Blog / Method

How to find unmet customer needs without guessing

An unmet need is a purchase that did not happen, of a product that does not exist, often by someone who never became your customer. Your own data cannot contain it. Here is where it actually shows up, and the exact language to look for.

By the Hugo team · Updated August 5, 2026 · 7 min read

Most "find the gap in the market" advice fails at the same point. It tells you to analyse competitors, which shows you what is being sold, and to survey customers, which shows you what people say about options you presented. Neither one contains the thing you are looking for.

Why your own data structurally cannot help

This is worth stating precisely, because it is the reason smart teams keep looking in the wrong place.

Every dataset you own describes transactions inside your catalogue. Sales tell you what people bought from what you offered. Returns tell you what disappointed them about what you offered. Support tickets and reviews come from people who already found you and already bought. All of it is a record of your own assortment being consumed.

An unmet need is the opposite of all of that: demand that produced no transaction, for a product that does not exist, frequently from someone who never entered your funnel at all. There is no row in your database for the thing nobody could buy. Adding another analytics tool does not fix this, because the gap is not in your reporting, it is in what your data can physically record. We wrote about the broader version of that trap in what your bestsellers can't tell you.

Unmet demand does not show up as a number. It shows up as a sentence, written by a person who wanted something and could not find it.

The four language patterns that carry the signal

Unmet needs get expressed in public constantly, in a fairly small set of recognisable phrasings. Search these alongside your category terms across forums, review sections, comment threads and social.

1. Requests

The most direct form, and the rarest. People asking outright for something that does not exist.

"does anyone make", "looking for something that", "why is there no", "I wish someone would", "recommendations for" followed by a spec nobody meets.

2. Workarounds

The highest-value pattern, and the most underused. Someone who has built a workaround has already proved the need is worth effort, which is a much stronger signal than someone who complained.

"I ended up using X instead", "I just modify it myself", "I buy two and combine them", "the hack I use is".

3. Compromise purchases

People who bought something adjacent and are quietly unsatisfied. This is where category-adjacent whitespace hides.

"settled for", "it's fine but", "closest I could find", "does the job I suppose", "good except for".

4. Abandonment

The hardest to find and the most valuable, because these people never appear in anyone's data anywhere.

"gave up looking", "ended up not buying anything", "not worth it at that price", "just went without".

Notice that only the first pattern is a complaint. The other three read as neutral or even positive, which is exactly why sentiment-based tools miss them: a person cheerfully describing their workaround registers as satisfied.

Group by the job, not the product

The most common analysis mistake is clustering findings by the product people mention. Twelve people naming twelve different products look like twelve niche complaints. Grouped by what they were trying to achieve, they are often one clear unmet need with twelve inadequate substitutes, which is a completely different conclusion and a much better one.

Practically: for each piece of evidence, write down the outcome the person wanted in their own words, ignore the product entirely, then cluster those outcome statements. The gap usually appears at this step rather than during collection.

Three tests before you build anything

Finding a plausible unmet need is easy. Most of them are not worth acting on. Apply these in order.

One caution that applies to all three: unsolicited public conversation skews to the extremes. Treat frequency as directional, never as incidence, and do not put a percentage in front of your board on this basis. If you need to know how many people share a need rather than what the need is and why, that is a sizing question and it needs a representative sample.

Doing it at scale

The manual version of this works and is how most teams should start: pick your two or three closest competitors, pull a few hundred reviews and forum threads, tag against the four patterns above, cluster by job. A focused afternoon beats most commissioned studies for this specific question.

The limit is refresh rate. Doing it by hand means doing it once, and unmet needs are not static: they close as competitors ship and open as expectations move. The version that compounds is the one you rerun quarterly and compare.

That is the job Hugo does. It reads live consumer conversation, validates each result against the question rather than against a keyword, and returns the demand language with segments and citations you can open and check. The validation step matters more than it sounds: keyword searches on any category return a large majority of irrelevant matches, and in a run we published, 78% of what came back was discarded before anything useful remained.

Bring a category and we'll show you the gap.

Pick a category you are considering. We will run the demand-language pass live on the call, and you can judge the evidence yourself.