Skip to content

tags: [] - coffee/tasting - coffee/tasting/evaluation aliases: - Anchoring bias - Anchoring effect (coffee)

Anchoring

Tags: #coffee/tasting #coffee/tasting/evaluation Aliases: Anchoring bias, Anchoring effect (coffee) Related: Tasting and Evaluation MOC | SCA Cupping Protocol | Confirmation Bias | Panel Management Status: ✅ Complete

Overview

Anchoring is a cognitive bias in which an initial piece of information — the "anchor" — disproportionately influences all subsequent judgements. In coffee sensory evaluation, anchoring manifests when the first coffee tasted, a known price, a prior score, or a descriptive suggestion sets a reference point that distorts how evaluators score later samples. Managing anchoring is a core concern of professional cupping protocol design.

How Anchoring Affects Evaluation

Score Compression

When the first coffee of a session is assigned a score, subsequent coffees tend to cluster around that initial figure rather than reflecting true quality differences. A panel anchored to an 83-point opening sample may score a genuinely excellent coffee at 85 and a genuinely poor one at 81 — compressing a real 80–90 range into a 81–85 range. The anchor constrains the evaluator's willingness to deviate substantially from the opening reference.

Sequential Effects

Sample order creates systematic bias. When the best coffee is presented first, subsequent samples register as inferior by comparison; when the worst is presented first, later samples receive inflated scores. A moderate-quality opening sample produces the most balanced distributional outcome across a flight. The SCA Cupping Protocol mitigates sequential effects through multiple cups per sample, evaluation at different temperature intervals, blind tasting, and randomised sample presentation.

Reference Dependence

Evaluators carry both numerical and conceptual anchors into a session. The convention that 85 points represents specialty grade can itself become an anchor — causing scores to cluster around that threshold rather than reflecting absolute quality. Similarly, knowledge of a coffee's price, origin reputation, or brand history creates expectation anchors: a high price tag primes higher scores, and a known origin reputation (such as "Kenya is bright") biases evaluators toward confirming that expectation.

Types of Anchors

Numerical anchors include previous session averages, a colleague's score shared before independent evaluation, a published score, or a target score for purchasing decisions.

Verbal anchors include phrases such as "this should be really good," "last year's crop was better," or "typical Colombia profile" — statements that prime certain descriptors and set expectation direction before the cup is assessed.

Experiential anchors are long-term reference points: the first coffee ever cupped, a memorable exceptional cup, or training standard coffees encountered during certification. These can serve as useful calibration references when actively managed, or as sources of rigidity when unconsciously applied.

Mitigation Strategies

Randomisation of sample presentation order — using a computer-generated sequence, rotated between judges — prevents sequence effects and removes systematic positional bias. Documentation of the randomisation scheme allows post-session analysis of order effects.

Calibration standards counter anchoring more effectively than a single unknown opening sample. Beginning a session with coffees of known, distinct quality levels (for example, samples representing 75, 82, and 88 on the SCA scale) provides multiple reference points rather than a single anchor, and resets evaluators to absolute rather than relative standards.

Blind protocols remove information anchors by concealing origin, price, processing method, brand identity, and previous scores. Samples are coded with neutral identifiers, changed between sessions to prevent evaluators from inferring identities across rounds.

Independent scoring — recording scores privately before any group discussion — prevents social anchoring, in which evaluators gravitate toward publicly stated scores. Group calibration follows independent scoring: extreme differences are identified and reasoning discussed; re-evaluation is then offered if warranted.

Full-scale training familiarises panellists with the extremes of the scoring range through periodic exposure to defective, commercial, specialty, and exceptional samples. Confidence in assigning scores at the extremes (below 80 or above 90) reduces the tendency to compress scores into a narrow middle range.

Statistical Detection

Anchoring susceptibility can be detected quantitatively. High correlation between the first sample score and the session average indicates anchoring. Abnormally low standard deviation across a session's scores indicates range compression. Plotting score against sample position and testing for sequence-dependent patterns identifies systematic order effects. These analyses are part of Panel Management frameworks for ongoing panellist performance monitoring.

Key Facts

  • Anchoring causes evaluators to cluster scores around an initial reference point, compressing the assessed range below its true extent
  • The first coffee tasted, a known price, published scores, origin reputation, and prior verbal descriptions all function as anchors
  • The SCA cupping protocol addresses anchoring through blind tasting, randomised order, and multiple cups per sample
  • Calibration with multiple known quality standards is more effective than a single opening cup because it establishes several reference points rather than one
  • Anchoring susceptibility in panels can be detected statistically through correlation analysis, range analysis, and sequence-effect testing

References

Changelog

Date Change
2026-04-29 Compliance review: complete rewrite — added frontmatter, metadata block, all required sections; converted dense AI-format to prose; removed ../wikilinks; applied Australian English; added copyright notice

This article is part of All-About-Coffee.com - The comprehensive coffee knowledgebase.

Copyright © Matthew Clairmont 2026