Testing a Partner Category Before Betting on It

The starting condition

Gyms were one of several plausible distribution environments for Dokitami's telehealth product, alongside pharmacies, supermarkets, salons, estates, and workplaces. "Plausible" is doing a lot of work in that sentence. Gyms have health-adjacent members, staff who talk to those members regularly, and physical space where a service prompt could sit. All of that is a reasonable hypothesis. None of it is evidence.

The temptation in a growing partnerships function is to treat a reasonable hypothesis as settled and commit real field resource — dedicated outreach time, a tailored pitch, follow-up capacity — to a category because it sounds right on paper. That's an expensive way to find out you were wrong.

The question that mattered

The question wasn't "are gyms a good segment." That's not answerable directly, and asking it directly invites confirmation bias — you go looking for the two anecdotes that support the story you already believe. The answerable version of the question was narrower: can the team reach the right people at gyms, and when it does, does it produce real interest, or does the conversation die on contact?

That's a contactability and early-intent question, not a market-sizing question, and it was deliberately scoped that way. A small, observable test that answers a narrow question honestly is worth more than a large, unstructured effort that answers a broad question vaguely.

What was tried, and what the evidence showed

Thirty-five gyms were called. That's the denominator, and it was chosen to be small enough to run quickly and cheaply, and large enough that the result wouldn't be noise from two or three unusually good or bad calls.

Twenty-six of those calls became conversations — someone was reached and a real exchange happened, as opposed to a call that went unanswered or was deflected immediately. That gap between 35 and 26 is itself a piece of evidence: it says something about how findable and reachable decision-makers are at gyms specifically, distinct from whether they're interested once reached.

Of those 26 conversations, ten gyms were marked interested. Interested here means a specific, defined thing — expressed intent to explore the partnership further — not a signed agreement, not an onboarded partner, not an activated channel. Keeping that definition narrow was the discipline that made the number trustworthy. It would have been easy to round "polite and non-committal" up to "interested" and inflate the result. The system was built to resist that temptation by making interest a distinct, deliberately conservative stage rather than a vibe.

The operating system that carried the test

The gym test didn't need a bespoke process. It ran on the same core partnership infrastructure used for every other category — target list, contact details, outreach disposition, conversation outcome, interest flag, follow-up owner and date — with one adjustment: category-specific prompts. What you say to a gym manager and what qualifies as a relevant next step for a gym are different from what works with a pharmacy or an estate, even though the stage structure underneath stays identical.

That consistency was the actual design choice worth naming. It would have been simpler to run gyms as a one-off experiment with its own tracking sheet, get a quick answer, and move on. Instead the test was built to be comparable — a gym's funnel snapshot sits next to a pharmacy's or a salon's in the same reporting view, using the same stage definitions, so the ten interested gyms mean the same thing structurally as ten interested pharmacies would. Without that, category tests accumulate into a pile of results that can't be compared to each other, and the organisation ends up making channel decisions off gut feel again, just with more meetings behind it.

The category-specific layer sat in the qualification prompts rather than in the funnel stages themselves. For gyms, that meant asking about member volume, whether the gym had a communal noticeboard or screen where a prompt could sit, and who — a front-desk staff member, a trainer, the owner — would be the one actually mentioning the service to members. None of that changes what "interested" means as a stage. It changes what evidence would need to exist before an interested gym could credibly move to qualified, which is the right place for category nuance to live: in the substance of the conversation, not in a parallel set of stage names that would make gyms incomparable to everything else in the pipeline.

What it changed, honestly stated

The direct result of the test is this: outreach to gyms produced conversations at a rate worth noting, and produced an early interest signal from roughly a third of those conversations. That's a real finding, and it's the finding the evidence supports.

What it doesn't tell you, and what I'm not going to claim it tells you, is whether those ten interested gyms went on to qualify, onboard, or activate — whether the channel converted into anything a member of a gym actually saw or used. The test was scoped to answer a contactability-and-interest question, and it answered that question. It wasn't scoped to answer a full-funnel or revenue question, and treating "ten interested" as if it settled the category would be exactly the overclaim the whole design was built to avoid.

The honest way to describe the status of gyms as a channel, based on this evidence alone, is: worth continuing to invest follow-up in, not yet proven as a working channel. Those are different claims and the difference matters more than it might seem, because a lot of commercial storytelling collapses the two.

What I would do differently, and what this generalises to

If I ran this test again, I'd set a qualification checkpoint on the ten interested accounts within a fixed window — say two weeks — so the test produced a second data point automatically rather than requiring someone to remember to chase it. A micro-funnel test that stops at "interested" risks becoming its own kind of vanity metric if nobody deliberately extends it.

The wider principle is one I'd apply to any new channel, category, or market before committing real resource to it: run the smallest test that can produce an honest answer, define your intermediate stages narrowly enough that they can't be gamed by optimism, and resist the pull to let a promising early signal stand in for a result you haven't actually observed yet. Thirty-five calls is a cheap way to learn something true. A department-wide bet on an unvalidated category is not, and the whole point of running the small test first is to make sure you never have to find that out the expensive way.

There's a related discipline worth naming, because it's easy to get backwards: a category test should be sized to the question, not to the ambition behind it. Thirty-five is small enough to run in an afternoon and large enough that a single unusually good or bad call doesn't swing the read. A test of five would have been cheaper still and worth almost nothing — you can't tell signal from noise at that size. A test of two hundred would have answered the same question with more confidence but at a cost that defeats the purpose of testing before committing. Matching sample size to the decision it's meant to inform, rather than to how important the category feels, is most of what makes a test like this worth running at all.