Needs Map
Why spell out needs — and how
Everyone in the room can name their customers' needs. Writing them down so a number can be attached to each one is the part that gets skipped — and it is the part that decides whether anything downstream is worth reading.
A need list is not a deliverable. It is the thing three later decisions rest on, and each of them fails quietly if the list is loose.
It separates what customers want from what you already sell
Written as an outcome — what someone is trying to get done — a need survives your product being wrong about it. Written as a feature, it cannot: it has already assumed the answer, and every score you put against it is really a score for your own roadmap.
It is the only way a gap can be measured
How much it matters and how well it goes today are two different questions, and neither one alone is an opportunity. A need everybody rates critical and already handles well is the price of entry. A wide gap on something nobody cares about is nothing. Only both numbers, on the same need, tell you which you are looking at.
It is what the rest of the case is built on
The value proposition answers a need, the economic value prices one, and the entry case argues that enough people have one. Each of those inherits whatever precision this list has — so a need phrased loosely here is a soft number three tools later, with nothing in between to catch it.
So the work is not listing needs. It is writing each one so that two people would score it the same way, and then saying what stands behind the score.
1 · What a need is
A need is an outcome the customer wants, in the customer's own words. Not a feature, not your product, and not a benefit statement.
The test is whether you could measure it without knowing what anyone sells. “Hold gloss batch to batch after reformulating” is measurable by somebody who has never heard of your product. “Better additives” is not measurable at all, and “higher resolution” is measurable but is a specification — it describes a thing, where the need underneath it is seeing detail in a dark scene without squinting.
“Better additives”
Names a product, and asserts an improvement with no unit. There is nothing here to measure and nothing to disagree with.
The tool says: This names a product or a spec, not what the customer is trying to do with it — try naming the outcome instead (what does it let them do?).
The tool says: “Better” on its own isn’t measurable — better at what? Quicker? Cheaper? Right first time?
“Higher resolution”
A specification rather than an outcome. It is measurable, which is what makes it convincing — but it describes the thing you would sell, not what the customer is trying to do with it.
The tool says: This names a product or a spec, not what the customer is trying to do with it — try naming the outcome instead (what does it let them do?).
“Minimize the variability in gloss from batch to batch”
An outcome, with a unit. Two people would score this the same way, and a competitor's product could satisfy it — which is exactly the property that makes it worth scoring.
Every well-formed need is better at something, so “better” is the parent rather than a seventh option — it is what you say when you have not yet named the unit. Seven units cover what can actually be measured; picking one is what turns a sentence into a claim.
The 7 units, with the sentence each one starts
One unit per need, and a need that honestly spans two is a need to split. There is one importance slider and one satisfaction slider, so a need tagged with three has two sliders describing a blend of none of them.
2 · What the score means
Two ratings, one to ten: how important the outcome is, and how well it goes today. The score is the importance plus whatever gap sits under it.
importance + (importance − satisfaction, when positive)— 0 to 20, both inputs rated 0–10
That is Tony Ulwick's opportunity algorithm, and the reason it doubles the importance is that a gap is only worth as much as the thing it is a gap in. A need rated ten that nobody serves scores at the top; the same gap on something rated four barely registers, and should not.
A need nobody has scored has no score, no band, no dot on the map and no place in the ranking. Not a zero — a zero reads as an answer, where the truth is that nobody has said yet.
The four bands, and where each one starts
Where a score lands is a band, and the four are the textbook's. Above the underserved line is where the opportunities are; the two below it are the ones this tool argues with.
Three places this departs from the textbook — and why
Outcome-driven innovation puts everything below the moderate line in one bucket called overserved, because it is normally applied to a set of outcomes somebody has already filtered. This tool lets anyone score anything, so that bucket collects three different situations — and calling all three over-service is wrong in a way that costs you a real finding.
Table stakes
Importance 7 or above. A need that is already well served here is a price of entry.
It matters, and doing it well wins you nothing — but not doing it loses you the deal. That is a finding worth having, and it is the opposite of what “overserved” suggests. The worked example carries the case exactly: cutting VOCs to clear tightening regulation is rated eight for importance and seven for satisfaction, and reading that as a market that does not want it would be a plain misreading.
Overserved
Satisfaction 5 or above. Below the table-stakes line, this is a market genuinely covering the need.
Under that, the score is only low because moderate importance caps it — nothing about being served three out of ten is over-service. The honest boundary is satisfaction meeting importance; a fixed line was chosen for simplicity and costs exactly one cell, where the two are equal and either reading is arguable.
Low priority
Importance below 4. Nothing else about the need is worth reading.
A wide gap on something nobody wants is still nothing to build on, so the gap carries no information down here and there is no reading of the score worth reporting. It is a floor rather than another diagonal for that reason.
3 · What makes a need believed
The score is two judgements you made. What stops that being the whole answer is a separate reading, computed only from evidence actually logged against the need — and there is no way to argue it up.
It counts categories rather than entries, because five friendly yeses are one category and one point of view, while one interview plus one workaround somebody is already paying for are two independent reads of the same claim. Volume is not corroboration — though two different customers who each went as far as agreeing they would pay are two reads on their own, since what separates those is not how many people spoke but how far each one went.
The four confidence tiers, and what each requires
Four of the seven carry weight the others cannot: an interview, a workaround, a support ticket, a win/loss note. Each of them is somebody who has the need, on the record — and the top rung is unreachable without one, however much desk research stands behind it. A report and a survey are somebody else's read of a population, and a regulation lands on a whole class of products: it proves something must change, never which of your needs is the answer.
Of the four, one is something a customer says when you ask: an interview. The other three are something they did anyway — a workaround they built, a ticket they filed, a deal they walked from. That is why the second kind carries more weight: it is something they did, not something they said.
The four questions that qualify a workaround
A workaround is the strongest of the three, and only as strong as what you know about it. Four questions turn "they use a spreadsheet" into evidence.
What it costs them
Time, people, money — whatever they spend keeping the workaround running. A need with no cost behind it is a preference, not a problem.
Did they build it, or just put up with it
Someone who built their own fix has already spent budget nobody made them spend. Someone who tolerates a bad process is telling you something weaker — that fixing it has not yet been worth the trouble.
How long has it run
A workaround that has lasted years proves the problem is real and durable. It also proves they have lived with it — so whatever finally displaces it has to be a step change, not an improvement.
What would have to change for them to stop
If nobody can answer this, the switching cost is higher than the pain. A blank answer is itself a reading, not a missing one.
The 7 categories a need can be backed in
Customer interviews
What someone told you, attributed. Stated preference — real, but weaker than behaviour you watched.
Observed workarounds
Something the customer already does to hack around the gap — a manual step, a spreadsheet, a person hired to bridge it. Revealed behaviour, not a stated opinion, so it counts as strong evidence — once you can say what it costs them and whether they built it.
Market & desk data
Fill in what the source actually discloses. A figure with no named author, no stated method and no date is a data point, not a source — the read below follows what you enter, so there is nothing extra to tick.
Regulatory triggers
A ban, mandate, or compliance deadline — the strongest “why now” you can cite, since it forces a change on a fixed timeline whether or not anyone asked for it. It lands on a whole class of products, though, so it cannot say which need is the answer: pair it with a customer read.
Support tickets & reviews
Sales win/loss notes
Survey / quantitative data
4 · What the list ends up as
The list ends as one rating — Unmet Needs, on the screen every project carries — and the two numbers behind it are not interchangeable.
Clearing the underserved line opens the top of the scale; what separates a strong rating from the top of it is the confidence tier, not a higher score. That is deliberate: the score is nothing but your own two sliders, so letting it carry the top of the scale would make the strongest thing this tool can say something anyone could type in thirty seconds.
The AI may name a need. It may never score one.
A suggestion arrives with no importance, no satisfaction and no evidence — there is nowhere in its shape to put them. Accepting one adds a name, and every number on it stays yours.
Nothing here is filled in for you.
A found source is staged until you place it on a need, and it arrives with its methodology boxes unticked. The tool will tell you what a source discloses about itself; whether that is good enough is a judgement it does not make.
None of this makes a need real. What it does is make the claim checkable — by you in three months, and by whoever asks where the number came from.