New

Now in Claude, ChatGPT, Cursor & more with our MCP server

Back to docs
Research Methods

Customer Needs vs. Wants: How to Tell What Users Actually Need (2026)

Customers ask for wants but buy to satisfy needs. This guide explains the difference between customer needs and wants, why building to stated wants leads to feature bloat, and how to uncover the underlying need with Jobs-to-Be-Done, laddering, and AI-moderated research.

Customer Needs vs. Wants: The Short Answer

A customer need is the underlying goal, problem, or outcome a person is trying to achieve — it is stable, often unspoken, and it is why they buy. A customer want is a specific solution or feature they ask for — it is a surface-level preference, easy to state, and frequently wrong about what will actually satisfy the need.

The classic illustration comes from Harvard marketing professor Theodore Levitt:

"People don't want a quarter-inch drill. They want a quarter-inch hole." — attributed to Theodore Levitt (who in turn credited Leo McGivena)

The drill is the want. The hole is the need. And if you only listen to the want, you will keep improving drills while a competitor sells the customer a laser level, an anchor that needs no hole, or a service that hangs the shelf for them. Great products are built by satisfying the need, not by shipping the literal want.

The most-quoted version of this idea — the apocryphal Henry Ford line, "If I had asked people what they wanted, they would have said faster horses" — makes the same point, even though there is no evidence Ford ever said it. Customers describe their desired solution in the vocabulary of what already exists. Your job is to hear past the want to the need underneath.


Needs vs. Wants: A Side-by-Side

DimensionCustomer NeedCustomer Want
What it isUnderlying goal or problem to solveA specific requested solution or feature
StabilityStable over timeShifts with trends and available options
AwarenessOften unspoken or unconsciousEasy to state out loud
Example"Reduce the time I spend on expense reports""Add a bulk-upload button"
Risk if ignoredYou solve the wrong problemYou build features nobody adopts
How you find itProbing, observation, Jobs-to-Be-DoneJust ask — customers volunteer these

The trap is that wants are loud and needs are quiet. Customers hand you wants for free in every feedback form. Needs have to be excavated.


Why Building to Wants Is So Dangerous

When teams build directly to stated wants, they get feature bloat — a long tail of features that individual customers requested but almost no one uses. The data is sobering. The Standish Group's classic feature-usage research found that 45% of features are never used and another 19% are rarely used — roughly two-thirds of everything built. Pendo's product-benchmark analysis reported that about 80% of features in the average software product are rarely or never used.

Much of that waste is a needs-vs-wants failure. A customer wanted a button; the team built the button; the button did not address the actual job, so it sat unused. Meanwhile the real need — maybe "help me trust that this report is correct" — went unmet, and churn crept up for reasons the roadmap never named.

The Kano Model formalizes why "just give customers what they ask for" fails. It sorts needs into three types:

  • Basic (must-be) needs — Expected and rarely articulated. Their absence causes anger; their presence earns no praise. Customers almost never ask for these, yet failing them is fatal.
  • Performance needs — The stated wants. More is better, and customers will happily list them. These are table stakes, not differentiation.
  • Delighters (excitement needs) — Unspoken needs the customer cannot articulate because they do not know the solution is possible. These create loyalty — and you can only find them by understanding the need, never by asking what the customer wants.

Kano's uncomfortable lesson: the features that differentiate you are precisely the ones customers will never request. If your roadmap is just a tally of stated wants, you are building a product with no delighters and, often, unexamined gaps in the basics.


The Jobs-to-Be-Done Lens

The cleanest way to separate need from want is to ask what job the customer is "hiring" your product to do. Jobs-to-Be-Done reframes every feature request as evidence of a deeper goal:

  • Want: "I want dark mode." → Possible job: "Help me use this late at night without eye strain."
  • Want: "I want more export formats." → Possible job: "Let me get my data into the tool my boss actually reviews."
  • Want: "I want faster onboarding." → Possible job: "Help me look competent to my team in week one."

Once you know the job, you can evaluate whether the requested want is even the best way to do it — or whether a different, unrequested solution would serve the need far better. That is exactly the space where category-defining products are built.


How to Uncover the Real Need Behind a Want

You cannot get to needs by asking "what do you want?" — that question only ever returns wants. Use these techniques instead:

1. Ask about the last time, not the ideal future

Anchor every conversation in a real, recent event: "Walk me through the last time you ran into this." Past behavior reveals needs; hypothetical wishes reveal wants. This is the core discipline of the Mom Test.

2. Ladder with "why"

When a customer states a want, ask "why" (or "what would that let you do?") three to five times. Each answer moves up the ladder from feature → benefit → underlying need. The want "add a bulk-upload button" ladders to "because I upload 200 files a week" to "because month-end close is a nightmare" to the real need: "help me finish the close faster and with fewer errors."

3. Watch what people do, not just what they say

Observe the workarounds. A spreadsheet someone maintains on the side, a manual copy-paste ritual, a sticky note on a monitor — these are needs made visible. People under-report the workarounds they have normalized, so behavior often reveals needs their words omit.

4. Separate the ask from the outcome

For every request, capture two things: the want (what they asked for) and the outcome (what they are trying to achieve). Prioritize on the outcome. Multiple different wants often ladder to the same need — and solving the need once can retire a dozen feature requests.


The Modern Approach: Uncovering Needs at Scale with AI

Laddering to a real need takes skill and patience — the interviewer has to notice a want, resist taking it at face value, and probe another layer deeper. Doing that well across enough customers to trust the pattern has traditionally required a trained researcher and weeks of calls. Most teams do not have that, so they default to counting wants in a feedback tool.

AI-native platforms like Koji change the economics. A Koji AI-moderated interview is built to ladder: when a participant states a want, the AI automatically asks why, when was the last time, and what were you trying to get done — the exact follow-ups that separate need from want — with every participant, over voice or text, in parallel.

  • Adaptive probing at scale. Instead of five interviews, you can run the same rigorous, need-seeking conversation with hundreds of customers and let the AI ladder each one to the underlying job. Teams using AI-assisted research report reaching insight in days rather than weeks.
  • Structured questions to size the need. Pair open-ended probing with ranking (which outcome matters most), scale (how painful is this today), and single_choice questions to quantify how widely a need is felt — not just how loudly one customer asked. See the structured questions guide.
  • Automatic thematic analysis. Koji clusters distinct wants that share an underlying need, so "I want bulk upload," "I want an API," and "I want a Zapier integration" surface together as one job: get my data out faster. That is the insight a roadmap can actually act on.

While a legacy survey tool like SurveyMonkey can tabulate which features customers say they want, an AI-native platform like Koji is designed to reveal the need underneath — and it democratizes that skill so any PM or founder, not just a research specialist, can do it.


Common Mistakes

  • Treating the feature request as the requirement. The request is a clue, not a spec. Ladder to the need first.
  • Averaging wants into a Frankenstein product. Ten customers wanting ten different features often share one need — build for the need, not the ten features.
  • Ignoring basic needs because no one asks for them. Reliability, speed, and trust are rarely requested and always required.
  • Confusing loud with important. The most vocal want is not necessarily the most widely felt need. Quantify before you prioritize.
  • Asking customers to design the solution. They are experts in their need, not in your product. Own the solution; let them own the problem.

Related Resources

Customers will always hand you wants. The teams that win are the ones who treat every want as a doorway to a deeper need — and who have the research muscle to walk through it at scale.

Related Articles

Customer Needs Analysis: How to Uncover What Customers Actually Want

A practical guide to customer needs analysis — how to identify, prioritize, and act on what customers genuinely need, with frameworks, research methods, and real-world examples.

How to Identify and Validate Customer Pain Points Through Research

A complete guide to discovering the real problems customers face — using AI interviews, structured questions, and proven frameworks to surface pain points that drive product decisions.

Jobs to Be Done Framework: The Complete Guide

The definitive guide to the Jobs to Be Done (JTBD) framework — its history, two schools of thought, how to write JTBD statements, famous examples, how to conduct JTBD research, and how AI interviews enable JTBD at scale.

The Mom Test: How to Talk to Customers Without Being Misled

Learn Rob Fitzpatrick's Mom Test methodology to ask questions that even your mother can't lie to you about.

Observational Research: How to Learn From What Users Do, Not What They Say

A complete guide to observational research — the family of methods that studies users by watching real behavior instead of asking. Covers the say-do gap, the main observational methods, how to run a study, and how to pair observation with AI interviews to capture the why.

Problem Validation: How to Prove a Problem Is Worth Solving (2026)

A step-by-step guide to problem validation — how to confirm a problem is real, frequent, and painful enough to build for, before you write a line of code. Includes interview techniques, signals to look for, and how to validate at scale with AI.

Structured Questions in AI Interviews

Mix quantitative data collection — scales, ratings, multiple choice, ranking — with AI-powered conversational follow-up in a single interview.

How to Write User Interview Questions That Surface Real Insights

A practical guide to writing user interview questions that uncover genuine insights — covering open vs closed questions, common mistakes (leading, double-barreled, hypothetical), and how Koji's 6 structured question types combine qualitative and quantitative research.

Value Proposition Canvas: The Complete 2026 Guide to Designing Products Customers Actually Want

Master Alex Osterwalder's Value Proposition Canvas — Customer Profile (jobs, pains, gains), Value Map (products, pain relievers, gain creators), the fit test, real examples from Uber and Airbnb, and how Koji populates the customer side from real AI interviews instead of guesses.