Mental shortcuts, or heuristics, help you make judgments without examining every possibility. They can be useful, but can also produce predictable errors.[1] A bad outcome alone does not prove bias: a reasonable decision can fail under uncertainty.

Our running example is a teaching scenario, not a reported campaign. Sales rose after launch. Before you credit the campaign, check whether prices, seasonal demand, or measurement changed. Timing alone does not identify the campaign’s effect.[8]

These seven cognitive bias examples follow the gaps between those statements. The research supports the mechanisms; the campaign examples and decision checks are applications of them.

You highlight the campaign’s strongest region and dismiss the weakest as “unusual.” In Wason’s rule-discovery experiment, participants often tested cases compatible with their hypothesis instead of effectively trying to rule it out.[2]

Check: Before opening another report, write what would weaken your claim. Then inspect that evidence with the same care you give favorable results. A contradictory result deserves investigation, not automatic belief.

Your launch forecast becomes the meeting’s reference point. Everyone adjusts it slightly instead of estimating from the available evidence. Anchoring experiments show that initial values can influence judgments even when they are uninformative.[1]

Check: Have reviewers estimate the campaign’s plausible effect independently before seeing the target. Then compare their assumptions. A useful baseline deserves weight; an arbitrary starting number does not.

A striking customer testimonial dominates your impression of the campaign. Availability is the tendency to judge frequency or probability by how easily examples come to mind.[1] The testimonial may be accurate while saying little about the typical customer.

Check: Inspect the relevant population and denominator. How many customers were eligible, exposed, and converted? Use memorable cases to generate questions, then check how widespread the pattern is.

A senior colleague calls the campaign a success. The next three reviewers agree—but all saw that first verdict. In a randomized experiment on a news aggregation site, prior ratings influenced subsequent ratings. Social feedback can shape the judgments it appears merely to summarize.[3]

Check: Collect written interpretations before discussion. Ask whether agreement comes from independent evidence or repeated exposure to one opinion. Agreement can be informative; its origin matters.

You are certain that more spending will reproduce the sales increase. Moore and Healy distinguish overestimating performance, overplacing yourself relative to others, and excessive precision in beliefs. These are related but different phenomena.[4]

Check: Record a plausible range and its assumptions, not just a best guess. Revisit predictions against outcomes across decisions. In this campaign, say which uncertainty remains: attribution, repeatability, or profitability.

After sales rise, you remember having expected the increase all along. Fischhoff found that knowing an outcome increased its judged prior likelihood; participants were largely unaware of that influence.[5]

Check: Save your prediction before launch, including alternatives and doubts. During the review, compare the result with that record. Judge the reasoning using what was available then, and separately ask what the outcome teaches now.

The campaign evidence is weak, but you argue that stopping would waste the work already done. Arkes and Blumer documented an increased tendency to continue an endeavor after investing money, effort, or time.[6]

Check: Ask whether you would fund the next phase today, given its future costs and expected benefits. Existing reusable assets and cancellation costs still matter. Irrecoverable past spending cannot, by itself, justify another round.

“Be objective” is vague. In two experiments, asking people to consider the opposite reduced biased judgments more effectively than general instructions to be fair. That finding supports a specific challenge to your explanation, not a promise that a checklist removes bias.[7]

For campaign attribution, a well-designed randomized comparison can estimate incremental impact. It requires valid assignment, trustworthy measurement, and adequate statistical power; contamination between groups can undermine the comparison.[8]

If randomization is unavailable, another comparison needs explicit assumptions. A simple before/after chart does not become causal evidence because you need a budget decision. You may still act under uncertainty; record that uncertainty.

Observation: Sales increased after launch.
Claim: The campaign caused additional sales.
Alternative: Demand or prices changed too.
Missing evidence: A credible comparison.
Next step: Design a test before scaling.