A High Expectation Customer, or HXC, is the most discerning person within your target market: the person who recognizes and enjoys your product for its greatest benefit, and whose judgment other people in that market respect. The framework comes from brand strategist Julie Supan, who described it in First Round Review in 2016.

Her definition, in her own words: "The high-expectation customer, or HXC, is the most discerning person within your target demographic. It's someone who will acknowledge, and enjoy, your product or service for its greatest benefit."

That is a narrower and more useful idea than the one usually attached to the term. An HXC is defined by discernment, not by how much they complain.

The term is sometimes written HEC. Both primary sources and this article use HXC.

Who originated the framework, and who applied it

Supan introduced the HXC in October 2016. First Round's write-up is explicit about authorship: the first step to positioning is "to define your ideal user, what Supan has coined 'the high-expectation customer.'"

Supan was YouTube's first head of marketing and communications, from December 2005 to 2009, and later led positioning work for Dropbox, Airbnb and Thumbtack ahead of their launches.

Rahul Vohra, CEO of Superhuman, is the reason most founders have heard the term. He applied it to product-market fit measurement in his own First Round article in November 2018, and he credits Supan directly: "I turned to Julie Supan's high-expectation customer framework as a tool to do just that. Supan notes that the high-expectation customer (HXC) isn't an all encompassing persona, but rather the most discerning person within your target demographic. Most importantly, they will enjoy your product for its greatest benefit and help spread the word."

Worth stating plainly because the attribution is often reversed: Supan originated the framework, Vohra operationalized it. His contribution was combining HXC profiling with Sean Ellis's "very disappointed" survey question, and using the result to split a roadmap. That contribution is real and it is his. The framework itself is hers.

Discerning is not the same as demanding

This is the distinction that changes what you do on Monday.

A demanding user has high standards and will leave if you miss them. Useful to know about, and worth listening to. But demandingness says nothing about whether that person values the thing your product is actually good at.

A discerning user recognizes your product's greatest benefit specifically. They can articulate what it does for them, in words that other people in the same market recognize. Supan adds a second requirement that is easy to skip: the HXC "needs to be a person who others aspire to emulate because they see them as clever" and insightful. The profile has to be relatable or aspirational inside your market, because part of its job is telling you who to talk to and how to talk to them.

Vohra's two examples show how specific this gets. "Airbnb's HXC doesn't simply want to visit new places, but wants to belong. For Dropbox, the HXC wants to stay organized, simplify their life, and keep their life's work safe."

Neither of those is a description of a difficult customer. They are descriptions of what the product is for, expressed through a person.

	Discerning (Supan's HXC)	Demanding (a different thing)
What defines them	Recognizes and enjoys the product's greatest benefit	Holds high standards and enforces them
What they tell you	What your product is actually for, in market language	Where your execution falls short
Relationship to the market	Others in the segment recognize their judgment	May or may not be representative of anyone
Useful for	Positioning, segment definition, roadmap direction	Quality bar, bug surfacing, support gaps

Both columns matter. Only the left one is the HXC.

An HXC is not simply everyone who said "very disappointed"

If you have run the Sean Ellis survey, you have a group of respondents who say they would be very disappointed if your product disappeared. It is tempting to call that group your HXCs and move on. Do not.

Being very disappointed is evidence, not a definition. The very-disappointed group is where you look. What you are looking for inside it is a recurring pattern: a consistent account of what the main benefit is and who benefits most. That pattern is the HXC profile.

The distinction has practical consequences. Your very-disappointed group will usually contain several different people. Some are there because your product does one thing extremely well for them. Some are there because they are locked in, or because switching is annoying, or because they are early and invested in you personally. An HXC profile that averages all of them together is a blur, and a blurred profile cannot direct a roadmap or a positioning decision.

It also is not automatically your earliest adopters or your heaviest users. Volume of usage and enthusiasm about a specific benefit are different signals.

How to find your HXC profile in survey data

The scoring question tells you how many. It does not tell you who. For that you need the open text.

FitSignal's PMF survey has seven fixed questions, and only the first is scored. Three of the other six do most of the HXC work:

What is the main benefit you receive from using [product]?

What type of person do you think would benefit most from [product]?

What is your job title?

Vohra's note on why the "who benefits most" question is unusually productive is worth keeping in mind: "happy users will almost always describe themselves, not other people, using the words that matter most to them." You get a self-description with the self-consciousness removed, plus the vocabulary that lands with that person.

A workable sequence:

Filter to very-disappointed respondents first. Keep the raw, unfiltered score visible next to whatever you produce from here. You will want it later.
Read the main-benefit answers as a set. You are looking for one benefit that recurs in different words, not a list of features. "I stop losing track of things" and "I finally trust my own notes" are the same benefit.
Read the who-benefits-most answers against it. If the benefit is consistent but the described person varies wildly, you probably have two or more segments rather than one HXC.
Use job title to bound the segment, not to define the person. Title is a filter and a sanity check. It is not a profile, and a title cluster can be an artifact of how you acquired those users.
Write the profile in specific prose, with a name and a day. Supan's method produces a person, not a demographic bracket. Vohra's Superhuman profile ran to several paragraphs about a specific working day.
Check it against people, not only text. Survey text is what respondents chose to type. Interviews, observed behavior, and sales or support conversations are how you find out whether the profile is real.

Step six is not optional politeness. A profile built only from free text can be confidently wrong, and it will be wrong in a direction that flatters your existing product.

If you want the survey wording itself and what each question is for, the seven questions are broken down here.

What to do with the profile

Once you have an HXC profile, Vohra's roadmap split is the most concrete thing to do with it. He devoted half the roadmap to deepening what the very-disappointed users already loved, and the other half to removing what held back the somewhat-disappointed users who wanted the same benefit.

His reasoning for the balance, verbatim: "If you only double down on what users love, your product-market fit score won't increase. If you only address what holds users back, your competition will likely overtake you."

Two things to keep straight about the 50/50 figure. It is a starting heuristic, not a theorem, and Vohra presents it as the balance that worked for him rather than a derived constant. The part that generalizes is the two-lane structure. The exact split is a dial, and it is reasonable to run it at 70/30 in either direction for a quarter if you can say why.

The filter on lane two matters more than the ratio. You are not building everything the somewhat-disappointed group asks for. You are building for the subset of them who named the same main benefit your HXC named, and who are blocked by something specific. Somebody who wants your product to become a different product is not in that group.

More on sequencing that work: how to improve a PMF survey score.

The Superhuman example, accurately

Superhuman is the case study everyone cites, and the numbers get mangled in the retelling. Here is the sequence as Vohra reports it.

Starting in the summer of 2017, Superhuman's initial score was 22%. "With only 22% opting for the 'very disappointed' answer, it was clear that Superhuman had not reached product-market fit."

He then segmented down to the very-disappointed users who loved the product most. That step alone moved the number: "By segmenting down to the very disappointed group that loved our product most, our product-market fit score jumped by 10%. We weren't quite at that coveted 40% yet, but we were a lot closer with minimal effort." That takes them to 33%.

Then the product work: "Within just three quarters of our work to improve the product, the score nearly doubled to 58%."

Two steps, two different mechanisms, and collapsing them is the most common error in the retelling. The move from 22% to 33% is segmentation. Getting from there to 58% is three quarters of shipping.

Segmentation deserves a warning label. A 10-point jump from redefining who counts is genuinely useful information, because it tells you a pocket of stronger fit exists. It is not an improvement to the product, and the same arithmetic will raise almost any score if you cut the denominator narrowly enough. Keep the raw baseline and every segment rule visible, side by side, permanently. If you cannot state the rule that produced a number, the number is not reportable.

On sample size, Superhuman had between 100 and 200 users to poll. Vohra's guidance is that "you start to get directionally correct results around 40 respondents, which is much less than most people think." Ellis puts it differently: a minimum of 30 responses before the survey is directionally useful, and he is much more confident at 100 or more. Those are two different people's rules of thumb from two different sources. Cite whichever you are using, and do not average them into a number neither of them said.

What the framework does not license

The HXC idea is easy to stretch into permission for things it does not support.

It does not license ignoring your non-fans. There is exactly one question where deprioritizing them is defensible: what to build to deepen love in your core. For churn, safety, accessibility, support load, pricing, market sizing, and messaging, the people who are not in love with your product are often the only ones who can tell you what is wrong. "Not disappointed" respondents are a churn signal and a positioning signal even when they are irrelevant to this quarter's differentiation work. Filing them under "ignore" throws away the cheapest research you have.

It does not license a hard reading of 40%. Forty percent is the benchmark to aim for and a useful directional signal. Sean Ellis derived it by comparing results across nearly 100 startups, and he was candid about the line itself: "Admittedly this threshold is a bit arbitrary, but I defined it after comparing results across nearly 100 startups. Those that struggle for traction are always under 40%, while most that gain strong traction exceed 40%." Note the asymmetry in his own sentence. Strugglers were always below it. Only most strong performers were above it. That is not a symmetrical divide, and 39% is not a verdict. More on how to read the number: what the 40% test does and does not tell you.

It does not license treating a survey score as product-market fit. A disappointment score measures stated attitude in one declared cohort at one point in time. Fit needs converging evidence: activation, repeat use at the job's natural cadence, retention, payment and renewal, referrals, and acquisition you can repeat. The survey is the cheapest instrument in that stack, not a substitute for it. More on what else has to be true.

It does not license claims about what HXCs will do commercially. You will read that high expectation customers pay premium prices and expand usage. Those are plausible tendencies, and they may turn out to be true for your product. They are not established properties of the profile, and treating them as given is how a positioning exercise turns into a revenue forecast. Test them.

It does not license a fixed survey rhythm. FitSignal prescribes no resurvey cadence. A first survey is a state, not a baseline. If you want a trend, decide deliberately whether you are measuring newly eligible users or re-asking the same ones, and keep that choice stable, because answers from the same person are not independent readings.

The narrowing decision, stated honestly

Building for a discerning segment does mean some people will like your product less than they would have liked a broader version of it. That is the actual cost, and it is worth naming rather than waving at.

Treat it as a deliberate, reversible choice with a stated price: you are accepting weaker scores in segments you have chosen not to serve well yet, in exchange for a sharper product for the segment you can describe. Write down which segments you are accepting that in, and what would make you change your mind. A narrowing decision you cannot articulate is not a strategy, it is drift, and it looks identical to a strategy on a dashboard.

Where FitSignal fits

FitSignal runs the seven-question PMF survey and breaks the results down by persona and segment, so you can read the very-disappointed group's main-benefit and who-benefits-most answers as a set rather than one response at a time. The word cloud view pairs what that core group loves with the blockers named by somewhat-disappointed respondents who cite the same benefit, and the AI analysis clusters free text into themes ranked by frequency, severity and persona weight, each theme linked back to the verbatim quotes behind it.

It surfaces the material an HXC profile is built from. Writing the profile, and deciding what it means for your roadmap, is still your judgment call, and it should be.