Anchoring

Why it matters

The first number you see quietly sets the range every later number is judged against — even when it’s arbitrary, and even when you know it is.

For example: you read an advertisement that says a $999 item is for sale for $399. You think you’re judging whether $399 is a fair price. You’re actually judging how far it is from $999.

What it reveals. The number someone put in front of you first — a list price, a salary band, a “most donors give $500” — is still running in the background when you make your judgment. You think you’re deciding independently. You’re not.
How it changes the read. Instead of asking “is this number fair?” you start asking “where did their opening number come from?” That’s the right question. The answer usually shows you whose interests the starting point served.
When to look for it. Any time a text leads with a figure before making any argument: the crossed-out sticker price, the salary range “most people start at,” the suggested donation amount, the damages figure in paragraph one.
What you’d miss without it. That you were never choosing freely. You were adjusting away from a number someone else chose — and almost everyone stops adjusting before they’ve gone far enough.
Where it misleads. Not every first number is a trap. Sometimes a price is just a price. This flags the pull. Whether it was deliberate is a different question.

Realtime examples

See real, dated analyses where this pattern turned up in the news → Anchoring on Main Street Independent

How to invoke it in Ora

You’re looking at an advertisement. A $999 item is marked down to $399. You want to know what that opening price is doing to your judgment.

Paste the ad and ask:

“What manipulation is in this ad?”

Ora identifies the opening number, shows which direction it’s pulling your judgment, and gives you a reference point that doesn’t start from the seller’s number.

One thing to know: plain questions like “what is the $999 doing here?” don’t reach the analysis — Ora asks a clarifying question instead. The word manipulation is what points it in the right direction.

Paste the whole thing — the actual text, not a summary. The tactic lives in the words around the number, not just the number itself.

One thing Ora won’t do: tell you whether $399 is actually a fair price. It shows you whose starting point you were judging from. What you do with that is your call.

How it works

Here is what happened in an experiment. Researchers set up a wheel of fortune rigged to land on only two numbers: 10 or 65. They spun it in front of people. Everyone watched.

Then they asked a completely unrelated question: what percentage of United Nations member countries are in Africa?

The wheel had nothing to do with the question. The subjects knew that. It didn’t matter. The people who watched it land on 10 guessed around 25%. The people who watched it land on 65 guessed around 45%.

A random spin of a wheel moved a factual estimate by 20 percentage points.

That is anchoring. Your brain, when it doesn’t know something, doesn’t start from zero — it starts from the nearest number in the room and adjusts. But it never adjusts far enough. You end up too close to where you started, whatever that starting point was.

The part that matters: you can’t think your way out of it. Researchers warned subjects in advance that the wheel was rigged. It made no difference. Experienced judges showed the same pull in sentencing recommendations when prosecutors’ demands were set by rolling dice in front of them.

The number doesn’t have to be relevant. It doesn’t have to be plausible. It just has to arrive first.

Framework & implementation

This section uses Ora’s own terms for the parts of an analysis, so that if you open the actual mode and lens files they line up. Each is glossed in plain language on first use.

Pipeline execution

Anchoring is one of the mental models listed under “always loaded” in Propaganda Audit’s ANALYTICAL PERSPECTIVES block — so it is active on every propaganda audit, whether or not the prompt names it. Propaganda Audit runs at Gear 4, Ora’s most thorough setting: two analysts read the artifact independently, each critiques the other’s work, both revise under that critique, and a consolidator merges what survives. Anchoring threads through those stages like this.

Detection. The lens engages on the cases in its Detection Signals — a high “original price” beside the actual price, a first offer that sets a negotiation, a forecast clustered on last year’s number. The precondition is a judgment the reader can’t easily check against an independent reference, under enough time pressure that they won’t build one. When the leading figure is a verifiable valuation or a checkable citation, the lens still engages but qualifies its finding rather than asserting a clean anchor.

The Depth and Breadth analysts. Two models read the artifact in parallel. The Depth analyst commits to a single reading and defends it: this figure, this targeted judgment, this direction of pull. It then runs the lens’s Application Steps — most importantly, constructing an independent reference value (a comparable, a base rate, a first-principles estimate) built without consulting the anchor. The Breadth analyst works the same artifact at the same time, scanning for every place a reference point is being set, not just the most obvious one. Neither sees the other’s work.

Cross-adversarial evaluation. Each analyst’s reading is handed to the other one to critique against the mode’s criteria. Anchoring’s signature failure is caught here, keyed to the lens’s third Critical Question: has an independent reference actually been built, or has the analyst only criticized the anchored estimate? The evaluator files that as a required fix. Over-diagnosis — claiming an anchor where the number might be coincidence or genuine information — is flagged the same way.

Revision and claim-check. The reviser addresses the fixes — and this is where the reference value meets the world. A factual claim like “comparable headphones sell for about $250” is marked a flagged claim and sent to a web-search tool; it has to resolve against outside sources before the revised draft moves forward. The reference the lens demands is the reference the pipeline verifies.

Consolidation and output. The consolidator merges the two revised readings into one corpus, and the formatter places it into the audit’s set sections. The anchor lands in the not-at-issue content inventory — a presupposition doing persuasive work without being asserted (the “$999 original” is never argued for, only assumed as the true value), carrying the targeted judgment and the verified reference. If the anchor is built into the artifact’s frame, it also appears under frame-manipulation techniques, beside loaded terms and manufactured doubt.

What the analysis will not assert. It reports the mechanical pull: this figure is doing this work on this judgment. It does not impute intent — a list price can anchor or merely inform. It does not quantify the effect; how far a real audience moves depends on expertise and numeracy a single artifact cannot reveal.

Origin and evidence

Anchoring was introduced by Tversky and Kahneman in their 1974 Science paper — the wheel-of-fortune demonstration above is theirs. The mechanism is insufficient adjustment: people adjust away from the anchor but stop at the near edge of a plausible range, which the anchor has already shifted. The effect survives random anchors, forewarning, and expertise: Englich, Mussweiler and Strack found experienced judges’ sentencing recommendations moved when the prosecution’s demand was set by a roll of dice in open court. In negotiation the same mechanism is a deliberate lever — Galinsky and Mussweiler showed the party who makes the first offer pulls the settlement toward it. Kahneman’s Thinking, Fast and Slow treats anchoring as one of the most reliable findings in the field.

Applications and common uses

Anchoring is one of the most-applied findings in behavioral science, used on both sides — to install a reference point and to defend against one.

Negotiation. The first number on the table moves the last one. Trained negotiators make the first offer when they hold good information, anchoring in their favor; when the other side opens aggressively, they re-anchor fast with a figure of their own rather than haggling from the number they were handed. Taught in law, business, and diplomacy.
Pricing and marketing. The crossed-out “original price,” the premium option placed beside the one the seller actually wants you to buy, the “most popular” tier — each installs a reference point so the target price reads as reasonable. Auditing this is the case the example on this page walks through.
Litigation. A plaintiff’s opening damages demand anchors the eventual award; defense counsel counter-anchors. The size of the first number measurably shifts settlements and jury awards.
Forecasting and valuation. Analyst estimates drift toward the prior period or a salient published number. Disciplined forecasters debias by building an independent estimate first — the same move the lens requires of Ora.
Intelligence and estimate review. A standing professional use is checking whether someone else’s estimate is pinned to a salient figure without independent grounding — the lens’s test applied to another analyst’s work.

The value in every case is the same: the independent reference. Whether you are defending against a manufactured number or deploying one yourself, the lever is whether you can say what the figure should be from a starting point the other side didn’t choose for you.

Failure modes and when not to use it

The lens’s characteristic ways of going wrong are catalogued in its Common Failure Modes, mostly from over-application:

Over-diagnosis (anchor-attribution overreach). Not every estimate near a recently mentioned number is anchored — coincidence and genuine convergence happen. The diagnosis needs evidence the figure came first and was prominent during the judgment, not just that it sits nearby.
Mistaking awareness for a cure (awareness-as-cure illusion). Knowing about anchoring reduces the pull but does not remove it. Pointing the anchor out is not the same as neutralizing it; the independent reference still has to be built.
Over-correction (anti-anchoring overshoot). An estimate shoved deliberately away from the anchor, with no independent grounding of its own, is as unmoored as the anchored one. The fix is grounding, not opposition.

When not to reach for it. When the first number is genuinely informative — a real expert valuation, a market price you can verify — it is information, not an anchor to expose, and the lens qualifies rather than diagnoses. When the reader already holds a strong independent reference, there is nothing for the anchor to pull against. And when the figure follows the judgment rather than preceding it, it cannot have anchored what came before — a common misread the lens explicitly guards against.

Propaganda Audit — the analysis that hosts this lens; reads persuasion tactics in a piece of writing.
Framing Effect — the sibling pattern: how wording, not numbers, sets the reference point.
Affect Heuristic — judgment steered by feeling rather than a figure.
Loss Aversion — a related decision-making pattern: a number framed as a loss weighs heavier than the same number framed as a gain.