Spotting the constraints that quietly kill a v1

The short answer: watch practitioners work and mark every point where the workflow bends around something external — an approval, a retained record, a mandated format, a required order — because those bends are the constraints a first release has to respect. Twelve to twenty independent sources is usually enough to map them.

Software built for an unfamiliar vertical tends to fail on a requirement nobody wrote down. The features work, the interface is better than the incumbent’s, and then a prospect explains that the approval has to be recorded with a name and a date, or that the export has to be in a format their auditor accepts, and the elegant model does not accommodate it.

Constraints show up as bends in the workflow

A constraint rarely announces itself. It appears as a moment where a competent person does something that looks inefficient, and does it anyway, without complaint. That combination — visible extra work, no irritation — is the reliable tell, because people only accept friction calmly when they know it is required.

The verbal marker is equally consistent. “We have to” precedes a constraint; “I like to” precedes a preference. Sorting your notes on that distinction alone removes most of the ambiguity in a first pass.

The retention row has the longest reach into architecture. A vertical where records must survive for years cannot be served by a product whose deletion story is “we remove it when you cancel”, and retrofitting that later touches the data model, the billing logic and the export path at once.

Ranking constraints by whether they bind

Not every constraint belongs in a v1. Rank each by three things: how often it recurs across sources, what happens when it is broken, and how much manual work people currently spend satisfying it. High on all three is a headline feature; low on all three is a note in the backlog.

The manual-effort dimension is the commercially interesting one. A constraint that everyone satisfies with a spreadsheet, a shared folder and a reminder is a constraint the incumbent has pushed onto its users, and absorbing it is the clearest value a vertical product can offer.

Whose videos to trust

Source selection matters more here than in almost any other research pass. Vendor content actively suppresses constraints, because a demo that pauses for an approval step is a worse demo. Practitioner content shows them, because leaving them out would make the walkthrough wrong.

Training material, conference talks and internal-process explainers are unusually dense sources: their entire purpose is to teach the rules to somebody who does not know them yet. Those formats are covered in using conference talks and podcasts as research sources, and they earn their place in a constraint map even when the production quality is poor.

Entering a vertical you do not know

Constraint mapping is really the first phase of entering an unfamiliar market, and it sits alongside learning the vocabulary and the buying structure — the sequence in building a research plan for an unfamiliar vertical. It also changes who your user is: where an approval exists, there are at least two roles with different needs, which the ICP work in defining your ICP from video research has to account for.

Agencies and consultants working across several clients are the best single source in a constrained vertical, because they have to satisfy the same rule in multiple organisations and therefore talk about the rule rather than their own habit. That is part of why the pattern in client market research for agencies generalises so well.

Constraints move, slowly and then suddenly

Rules change on someone else’s schedule, and practitioner content reflects the change fast because the people affected start explaining it to each other immediately. A quarterly re-read of the same sources catches a shift early enough to be a roadmap item rather than an emergency.

Keeping the source attached to every constraint you record is what makes that re-read cheap. A note saying “approval required” cannot be re-verified; a note pointing at the moment somebody explained the approval, with the date, can be checked in a minute — the same durability argument as making research survive contact with your team, and the reason the recurring habit in monitoring a niche with recurring research pays off here more than anywhere.

Constraints are the moat, not the tax

Founders tend to treat vertical constraints as overhead: work that has to be done before the interesting product exists. That reading is backwards in most regulated categories, because the constraints are precisely what a horizontal competitor cannot be bothered to implement.

A general-purpose tool serving forty industries will not enforce your vertical’s approval sequence, because doing so would break the other thirty-nine. It offers configuration instead, and configuration means every customer rebuilds the rule by hand and maintains it forever. Shipping the rule as a default is a small piece of engineering that a much larger company is structurally unable to copy.

That asymmetry also shapes the sales conversation. Demonstrating that you already know the constraint — using its real name, in the right order — does more for credibility than any feature comparison, because it signals you have talked to people like the buyer before.

Two ways this pass goes wrong

The first is mistaking one organisation’s internal policy for an industry rule. Plenty of observed friction is a particular employer’s caution rather than a requirement, and building a rigid product around it makes you unusable everywhere else. Recurrence across unaffiliated sources is the test; a single confident practitioner is not.

The second is over-reading the consequence. Not every rule carries the same weight, and treating a bookkeeping convention with the same severity as a genuine regulatory requirement produces software that feels obstructive. Ranking by what actually happens when a rule is broken keeps the enforcement proportionate — and where a rule turns into a claim on your website or a promise to a customer, it needs verifying with the relevant authority or a qualified professional rather than with a transcript.

What the pass costs

Twelve to twenty practitioner sources, watched for bends rather than for features, produces a ranked constraint map and a list of things worth checking with a professional. As of August 2026 that fits the $19 a month plan with 25 videos and 2 projects; $59 covers 80 videos and 8 projects, and $199 covers 250 videos, 20 projects and 3 seats. See the pricing page.

Closing thought

Nobody in a regulated vertical will list their constraints for you, because to them the constraints are not notable — they are just how the work is done. That is precisely why watching the work beats asking about it.