A one-week research plan for a vertical you know nothing about
Vocabulary, then workflow, then economics. Run those out of order and you will produce confident conclusions about a job you cannot describe.
The short answer: three passes, in this order — vocabulary, workflow, economics. Fifteen to twenty sources over about a week gets an outsider from nothing to being able to ask a practitioner an intelligent question, which is the actual objective. Evaluating opportunities before you can name the steps of the job produces conclusions that sound confident and are not.
The instinct on entering a new vertical is to look for the gap immediately. It fails for a structural reason: without the workflow in your head you cannot tell an unusual complaint from an ordinary one, and every complaint therefore looks like an opportunity.
The three passes
Each pass has a different question and a different stopping condition. Running them as one blended pass is the same as not running them.
| Pass | Question | Sources | Stop when |
|---|---|---|---|
| 1. Vocabulary | What do these people call things? | 4-5 introductory or day-in-the-life sources | You can read a trade discussion without lookups |
| 2. Workflow | What are the actual steps, in order? | 6-8 walkthroughs of real jobs | You can draw the sequence and name the handoffs |
| 3. Economics | Where does money enter and leave? | 4-6 business-side or long-form conversations | You know what a good week pays and what kills margin |
Fifteen to twenty sources total. The temptation is to skip pass one because it feels like it teaches you nothing — it teaches you the vocabulary that makes passes two and three legible, which is not nothing.
Pass one: the words, and how loosely they are used
Every trade has terms outsiders use interchangeably that practitioners keep strictly separate, and terms outsiders treat as precise that practitioners use loosely. Both mistakes are audible immediately.
Collect three things in this pass: the terms themselves, which distinctions actually matter to the people doing the work, and the informal names for steps that have no official name. The last category is the valuable one — an unofficial name for a step usually means the step is annoying enough that people needed a shorthand for complaining about it.
When a term is used differently by two roles in the same industry, that boundary is almost always a handoff — and handoffs are where the coordination failures live. A vocabulary clash between roles is a free pointer at the seam in the workflow.
Pass two: the sequence, including the invisible steps
Now watch complete jobs, start to finish, in as unedited a form as you can find. The objective is a sequence you could draw: what happens first, what triggers the next step, who hands what to whom, where waiting happens.
The steps that matter most are the ones nobody narrates because they have stopped noticing them — re-keying a number into a second system, taking a photo as a record, texting to confirm something a system already knows. Those are invisible in interviews, because people describe the job they think they do rather than the one they perform, and they are visible in recordings. That gap is the whole argument for watching rather than asking, developed in YouTube versus forums for product research.
- ✗Four clean steps with obvious handoffs
- ✗One system of record
- ✗Decisions made on information at hand
- ✗The painful part is the visible one
- ✓Eleven steps, three of them undocumented
- ✓A system, a spreadsheet and a group chat
- ✓Decisions made while waiting for someone
- ✓The painful part is the waiting, not the work
Pass three: where the money moves
Only now does the opportunity question become answerable, because you can finally locate a complaint inside a process and see what it costs.
Three things to establish: what a job is worth, what the recurring costs are, and which step destroys margin when it goes wrong. That third one is where vertical software usually earns its price — not by making a good day faster, but by making a bad day less expensive.
Two structural details are worth capturing alongside the numbers, because they decide whether a product is sellable at all. The first is who actually authorises spending: in many trades the person suffering the problem cannot buy anything, and a tool designed entirely around their experience has no route to a purchase. The second is seasonality — trades with a compressed busy season buy software in the quiet months and will not take a call during the rush, which changes launch timing more than most positioning decisions do.
Long-form conversational sources carry these numbers more reliably than anything produced, because the specifics arrive as digressions rather than as claims — the property discussed in conference talks and podcasts as research sources. Treat every figure as an anecdote until it repeats across independent voices.
Choosing sources when you cannot yet judge quality
The awkward part of entering a vertical is that source selection requires exactly the knowledge you do not have. You cannot tell a competent practitioner from a confident amateur, and search rankings will not help — they favour production quality, which in most trades is inversely related to being busy with actual work.
Three proxies get you through the first pass. Look for unedited length: someone showing a full job in real time, including the boring parts, is demonstrating something an amateur cannot fake. Look for equipment and environment consistency across uploads, which distinguishes a working practitioner from someone producing content about the trade. And look at who the comments treat as a peer — practitioners correcting each other in specific technical terms is the clearest available signal that you are in the right room.
By the end of pass two you will be able to judge quality directly, and it is worth revisiting your pass-one sources at that point. Discovering that your first four sources were the polished outsiders of the vertical is common, harmless if caught, and quietly damaging if not — the sampling skew described in what AI video research gets wrong.
Knowing when the model is wrong
The point of a week of orientation is to make your first practitioner conversation productive. The signal to watch in that conversation is what kind of correction you receive.
Factual corrections — a wrong number, a mis-ordered step — mean the model is broadly right and improving. Framing corrections, the “that is not really how this works” kind, mean the model is wrong at its base and no amount of additional sources on the same track will repair it. The correct response is to restart pass two with sources chosen from the framing you were just given, which usually costs a day and saves considerably more.
A week of video research makes you a well-briefed outsider, not an insider. It buys better questions and a shorter path to a credible first conversation. Vertical products that fail on domain detail usually failed because someone mistook orientation for understanding.
What to write down
The output that survives is not a report. It is a glossary, a drawn workflow with the handoffs marked, and a short list of moments where money is at risk — each item carrying the source it came from so it can be re-checked when a practitioner disputes it.
Structuring the notes that way also makes them directly usable downstream if the vertical turns into a build, since a workflow with marked handoffs is most of a specification already — the shape argued for in turning research notes into a coding-agent spec.
What the week costs
Fifteen to twenty sources, one project, about a week of part-time attention. As of August 2026 that fits the $19 Hobby plan with its 25 videos a month and 2 active projects.
The tier question arises for people who orient in a new vertical regularly — agencies taking on clients in unfamiliar industries, or founders evaluating several verticals before committing. Pro at $59 carries 8 projects and 80 videos; Studio at $199 carries 20 projects and 250 videos with team seats. Detail is on the pricing page, and the client-facing version of this workflow is in running client market research at agency scale.
One project per vertical, the three passes in order, and a glossary plus a workflow map with sources attached at the end of the week. 7-day free trial.
Closing thought
Outsiders fail in unfamiliar verticals less often from ignorance than from arriving at the opportunity question first, while still holding a four-step mental model of an eleven-step job. A week spent in the right order does not make you an insider. It makes you someone an insider is willing to keep talking to, which is the only thing that was ever available.
Frequently asked
How do you research an industry you know nothing about?
In three ordered passes: vocabulary first, then workflow, then economics. Trying to evaluate opportunities before you can name the steps of the job produces confident nonsense, because you cannot tell which complaints are ordinary and which are unusual.
How long does it take to get usefully oriented in a new vertical?
About a week of deliberate work for most trades and professional workflows — roughly fifteen to twenty sources. That gets you to the point of asking a practitioner an intelligent question, which is the real goal, not to expertise.
What is the first thing to learn about an unfamiliar industry?
Its vocabulary, including which words are used loosely and which are precise. Mismatched terminology is the fastest way to lose credibility with practitioners and the most common reason outsider-built products read as obviously outsider-built.
Why is workflow research more important than market sizing?
Because market size tells you whether a win would be worth having, and workflow tells you whether a win is available. Outsiders usually have adequate access to the first and almost none to the second, so that is where research time earns the most.
How do you know when your understanding is wrong?
When practitioners correct your framing rather than your facts. Factual corrections mean you are close; framing corrections — 'that is not really how the job works' — mean the model is wrong and more sources on the same track will not fix it.
Is second-hand video research enough to build in a vertical?
No, and it is not meant to be. It compresses the orientation phase so the conversations you have with actual practitioners start at a useful depth instead of at the vocabulary. The interviews stay necessary; they just stop being wasted on basics.
What does an orientation pass cost?
As of August 2026 fifteen to twenty sources fits inside the $19 entry plan's 25 videos a month, with 2 active projects. Agencies orienting in a new client vertical every month or two are the case for the $59 and $199 tiers, which carry 80 and 250 videos.