Real Needs vs. Fake Needs
Telling apart what they say they want from what they will actually use.
Tell apart what they say they want from what they will actually use. A fake need isn't the user lying to you; it's the user answering a hypothetical question—and people are always too optimistic about their hypothetical selves.
What you'll run into:
- Everyone says it's great in research, and after launch the usage rate is in single digits
- You spent three weeks on a feature, only to find the person who requested it is the only one using it
- Features keep piling up, but the core metric doesn't move
Where fake needs come from
| Source | What it looks like |
|---|---|
| Poor expression | They can't say what they want, or what they describe is some solution they've seen |
| Over-designing | Buried in details and features, they forget the original problem |
| Wrong assumptions | An unvalidated conclusion pushed out of a report or a competitor's move |
| Tech-driven | A new technology is looking for a problem to attach to. Especially common in this round of AI features |
The fourth is the most common now. "Add an AI assistant to our product" is not a need; it's a solution looking for a problem.
Telling real from fake is hard for one reason: saying "I want it" costs nothing. He doesn't lose anything by claiming a need, and you don't lose anything if he's lying. Only when he has already spent time, money, or effort on it does the claim carry weight—the four sieves are how you ask that weight out.
The four sieves
No user research needed—these four questions get you most of the way there on the spot:
- When did they last hit this problem? If they can't name a concrete time and scene, it doesn't bother them that much.
- How do they cope with it today? A workaround means a real problem; no workaround at all usually means they don't care either.
- Have they spent money or time on it? Already paying a cost is the strongest signal. Saying they need it doesn't count.
- If it existed tomorrow, would they pay first? The harshest one. When they go vague, the answer is usually already out.
All four questions share one trait: they ask about facts that already happened, not hypotheticals. Asking about the past gives reliable answers—that's also the core of The Mom Test: only ask what people have actually done and how, never their opinions about the future.mom-test
You don't need to run all four at once. Two and three alone already filter out most fake needs; save the fourth for the decisions with the most money on the line.
Typical cases
| The need as stated | What's actually going on | What you should really do |
|---|---|---|
| I want to customize the theme color | The requester is a designer; regular users won't touch it | Make the default palette great; skip the settings panel |
| I want to export every format | Nine out of ten users use only one | Make that one format excellent |
| I want an AI assistant | They saw it in another product and can't say what they'd do with it | Find the step they repeat by hand most, and do only that |
| I want free drag-and-drop layout | They moved things once and never again | Get the default layout right; two or three presets are enough |
These four share one shape: users ask for "more possibilities," while what actually improves the experience is "better defaults." Giving options is easier than making decisions, so offering options is often the lazy path.
One rule to judge by: a need has to be falsifiable. If a requirement would count as satisfied no matter what you do, it isn't a need—it's a wish. Score each sieve; if a claim fails more than two of the four, don't build it—it will most likely become a feature nobody uses.