counterintuitive-tests

Counterintuitive Tests — invert the obvious, then watch behaviour

Two moves: (1) deliberately generate the ideas that sound wrong, because the sensible ideas are already taken; (2) never trust what people say about them — run a cheap, reversible experiment and watch what they do.

Why the obvious idea is the wrong default

Logical optimisation converges: every competent competitor finds the same "best practice," so it stops being an advantage. The remaining opportunities are the moves that don't survive a rational business case — which is exactly why nobody has tried them. "Doesn't make sense" is not a reason to reject an idea; it's a reason it might still be available.

• Red Bull launched a small, expensive can of bad-tasting drink. Every logical signal said fail. The "wrongness" was the positioning — it must do something.
• Raising the price sometimes raises sales (it changes the signal — see costly-signaling).
• Removing features/options/steps often beats adding them (paradox of choice).

Why you must test behaviour, not opinions

People don't think what they feel, don't say what they think, and don't do what they say.

Surveys and focus groups capture post-hoc rationalisations, not causes. Nobody in a focus group asked for Red Bull. Stated preference is cheap talk; revealed preference (what they actually clicked, bought, kept) is the only honest data. So your job isn't to ask whether the counterintuitive idea works — it's to expose people to it and measure behaviour.

How to apply

1. Write the obvious solution down. The thing every competitor would do. This is your "do not just do this" baseline.
2. Invert it on purpose. For each obvious move, generate its opposite and ask "under what circumstances would the opposite be better?" (Charge more. Offer less. Make it slower but nicer. Hide the price. Add friction. Remove a feature.) Keep the ones that aren't insane on inspection.
3. Design a cheap, reversible test. The whole point of a counterintuitive idea is you can't reason your way to the answer — so make the experiment so cheap and reversible that you don't need to. A/B test, a soft launch to one segment, a fake-door, a landing page, one store. Small, fast, undoable.
4. Measure behaviour, not approval. Define the behavioural metric before running (clicks, purchases, retention, completion). Ignore "would you…?" survey answers about the idea.
5. Let surprising-but-real results win. If the illogical version beats the logical one on behaviour, ship it — even if you can't fully explain why. It doesn't have to make sense to work.