Short answer
The first five sites should not all prove the same thing. Give each site a learning job across workflow, training, economics, evidence, or repeat use, then delay the next opening until the evidence from the current sites supports it.
The Inspire Medical Systems case
In the second quarter of 2018, Inspire Medical Systems activated 15 new U.S. medical centers for its implantable therapy for obstructive sleep apnea. That brought the total to 168 centers. In the third quarter, it activated 16 more and reached 184.
Those numbers look like a site-expansion story. The operating model behind them was broader.
Inspire was not selling a device that a physician could start using after a product demonstration. The commercial path joined several groups. A patient was usually diagnosed and managed through a sleep center. An ear, nose and throat surgeon implanted the system. A clinician activated it about a month later and adjusted the stimulation. The patient then returned for a sleep study and later follow-up.
Payment added another step. Inspire worked on positive coverage policies while helping centers obtain prior authorization for individual patients. In the third quarter of 2018, it reported 661 patient prior authorizations, 49% more than in the same period of 2017. In the prior quarter, Aetna had issued a coverage policy for about 22 million members.
The product, patient, referral route, implant, activation, follow-up, and payment process all had to work together. A center could be strong at one part and weak at another.
The evidence job was also spread across sites. Inspire's 2018 annual report discussed an independent study from Thomas Jefferson University Hospital and the University of Pittsburgh Medical Center. The study covered 97 patients. The company reported that mean apnea-hypopnea index fell from more than 35 events per hour to about six after three months at both institutions. Those centers were doing a different job from a newly activated community program learning how to find patients and secure prior authorization.
Inspire ended 2018 with 46 U.S. territory managers. Revenue reached $50.6 million, up 77% from $28.6 million in 2017. The company attributed the increase to greater penetration in existing territories, expansion into new territories, and greater physician and patient awareness. That statement does not separate the contribution of each factor. It does show that opening centers and producing more use inside existing territories were different parts of the commercial model.
This is the useful lesson for a smaller launch team. The first sites are not one ranked list. They are a small portfolio of different operating tests.
Prestige is not a learning job
Most first-site lists are built from a mix of physician interest, brand name, expected volume, geography, and how quickly contracting might close.
Those factors matter. The problem is that they do not say what the company needs to learn.
Five prestigious academic centers may all answer the same question: can skilled specialists use the device with heavy support in a well-resourced environment? That can produce valuable clinical evidence and reference accounts. It may say little about training a community physician, fitting the procedure into a normal schedule, getting a value-analysis committee through the economics, or producing a second and third case without the launch team standing in the room.
A first-site portfolio starts with the gaps in the launch model, not the names at the top of the target list.
Five sites do not have to mean five fixed categories. Start with the unresolved decisions. One site may test whether the normal clinical team can absorb the workflow. Another may test whether the account can defend the economics. A third may show which patients are easiest to identify. A fourth may produce evidence with complete follow-up. A fifth may test whether a distributor can train the account, report what happened, and support a second use.
The mix changes with the product. The rule does not: do not spend two early sites answering the same question while another launch assumption remains untouched.
Give each site a primary job
An early site will always produce more than one kind of information. Giving it a primary job prevents the team from treating every observation as equally important.
| Learning job | Question the site should answer | Evidence to capture |
|---|---|---|
| Workflow | Can the procedure fit into the site's normal operating day? | Room time, setup steps, handoffs, delays, schedule changes |
| Training | Can the team use the product safely without launch-level support? | Time to readiness, support required, errors, retraining triggers |
| Economics | Can the account explain and defend the cost? | Budget owner, payment path, committee questions, case contribution |
| Evidence | Can the site produce usable clinical or operating evidence? | Patient criteria, data completeness, follow-up, protocol deviations |
| Repeat use | What turns one completed case into another? | Referral flow, eligible patients, reorder timing, second-case interval |
Patient fit or channel execution may deserve the primary job instead. A site can test whether the eligible population is findable, or whether a distributor can turn access into training, cases, and usable field evidence. The jobs should follow the launch risk, not the article's number.
A site can have a secondary job, but one should come first. Otherwise a missed case can be explained five different ways and the team learns nothing.
Suppose the workflow site completes its first case but needs two extra hours of room time. That observation should reopen the setup sequence, staffing assumptions, and training materials. It should not immediately be read as a demand problem.
Suppose the repeat-use site performs one successful case and then goes quiet for six weeks. The important question is not whether the physician still likes the product. It is whether there is another eligible patient, a functioning referral route, an approved schedule, inventory, and payment clearance.
The job shapes the review.
Do not confuse access with activation
A signed contract, an interested physician, and a stocked product do not make an active site.
The Inspire case makes this distinction visible because the patient pathway included diagnosis, prior authorization, implantation, activation, and follow-up. The company reported new center activation separately from penetration in existing territories. It also reported prior authorizations as an operating measure.
A launch team should define its own activation gate before naming a site as live. For example:
- contract approved;
- trained team signed off;
- first eligible patients identified;
- payment route confirmed;
- required inventory on site;
- procedure date scheduled; and
- follow-up owner named.
The exact gate changes by product. A capital device may require installation, biomedical engineering acceptance, and a first-case consumable order. A disposable may need par levels and reorder ownership. A diagnostic may need interface validation and a process for acting on the result.
The point is to separate commercial access from operating readiness.
This also changes forecasting. A site in contracting is not an activated site. An activated site with no eligible patients is not a utilization forecast. A first case is not evidence of repeat use.
Each is a different state with a different owner.
Sequence sites around evidence, not the calendar
Launch plans often open sites by month: two in April, three in May, five in June.
That sequence assumes learning will arrive on schedule.
A better sequence uses gates. Site two may open after the first team completes training without a major gap. Site three may open after the setup process fits the target room time. Site four may wait until the payment packet survives review. Site five may wait until one earlier site produces a second case from its normal referral flow.
This does not mean launching one site at a time. It means limiting how many sites share the same untested assumption.
If all five depend on one training model, opening them together multiplies the cost of finding out that the model is incomplete. If all five require the same prior-authorization documentation, one denial pattern can stop the group. If all five need the launch team's only clinical specialist at the first case, the calendar is not a plan.
The first sites should be different enough to test the model and close enough in sequence that one can improve the next.
Build a site learning contract
Write down what both sides are expected to learn and do. This is an internal operating record, not necessarily a legal agreement with the account.
Before a site opens, record seven things:
- Primary learning job. The uncertainty assigned to the site.
- Starting assumption. What the launch plan currently believes.
- Evidence required. What would support or weaken it.
- Owner. Who captures and reviews the evidence.
- Expand trigger. The result that supports repeating the model.
- Stop or reopen trigger. The result that prevents the same assumption being copied to another site.
- Linked decisions. The training, inventory, price, support, evidence, channel, or sequencing choices that depend on the result.
An expand trigger prevents a good first case from being treated as proof of repeatability. A stop trigger prevents the launch calendar from overruling what the site has already shown.
For a workflow site, the starting assumption may be that setup adds no more than 20 minutes. The trigger may be two cases above 35 minutes. The linked decisions may include the procedure guide, specialist coverage, and whether the next community site opens.
For an economics site, the assumption may be that the existing payment covers the disposable cost. The trigger may be a value-analysis request for a budget-impact model or a denial under a common payer. The linked decisions may include price, contracting, evidence, and which payer mix qualifies a later site.
For a repeat-use site, the assumption may be that the champion can identify four eligible patients a month. The trigger may be no second scheduled case within 30 days. The linked decisions may include referral development, inventory, and territory staffing.
This turns the first five sites into a controlled set of commercial tests.
What this does not prove
The public record does not show that Inspire chose its early centers through a five-job portfolio or used a site learning contract. By 2018, it already had far more than five U.S. centers. The record also does not show that activating more centers caused the company's revenue growth.
It does show that commercial progress depended on several distinct parts of a center: physician and patient awareness, reimbursement and prior authorization, implantation, activation, follow-up, and penetration inside existing territories. It also shows independent evidence coming from named implanting centers while the company continued to add other centers.
A first-site plan should preserve those differences before a team has 184 centers to manage.
Do not ask the first five sites to repeat the same test. Give each one a job, decide what evidence should change the plan, and open the next site when the current evidence supports it.
Sources
- Inspire Medical Systems, 2018 Form 10-K, filed February 26, 2019: sec.gov
- Inspire Medical Systems, "Second Quarter 2018 Financial Results," August 7, 2018: sec.gov
- Inspire Medical Systems, "Third Quarter 2018 Financial Results," November 6, 2018: sec.gov
- U.S. Food and Drug Administration, Inspire Upper Airway Stimulation PMA P130008: accessdata.fda.gov
Frequently asked questions
- Why shouldn't the first launch sites all be prestigious academic centers?
- Because five prestigious centers often all answer the same question: can skilled specialists use the device with heavy support in a well-resourced environment? That produces clinical evidence and reference accounts, but says little about training a community physician, fitting the procedure into a normal schedule, or surviving a value-analysis committee. The portfolio should start from the gaps in the launch model, not the names at the top of the target list.
- What is a site's primary learning job?
- The main uncertainty assigned to that site before it opens: workflow, training, economics, evidence, or repeat use. Giving each site one primary job prevents every observation from being treated as equally important — a missed case can otherwise be explained five different ways and the team learns nothing.
- What is the difference between access and activation?
- A signed contract, an interested physician, and a stocked product are access — not an active site. Activation needs an operating gate: trained team signed off, first eligible patients identified, payment route confirmed, inventory on site, a procedure date scheduled, and a follow-up owner named. A site in contracting is not an activated site, and a first case is not evidence of repeat use.
- Why sequence sites by evidence gates instead of the calendar?
- A calendar sequence — two sites in April, three in May — assumes learning arrives on schedule. Evidence gates limit how many sites share the same untested assumption: if all five depend on one training model or one prior-authorization packet, opening them together multiplies the cost of finding out the model is incomplete.
- What is a site learning contract?
- An internal operating record written before a site opens, with six fields: the site's primary learning job, the starting assumption, the evidence required, the owner, the reopen trigger, and the decisions linked to the result. It turns the first five sites into a controlled set of commercial tests instead of a ranked list.