The cost-containment pitch starts with the spend. Facility claims, specialty infusion, advanced imaging, high-cost surgery. The vendor shows where the money goes and a mechanism to move it: a reference price, a home setting, a lower-cost site, a center of excellence. Then the slide that closes the room: a benefit differential. Members who use the program pay less. Members who do not pay more. The model runs on the assumption that the differential moves people.
I have built these differentials into plans as a TPA, sold programs that depended on them, and answered the phone when a member found out what one meant. The gap between the slide and the claims file is where the diligence lives.
Is the differential in the plan document?
A differential is a plan design decision. The employer has to amend the plan document to pay the benefit one way for the program and another way outside it. Employers hesitate. HR does not want to be the reason a long-tenured employee's surgery cost more, and the broker knows it. So the program gets sold with the differential, and the plan gets amended without it, or with a softened version: a waived deductible for those who use it, no penalty for those who do not. The vendor's model assumed a stick. The plan installed a small carrot.
Ask what share of the client base has the full differential in the plan document as sold, what share has a softened version, and what share has none. Then ask for realized savings by tier. If the company cannot split it that way, it does not know which of its clients are producing the number.
Does the TPA enforce it at adjudication?
A differential written into the plan still has to be configured. The adjudication system needs a way to know, at claim time, that this claim was steerable, that the program was offered, and that the member chose otherwise. That usually depends on a prior authorization or a flag set by the navigation team before the service. If the flag is not set, the claim pays at the regular benefit and nobody notices. If it is set and the member complains, the account manager, the broker, or the employer asks for an exception, and the TPA grants it, because the alternative is losing the group.
Ask for the override rate: claims that should have paid at the reduced benefit and paid at the full one. Ask who can grant an exception and whether there is a log. A program whose differential is overridden a third of the time has a third of its model missing, and the vendor may never see it because the override happens inside the TPA.
Will the member do it?
This is the assumption that fails most often, and it fails for reasons that look obvious once named. The member is standing in front of a doctor they trust, at a hospital they have used for twenty years, being told where to go. The program is asking them to say no to that person and drive somewhere else. Most people will not, and the ones who will need someone on the phone with them at the moment of scheduling, not a letter in the mail after.
Home infusion. Moving a specialty infusion out of the hospital outpatient department can cut the cost of an episode substantially. It also requires the member to tell their hospital and their prescribing physician that they want care somewhere else, coordinate a nurse, drug shipment, and supplies, and open their home to a clinical visit. Nurse availability varies by market and can be the binding constraint. Home conditions introduce risks the hospital setting did not have. Ask for the share of identified candidates who converted, the time from identification to first home dose, the cancellation and readmission-to-hospital rate, and nurse coverage by market.
Advanced imaging navigation and stand-alone clinics. The heat map is compelling: freestanding imaging centers and clinics at a fraction of hospital rates, well distributed across the client's population. Then ask whether a member with a scheduled MRI will drive thirty-five miles past the hospital their doctor is affiliated with, to a facility that doctor has never referred to, in order to save money on a deductible they may already have met. Some will. Ask for the redirect rate at the point of scheduling, not the share of members who received an outreach message. Ask what the navigation team does when the ordering physician's office schedules the scan before the member is reached.
Centers of excellence and bundled surgery. Same pattern with higher stakes and travel. The savings per case are large. The candidate has to accept a surgeon they have not met, in a city they do not live in, with a caregiver taking time off. Conversion is low even with a generous differential, and the program's economics depend on a small number of cases actually traveling. Ask for identified candidates, referred candidates, and completed cases by year.
International sourcing. Sourcing specialty medications from outside the United States can cut the cost of a single drug dramatically, and the program looks simple on the slide: enroll the member, ship the drug. In practice the enrollment is paperwork the member and the prescriber both have to complete, the first shipment takes weeks, and the member cannot go without the drug while waiting. So the plan authorizes a bridge fill at the domestic pharmacy. Then the refill is late, or the international pharmacy does not stock that drug or that dosage this month, and there is a second bridge fill. By the third one, the bridge is the benefit, the member has never been asked to switch again, and the modeled savings on that member are zero while the program is still counting them as enrolled. Ask for the share of enrolled members who received their first international fill, the average days from enrollment to first shipment, the refill on-time rate, the bridge fill count per member, and the share of enrolled members whose claims are still paying domestically after ninety days.
Who intercepts the member, and when?
Every program above works or fails at one moment: when care is being scheduled. If the navigation team reaches the member before the appointment is booked, with the authority to book the alternative and handle the physician's office, the differential has a chance. If the first contact is a letter after the claim, the differential is a penalty the member did not understand, and the complaint goes to HR.
Ask how the program learns that care is being scheduled. Prior authorization is the usual trigger, and it is late and incomplete. Ask how many members answer the outreach, how many are reached before booking, and how many are booked into the alternative by the navigator rather than left to arrange it themselves. Ask for the average number of touches per converted member. That number is the real cost of the program, and it does not appear in the savings slide.
What should be in the data room?
Client base split by differential status: in the plan document as sold, softened, or absent. Realized savings by tier.
How the claim is identified as steerable at adjudication, and the override and exception rate with who granted them.
Funnel by program: identified, reached before scheduling, agreed, completed. By year and by market.
Member complaint, access-failure, and bridge-fill volume, and plan sponsor termination reasons recorded at the time.
Navigation staffing, touches per converted member, and clinical capacity constraints (nurse coverage, facility availability) by market.
Realized savings against modeled savings for the same population, restated after reversals and exceptions.
Is any of this a reason to pass?
No. Every program above can work, and the spend it targets is real. These are the questions that separate a company that has run the program from one that has modeled it. The useful signal in the room is how the founders respond to them.
A founder who has lived the bridge-fill problem, the exception queue at the TPA, or the member who would not drive past their hospital will answer with specifics: the rate, what they changed, what it cost. A founder who has a smooth answer for each one, or who has not encountered the problem, is telling you the model has not met the plan yet. That is not disqualifying on its own. It is a finding, and it should change something: the allocation of the round toward the navigation and operations layer where the friction lives, the projections, which usually assume conversion rates the market has not produced, or the decision to invest at all.