but it anchored their guess anyway.

And that’s the real lesson here, and it’s why anchoring is so dangerous. You don’t have to believe the first thing you see has anything to do with anything else. You just need to see that thing first.

Later researchers dug into why our adjustment falls short. Nick Epley and Tom Gilovich framed it as a resource trade-off. Adjusting away from an anchor takes mental effort, and effort is expensive. So we adjust until we hit the edge of a plausible range and we tend to stop, because “plausible” feels like “correct.” The anchor gives us a cheap place to start, and our brains are wired to prefer cheap.

Then there’s the finding that should worry anyone who makes decisions for a living: experts anchor too.

Birte Englich, Thomas Mussweiler, and Fritz Strack ran a study with real criminal judges. These were seasoned professionals with an average of more than a decade on the bench. Each judge worked through the same hypothetical case: a trial, complete with the victim’s account, an expert’s risk assessment, and the prosecution and defense arguments. It wasn’t a real defendant, but it was built to feel like the real thing, and the evidence was deliberately mixed so the judges felt genuinely unsure about the right sentence. Then, before deciding, each judge rolled a pair of dice that were secretly loaded to land high or low. Same as with Tversky and Kahneman’s experiments, a number that obviously had nothing to do with the case. Judges who rolled high recommended longer prison sentences than judges who rolled low. Remember, these were experienced professionals with the same case, the same facts. The only thing that changed was the number they rolled on a pair of dice.

This wasn’t the only experiment. Other experiments showed the same thing with real estate agents. Northcraft and Neale gave agents identical house listings with different asking prices, then asked for appraisals. The agents’ valuations tracked the arbitrary listing price, and the agents insisted the price hadn’t influenced them at all.

That denial is the key to why anchoring is so dangerous. It runs underneath awareness. Obviously, we don’t feel ourselves being anchored. We feel like we’re reasoning from scratch, like we’re weighing the evidence and landing on a fair number. Meanwhile, our starting point is steering the whole thing.

It’s useful most of the time. But it’s pretty dangerous when the starting point is garbage.

The version of anchoring that hurts teams isn’t a rogue number in a spreadsheet. It’s the quieter fact that whoever frames a problem first sets the boundaries everyone else reasons inside.

Take the first draft. The person who writes the initial PRD, the opening strategy doc, or the first version of the roadmap has done more than start the work—they’ve set the anchor. Every review after that becomes a set of adjustments to their framing. Comments cluster around what’s already written. Nobody re-asks the question the doc quietly assumed. The first draft wins not because it’s best, but because it got there first.

Redesigns show this vividly. When a team sets out to “rethink” a feature, the old design becomes the anchor almost immediately. The new version drifts a little from the old one, then stops—because the old layout defines what “this feature” even looks like in everyone’s head. You end up with a reskin when you meant to reinvent. The existing thing anchored the imagination for the new thing.

The status quo works this way in general. Whatever exists now functions as an anchor nobody deliberately chose. “How do we improve onboarding?” starts from the current onboarding and adjusts outward, which quietly rules out the possibility that the whole flow is wrong. The current state is the loudest anchor in the room precisely because nobody notices it’s there.

Even research synthesis isn’t safe. The first interpretation someone voices in a readout, especially a confident one, becomes the frame the rest of the team reasons from. Alternative readings of the same data get measured against that opening take instead of generated fresh, so the analysis narrows before the evidence has really been explored.

Everything so far is about not getting anchored. But the same mechanism runs the other way, and sometimes we want to be the ones using it… on purpose. If the first number always sets the range, then be the one who says it. The engineer who puts “realistically this is six weeks” on the table before leadership floats “can we do three?” has anchored the conversation toward the actual work instead of reacting to a deadline pulled from bag of wishes. Same with scope, staffing, or the size of a task—whoever names the honest figure first shapes what everyone treats as reasonable.

That same tactic works outside the team, too: in a salary negotiation, the first number on the table tends to pull the final one toward it, which is a large part of why recruiters push you to name your expectations before they name theirs. This isn’t about tricking anyone. It’s leverage that already exists in every estimation meeting, and the only question is who knows enough to use it. Leadership anchors aggressive timelines constantly, and they know exactly what they’re doing when they do it. Naming a grounded number first is how a team keeps a realistic baseline from getting defined by someone whose incentives point somewhere other than the people doing the work. Remember your corporate overlords are, usually, not looking out for you.

The through-line from the judge and real estate studies is that awareness and expertise barely help. You can know all of this and still anchor on the first draft, the old design, the current state. So the move isn’t to try harder to be objective. It’s to change the order and source of information on purpose: have people reason before they see the existing artifact, invite framings that start from a blank page instead of the current one, and treat the first version of anything, like a doc, a design, a take, as one input rather than the center everything else revolves around.