Building a Provider Network: Licensing, Credentialing, and the Capacity Math

Telehealth founders usually discover this the same way. Demand is fine, the funnel works, and then patients in three states start waiting four days for a consult because there are two providers licensed there and both are booked.

Provider capacity is the constraint that surprises people, because unlike servers you cannot add it in an afternoon.

The three things that must all be true

For a provider to see a given patient, all of these have to line up:

1. They hold an active licence in the state the patient is physically in. Not where the patient lives — where they are during the visit.
2. They're credentialed for whatever your model and payers require.
3. They're actually available at a time the patient will take.

Most capacity problems are the first one. You can have plenty of provider hours and still not be able to serve someone in Oregon.

Licensing is slower than you'd like

State licensure runs on each state's own timeline, and they vary a great deal. Some are relatively quick; others take months. Fees, requirements, and processing times all differ.

Two things help:

The Interstate Medical Licensure Compact offers an expedited route for eligible physicians across participating states. It's meaningfully faster than applying to each state separately. Not every state participates and not every provider qualifies, but where it applies it's the difference between months and weeks. Similar arrangements exist for nurses and some other professions.

Licensing services handle the paperwork for a fee. Worth it once you're doing this at volume, because the administrative load per application is real and it's not work your clinical lead should be doing.

The planning consequence: licensing has to run ahead of demand by a quarter or more. If you're planning a state launch for Q2, the applications go in now. Companies that wait until they see demand in a state spend the next four months apologising to patients there.

Credentialing is a separate process

People conflate this with licensing. It isn't the same thing.

Licensure is the state saying a provider may practise. Credentialing is an organisation verifying their qualifications, history, and standing — education, training, licence status, malpractice history, references.

If you're billing insurance, payer enrolment is another layer again, with its own timelines that are frequently measured in months.

Cash-pay models have it easier here, which is part of why so many telehealth companies start cash-pay. But you still want a real credentialing process, because it's your defence if something goes wrong and because your malpractice carrier will ask.

Employed, contracted, or a friendly PC

This is a structural decision that needs proper legal advice, and the right answer varies by state.

Many states have corporate practice of medicine doctrines that restrict non-physician-owned entities from employing physicians or controlling clinical decisions. The common structure in response is a professional corporation owned by a licensed physician, with a management services agreement between it and the technology company.

Whether that's necessary and how it should be structured is genuinely a question for healthcare counsel. Get it right early — restructuring after you have patients and revenue is expensive and disruptive.

What that means for your software: your system probably needs to model the practice entity separately from the platform, with the clinical record belonging to the right one. Building as though there's a single organisation is a common early mistake.

The capacity model you should actually build

Track supply and demand per state, not in aggregate. Aggregate numbers hide exactly the problem you have.

For each state, you want to know:

• Licensed providers, and how many hours they'll actually give you
• Current demand and its trend
• Time to next available appointment
• Licences in progress and their expected dates

The number that matters most is time to next available, by state. It's the leading indicator. It degrades weeks before it shows up in conversion or churn, which gives you time to act if you're watching it.

Licence expiry will bite you

Every licence has a renewal date. Miss one and a provider silently becomes unable to see patients in that state — except your booking system probably doesn't know that, so it keeps offering the slots.

Build this properly:

• Licences stored as data with effective and expiry dates, not as a spreadsheet someone maintains
• Availability queries read those dates, so an expired licence stops producing bookable slots automatically
• Alerts at 90, 60, and 30 days before expiry, to a named person
• Regular verification against primary sources, because a licence can lapse or be restricted between renewals

The automation here isn't a nice-to-have. A booked appointment that legally can't happen is a bad day for everyone.

Provider experience is a retention problem too

Providers leave for the same reasons anyone leaves a badly run job.

The complaints are consistent: bad tooling that adds clicks to every encounter, unpredictable volume, unclear escalation paths, and administrative work that should have been automated.

Concretely, providers stay where:

• The review queue is well-prioritised and doesn't require hunting
• Patient context is assembled before they open the case, not gathered by them
• Documentation is mostly pre-filled from structured intake
• There's a clear path for the case that needs escalating
• Volume is predictable enough to plan a life around

Losing a provider costs you their licences, which is to say it costs you capacity in specific states that takes months to replace. Treat provider tooling as infrastructure, not as an internal admin screen.

Sequencing a state launch

A workable order:

1. Pick states on demand signal, not alphabetically. Where is traffic coming from that you can't serve?
2. Confirm the rules for that state with counsel — modality, prescribing, practice structure.
3. Start licensing 3 to 6 months ahead.
4. Credential and onboard while licences process.
5. Turn on marketing only once you can actually serve people.
6. Watch time-to-next-available and start the next licence before it degrades.

Step 5 is the one companies skip. Running ads into a state where you have one provider produces a great cost per acquisition and a terrible patient experience, and the reviews from that follow you.

---

We build the scheduling, licensure logic, and provider tooling this runs on. See the telehealth build.