Anchoring bias is what happens when a number — any number — lands in your mind before you make an estimate and quietly becomes the gravitational center everything else orbits. A salary negotiation opens at $60k and you counteroffer at $68k. A product is "marked down from $499" to $299 and it feels like a deal even if $299 is the real price. A project estimate starts at "two weeks" and the final answer stays in that neighborhood even when the work clearly demands six. The anchor doesn't need to be reasonable to stick.
What changed in 2026
- AI-generated estimates and valuations have become a new anchor source — the first number an LLM returns for "how much should I pay for X" becomes the anchor for subsequent human judgment, which matters if the model was trained on skewed data.
- Dynamic pricing makes consumer-facing anchors more sophisticated: retailers now test personalized "original price" anchors in real time.
- Negotiation coaching tools trained on recorded deal outcomes have confirmed what labs showed — the party who names the first number wins more often than not, even controlling for position quality.
- Reference-class forecasting became a standard project management practice at tech companies, specifically as an anchor-mitigation technique.
How anchoring works
When you make a numerical judgment, your brain starts from an initial value and adjusts from it — but almost always adjusts insufficiently. The starting point is the anchor. Tversky and Kahneman demonstrated this with a roulette wheel: even a clearly random number (say, 65) shifted participants' estimates of the percentage of African nations in the UN. People anchored on 65 even knowing it was random.
| Scenario |
The anchor |
The distortion |
| Salary negotiation |
First number named |
Counteroffers cluster near it |
| Home purchase |
Listing price |
Offer price adjusts from list, not market value |
| Project scoping |
PM's first estimate |
Final estimate stays ~20% of original |
| Retail pricing |
"Was $199" label |
Judgment of $129 as "reasonable" |
| Performance review |
Prior-year rating |
This year's rating adjusts from last |
Where it costs you most
Salary and contract negotiations. If you don't anchor first, the other side will — and you will spend the whole conversation adjusting from their number instead of yours. The fix is simple: go first with a well-researched, slightly aggressive figure.
Project estimation. Teams anchored to a leader's early guess produce estimates that cluster within 15–20% of that guess regardless of actual scope analysis. Run estimates before revealing a target number.
Investing. The price you paid for a stock becomes a powerful anchor — you wait to sell "until it gets back to $80" even when the fundamentals say it won't. This is anchoring plus loss aversion working together.
Consumer decisions. "Original price" labels, subscription tiers designed to anchor on the most expensive option first, and "compare at" pricing all exploit anchoring. Retail is engineered around it.
How to correct for it
1. Generate your own estimate first
Before receiving any external number, commit your own estimate to writing. This establishes a competing anchor based on your own analysis rather than an externally supplied one.
2. Reference-class forecasting
Rather than adjusting from a prior estimate, ask: "What is the typical outcome for projects/deals/situations like this one?" Ground your estimate in a base rate, not a starting number.
3. Consider the opposite
When given an anchor, deliberately ask whether the true value could be significantly lower or higher. Forcing the opposite extreme into consideration breaks the one-directional pull.
4. Separate information from anchor
Ask: "Is this number giving me real information, or is it just the first number someone said?" Anchors from uninformed sources deserve zero weight — treat them as noise, not data.
5. Multiple independent estimates
Get separate estimates from team members before any group discussion. Average or deliberate from independent anchors rather than one shared anchor.
How to pick the right debiasing approach
- Negotiating a salary or contract? Research market rates, set your floor privately, then name your number first.
- Estimating a project? Run structured independent estimates before revealing any target date or budget.
- Making an investment decision? Clear the purchase-price anchor — ask "would I buy this today at this price with no history?"
- Evaluating a purchase? Strip the "was $X" framing. Ask: what is this worth to me at this price alone?
Common mistakes
Thinking awareness eliminates the effect. Studies consistently show that people who know about anchoring are still significantly anchored — they just feel less anchored. Knowledge reduces but does not remove the bias.
Only worrying about other people's anchors. Your own prior estimates anchor your current ones. Teams anchored by a PM's throwaway comment in a kickoff meeting are anchored by an insider, not an adversary.
Ignoring temporal anchors. The first salary you ever received anchors your sense of what you "should" earn. The first apartment you rented anchors your sense of what rent "costs." These historical anchors persist for years.
Countering an extreme anchor with a moderate one. If you're anchored too high, counter-anchoring far in the opposite direction is legitimate and effective — don't be afraid to go significantly opposite.
What to skip
- Anchoring with obviously fake numbers in negotiations — they signal bad faith and destroy trust for any long-term deal.
- Paralysis-by-analysis trying to remove all anchors before deciding. Use the tools above and move; perfect anchor-neutrality is not achievable.
- Ignoring anchors when you benefit from them — if you're the seller and you set the anchor well, that's informed negotiation, not manipulation.
FAQ
Does anchoring work in writing as well as spoken conversation?
Yes, and often more strongly — written anchors are re-read and persist longer than heard ones.
How do I break a bad anchor in a negotiation already in progress?
Make an explicit reference-class argument: "Let me set aside the initial number and look at comparables." Reframing to market data is the cleanest reset.
Are some people more susceptible than others?
Moderately. People under time pressure, cognitive load, or with less domain expertise anchor more strongly. Expertise reduces susceptibility somewhat but doesn't eliminate it.
Should I always name the first number?
In salary negotiation, yes — research consistently favors the first mover. In auctions or competitive bids, it depends on information asymmetry.
Where to go next
For related biases that amplify anchoring's effects, see Loss aversion explained in 2026, Confirmation bias explained in 2026, and The planning fallacy explained in 2026.