The easiest person to persuade is often yourself. You already know which outcome would feel good. You know which explanation protects your reputation, your investment, or your view of another person. Then you collect reasons.
Imagine a founder who believes the business needs a new website. Every confusing page becomes evidence. Every lost sale becomes evidence. A competitor’s attractive homepage becomes evidence. Nobody has checked whether prospects understand the offer, whether the right people are visiting, or whether anyone follows up after inquiries. The website might need work. The problem is that the investigation has been hired to defend the verdict.
That’s the practical danger of confirmation bias: you can do a remarkable amount of thinking without giving your conclusion a meaningful chance to lose. Knowing the term won’t protect you. You need a way of working that makes inconvenient information harder to avoid.
In Peter Wason’s 1960 study of hypothesis testing, participants had to discover a rule governing number sequences. The task exposed a problem with relying on examples that fit an initial guess: several different rules can explain the same examples. A confirming result doesn’t necessarily distinguish your explanation from its competitors. That limited experimental task doesn’t explain every human disagreement, but it makes the testing problem wonderfully clear.
Suppose you think clients leave because your prices are too high. One former client says the service was expensive. Your theory survives. But so do several others: the client didn’t see the value, didn’t use the service, lost their budget, or found a more suitable alternative. An observation can be compatible with your story without being especially diagnostic of it.
My interest is in the difference between a convincing experience and a dependable explanation. Mentalism makes that distinction impossible to ignore. The interpretation that feels obvious to an audience can be very different from the process that produced the experience. In business, the consequences of confusing those two can be expensive.
“People don’t value quality anymore” is difficult to test. It can absorb almost any disappointing result. “Operations directors at companies of this size won’t pay this amount to solve this specific reporting problem” gives you something you can investigate.
A useful claim names the people, the behavior, and the conditions. It also separates what happened from your explanation of why it happened. Five prospects didn’t reply is an observation. They thought the price was ridiculous is an interpretation, unless they actually told you that.
Write those on separate lines. Then list two other explanations that could reasonably produce the same observation. Your email might have gone to the wrong role. The proposed problem might be low priority. The recipient might have been dealing with something more urgent. You’re creating alternatives to compare, not excuses to preserve optimism.
This matters in relationships, too. A delayed reply can become “They don’t respect me” with astonishing speed. The interpretation may eventually prove justified. It still deserves evidence beyond the delay itself. If you need help separating the situation from the story, the same distinction is useful when both sides think they’re right.
A genuine test creates a contrast. If your explanation and every alternative predict exactly the same result, the test won’t tell you much.
Return to the website example. If poor design is the main obstacle, improving the page should help qualified visitors take the next step. If the offer is unclear, a plain page with a sharper explanation might perform better without a visual redesign. If the wrong audience is arriving, neither change may matter much. Each possibility suggests a different observation to seek.
You might interview recent prospects about what they understood, watch someone attempt the inquiry process, or run a small test with a clearer offer. None of those is guaranteed to settle the matter. Together they can expose whether the favored explanation deserves a major investment.
The question I’d put at the top of the page is simple: what result would make us substantially less confident in this explanation? Decide that before running the test. Otherwise, every disappointing result can become a reason to run a larger version of the same plan.
“Would you agree our onboarding is much easier now?” gives someone a conclusion to endorse. “Show me where you got stuck the last time you used it” invites an account of behavior. Both are questions. They produce very different kinds of evidence.
Ask about specific recent experiences rather than hypothetical approval. What did the person try? What happened next? What did they do instead? What did that cost them? Let them describe the event before offering your categories.
Be careful about whose answers you collect. Your enthusiastic customers can explain why the offer works for enthusiastic customers. They may tell you very little about people who walked away. Your friends may be generous with encouragement because they’re your friends. That isn’t dishonesty; it’s a different relationship with different incentives.
For a hypothetical consultant testing a new workshop, useful evidence might include a purchasing manager explaining the approval process, a former buyer describing why a similar program failed, and a prospect declining a specific proposal. A room full of people saying the concept sounds fascinating doesn’t answer the same questions.
Searching for disagreement isn’t enough if you immediately disqualify everyone who disagrees. You can build a beautifully balanced reading list and still treat every favorable source as informed and every unfavorable source as compromised.
Use the same standard in both directions. If a tiny sample is too weak to support a result you dislike, it’s also too weak to establish a result you want. If a customer’s complaint is worth investigating, a customer’s compliment should face similar questions about context and specificity.
Ask someone capable to make the strongest case against your plan. Give them the real constraints and the information you’re using. A person forced to argue from incomplete facts is easy to dismiss, which defeats the point.
Then summarize their case in language they would recognize. You don’t have to agree. You do have to understand what would make their explanation credible. This is one reason listening without immediately correcting someone is a serious thinking skill.
Before a consequential choice, record your main prediction, the evidence supporting it, the strongest alternative, and the conditions that would trigger a review. Keep it short enough that you’ll actually use it. A single page is often plenty.
Suppose you’re introducing a paid advisory service. You expect a certain type of buyer to purchase because a recurring problem is costly and urgent. Write down what a meaningful test would require: conversations with actual budget owners, a defined offer, an appropriate buying period, and clear records of acceptance or refusal.
Also record the limits. A handful of conversations won’t establish the size of an entire market. A failed outreach message won’t prove the problem is imaginary. A successful sale won’t prove your delivery model is profitable. Good evidence answers a particular question; it doesn’t grant universal permission to celebrate.
When the review date arrives, compare the result with the original prediction. Avoid quietly substituting a new definition of success. If the evidence changes your mind, preserve that change in the record. An accurate update is a useful outcome even when it’s uncomfortable.
A team can agree because the plan is strong or because challenging it is expensive. Those situations look similar in a meeting. To distinguish them, change how you collect views. Ask people to write their concerns before hearing the senior person’s conclusion. Ask each person which assumption seems least secure and what evidence they would want before committing.
Imagine a director who has already announced that a new product will launch in June. Asking whether June is achievable may invite reassurance. Asking the implementation team to estimate the remaining work independently gives the director a different kind of input. The estimates may still be wrong, but they’re less dependent on repeating the desired answer.
If someone identifies a real problem, don’t punish them for the inconvenience of discovering it. Thank them for making the decision more accurate, then investigate the claim. This doesn’t mean treating every objection as a veto. It means making the quality of the evidence more important than the comfort of the person who proposed the plan. A process that rewards useful corrections has a better chance of catching errors while they’re still affordable.
You can become so concerned about being biased that you refuse to conclude anything. That creates its own failure: important decisions drift while you search for a level of certainty the situation cannot provide.
Some explanations deserve more confidence than others. Repeated observations, independent evidence, and tests that distinguish alternatives can justify action. You don’t owe a weak claim equal weight merely because someone can imagine it.
Set your standard according to the stakes. Choosing a headline for a small experiment requires less evidence than committing your company’s remaining cash. The practical question is whether you know enough for the next step, with a clear understanding of what could change your view. That connects directly to making decisions with incomplete information.
For your next important decision, write two sentences before you ask for another opinion. The first names what you currently believe. The second names what would make you reconsider. If the second sentence is impossible to write, you may be defending an identity or a preference rather than investigating a claim.
You don’t need to become someone who never gets fooled. You need to make it easier to notice when your explanation has stopped earning your confidence. That’s how you keep your intelligence working for your future instead of serving as legal counsel for your past.