You are considering $200 of shares in Harbor Coffee, our fictional coffee business. Every shop you visit is busy. You like the coffee, and the crowds feel like evidence that the whole company is doing well.
Then you learn that most of Harbor's sales come from packaged coffee. You can check what that means for your idea, or find another glowing review of the shops.
Both feel like research. Only one tests the gap in your reasoning.
A useful belief must be able to change
Confirmation bias means searching for information, or interpreting it, in ways that favor what you already believe. You might accept a friendly review at face value while demanding much stronger proof from a critic. You can read both sides and still grade them differently.
Start with the exact claim: "Busy shops tell me how most of Harbor's business is doing." That gives you something to test. "Harbor is great" gives contrary evidence nowhere to land.
Disconfirming evidence is reliable information that weakens a specific claim. A complaint about a cold latte says little about where Harbor's sales come from. The company's sales mix speaks directly to your claim.
A criticism earns its place by testing your reason for buying. Being gloomy is not enough.
The overconfidence check counted all the results. This check asks whether you will weigh evidence that challenges your idea.
Try a test that could prove you wrong
In psychologist Peter Wason's 1960 rule task, participants saw the numbers 2, 4, 6. They could propose other triples and learn whether each followed his hidden rule.
Suppose your guess is "even numbers, two apart." The actual rule is broader: "any three numbers in increasing order." These two extra triples are examples chosen for this lesson:
| Test triple | Fits my guess? | Fits the true rule? |
|---|---|---|
| 8, 10, 12 | Yes | Yes |
| 1, 3, 8 | No | Yes |
The answer to 8, 10, 12 is yes under both rules, so it cannot tell them apart. But 1, 3, 8 gets a yes when your guess predicts no. That answer defeats your guess without establishing the true rule on its own.
One good test can rule out an answer before you know the right one.
In the original study, six of 29 psychology undergraduates found the rule without first announcing an incorrect one. The study tested students solving a puzzle; it did not measure how many investors have a bias.
Another review of a busy Harbor shop is your 8, 10, 12. For the $200 purchase, the useful test is whether shops account for most sales.
Apply one contrary fact
Harbor reports that 70% of its revenue, meaning sales before expenses, comes from packaged coffee. Shops supply the other 30%.
Your visits describe a few locations in the smaller part of Harbor by sales. More shop visits still leave packaged coffee unexamined. The observation can be true while the conclusion is too large.
The revised belief: "My visits tell me something about a few shops. I still need evidence about packaged coffee."
The visits alone do not justify the $200 purchase. You can leave the order unplaced. Finding a hole in your reason for buying is a useful result, even without a replacement reason.
The 70/30 mix shows where sales come from, not whether Harbor is failing or its stock is expensive. Price versus value explains why a promising business still needs a price check. For further research, the guide to company disclosures shows where to find the facts.
Write the failure story before the result
A bear case is a plausible explanation of how your reason for investing could fail. "The shares might fall" names an outcome; a useful bear case explains what breaks.
You might rescue the Harbor idea with a new claim: "People love the shops, so they will buy Harbor's packaged coffee too." That link needs evidence of its own. A new explanation is a new claim to test.
For its bear case, imagine those customers choosing a rival brand for home. The shops remain popular, but that popularity does not carry over to packaged sales. Liking the café and buying the bag are different decisions.
A pre-mortem, developed by Gary Klein, means imagining that a plan has failed and working backward to possible causes before you act.
For this purchase, imagine needing the $200 for a bill sooner than expected. Your money plan can fail even when the company does fine.
The 70/30 mix already defeats the original claim. An investing journal can preserve it beside the revised version: "My visits tell me about a few shops, not most sales." A later price rise would not make the original reasoning sound.
The same test works with funds. Suppose you believe another fund must add diversification, but its holdings match a fund you own: the same investments in the same proportions. A second fund name adds no new investments. You can reject that reason for adding it without deciding that either fund is bad.
After buying, your purchase price can become another stubborn belief about what a stock is worth.
In short
- A belief becomes testable when you can name evidence that would weaken it.
- A test that rules out your guess can help even before you know the right answer.
- Busy shops can be real evidence with a much smaller reach than your investment story.
- Decide what would change your mind before the share price gives you an outcome to defend.