Turning a Partner's Yes Into Something a Customer Could Actually See

The starting condition

A number of partnerships across Dokitami's channel mix — pharmacies, gyms, salons, estates, workplaces — existed in a state that looked, from a distance, like distribution. A partner had agreed to mention or refer the service. There was a relationship, a point of contact, and in some sense a deal. What there wasn't, in most of these cases, was any mechanism to know whether that agreement translated into an actual moment where a staff member said something to a customer, or a member saw a prompt, or anything observable happened at all.

This is a specific and easy-to-miss failure mode, because it doesn't look like failure. A signed-up partner shows up as a win on a pipeline report. The absence that matters — no visibility into behaviour, no way to confirm activation happened — doesn't show up anywhere unless someone goes looking for it.

The question that mattered

The instinct, when a partner has agreed in principle, is to move on to the next account. That instinct treats agreement as the finish line. The question that actually needed asking was different and more uncomfortable: what specific, observable action is supposed to happen at this partner's location, who is responsible for making it happen, and how would the team ever know if it didn't?

Without an answer to that question, a referral channel isn't manageable. You can't improve something you can't see, and a passive partnership — polite, real, and invisible in practice — is nearly indistinguishable from no partnership at all, except that it occupies a slot in the pipeline that looks like progress.

What was tried, and what it showed

Reviewing the existing partner relationships against that question surfaced a consistent pattern: willingness to participate did not reliably translate into frontline staff, or gym members, or estate residents, knowing what they were supposed to do or notice. A partner's owner or manager might be fully bought in, and that buy-in would simply not reach the person actually interacting with customers, because nobody had built the step that carried it there.

That's the diagnosis in one sentence: the weak point sat at the handoff between partnership acquisition and partner-side behaviour, not at acquisition itself. Getting the agreement had, in a real sense, been the easy part. Getting the agreement to change what happened on the ground was where the channel was quietly failing.

The operating system built

The response separated the partnership lifecycle into four layers, each with its own operating question: recruitment (is there a credible partner and contact), onboarding (does the partner understand the service and the specific action required of them), activation (has that agreed action actually been put into practice), and reporting (what evidence shows progress, stalling, or repeatability).

The critical design move was refusing to let "recruited" stand in for "activated." Those became genuinely separate, tracked stages inside the wider partnership CRM, rather than shades of the same status. A partner could sit at recruited or onboarded indefinitely and the reporting would show exactly that, instead of quietly rounding up to something that sounded more finished.

SOPs did the actual work of translating a partnership into behaviour. For each setting, the SOP specified where a service prompt should physically sit, who at the partner location was responsible for raising it, and what needed to be recorded as evidence that it had happened. This is the least glamorous part of the system and the part that actually closes the gap — a partnership agreement is an abstraction; a laminated prompt at a checkout counter with a named staff member responsible for it is not.

Activation measurement was tied directly to partner status, so the reporting view could show, category by category, how many recruited partners had reached onboarding, and how many onboarded partners had reached documented activation. That made the passive-referral problem visible as a number rather than a suspicion — a category where partners consistently stalled between onboarding and activation was now flagged by the system itself, rather than discovered by accident months later.

The last piece fed learning back into acquisition. If a category repeatedly produced partners who agreed but never activated, that was treated as a signal to change qualification criteria, onboarding content, or the activation offer itself for that category — not as a reason to simply recruit more partners of the same kind and hope the ratio improved on its own.

The SOPs themselves were written in the language of the setting, not the language of the CRM. A pharmacy SOP specified which staff member — usually whoever worked the counter during peak hours — was responsible for the mention, what exact form the mention took, and where a printed prompt or card was placed so it would actually be seen rather than filed under the till with everything else. An estate SOP looked completely different: it ran through the estate's existing communication channel, whether that was a residents' WhatsApp group, a noticeboard, or the estate manager's own announcements, because that's where residents already paid attention, and a new channel invented for this purpose alone would have competed with habits that already existed and mostly lost. Writing the SOP around the partner's existing behaviour, rather than around what would have been easiest to specify from a desk, was the difference between a document that got followed and one that got signed and ignored.

Ownership at the activation stage was assigned to a named person on the Dokitami side, not just left with the partner, because a mechanism with no one accountable for checking it is a mechanism that quietly stops. That person's job wasn't to perform the activation themselves — the partner's staff did that — but to confirm it was happening and to be the one who noticed, and acted, when it wasn't.

What it changed, honestly stated

What this built is a definition of "active" that means something specific and reportable, in a channel where "active" had previously meant, in practice, "someone at some point agreed to something." I'm not going to claim a final activation rate, a patient-volume figure, or a revenue effect from this redesign, because the underlying work doesn't support those numbers, and stating them would be exactly the kind of overclaim this whole case is arguing against.

What I can say is narrower and, I think, more useful to anyone evaluating whether this kind of thinking is worth applying to their own partnerships: the redesign made it possible, for the first time, to ask which partners had actually changed customer-facing behaviour and which had only changed a conversation. Those are very different assets, and before this system existed, the pipeline could not tell them apart.

What I would do differently, and what this generalises to

Given another run at this, I would build the activation SOP for each category at the same time as the recruitment pitch, rather than after — the two were developed somewhat sequentially here, which meant a few early partners were recruited before their category's activation path had been fully specified, and those accounts needed to be revisited.

The pattern generalises past telehealth distribution, and past partnerships as a function. Any relationship where value depends on a second party changing their behaviour — a reseller agreement, an integration partnership, an employee benefit rolled out through HR — carries the same risk. Agreement is not activation. The two get conflated constantly because agreement is the visible, celebratable moment and activation is the invisible, unglamorous follow-through. Building a system that measures the follow-through, deliberately and separately from the agreement, is the actual work. The handshake was never the hard part.

It's worth naming why this failure mode is so persistent, rather than treating it as a one-off oversight. Agreement produces a natural, satisfying checkpoint — a signature, a call that ends well, a partner who says yes. Activation produces no equivalent moment. It happens, if it happens at all, in small increments at a location nobody from the partnerships team is standing in, weeks after the deal was closed and everyone's attention has moved to the next account. A system has to be built deliberately to go looking for that evidence, because nothing about the natural flow of commercial work will surface it on its own. That's not a criticism of anyone involved. It's just what happens when the moment of commitment and the moment of execution are separated by distance, time, and a change of hands — and it's exactly the gap this system was built to close.