Cost model
Vendor projections are built to win business. This one is built to survive a finance review. Set your own assumptions, look at all three scenarios, and pay attention to what discontinuation does to cost per persisting member.
Your assumptions
5,000
25%
Share of census meeting BMI and comorbidity criteria.
7%
The number every projection gets wrong. Real-world first-year uptake commonly lands between 3% and 12%.
$950
$300
Net of admin fees, in the period it is actually received.
$52
45%
Drives everything downstream. A good clinical vendor moves this number more than it moves weight loss.
3-year net cost
$1,760,655
Blended PMPM
$9.78
Across full census
Cost per persisting member
$36,680
48 at 12 months
| Year | New | Continuing | Member months | Net drug | Vendor | Total |
|---|---|---|---|---|---|---|
| Year 1 | 88 | 0 | 818 | $531,700 | $54,912 | $586,612 |
| Year 2 | 35 | 48 | 906 | $588,900 | $52,042 | $640,942 |
| Year 3 | 22 | 46 | 755 | $490,750 | $42,351 | $533,101 |
| 3-year total | $1,760,655 | |||||
Scenario spread
3-year net costConservative3.9% uptake$959,501
Expected7% uptake$1,760,655
High uptake12.6% uptake$3,166,886
Present all three to your committee. A single point estimate gets treated as a forecast; a range gets treated as a decision.
Model assumptions
- Eligible pool: 1,250 members (25% of 5,000).
- New enrollment decays to 40% of year-one volume in year 2 and 25% in year 3, reflecting pent-up demand clearing in the first year.
- First-year members are credited 9.3 months of therapy on average, reflecting staggered starts and mid-year discontinuation.
- Vendor fees are billed on everyone enrolled during the year, including members who discontinue. Confirm this against your own contract — some vendors bill on engagement instead.
- Rebates are treated as realized in the same period as the drug spend. In practice they usually lag a quarter or more, which matters for cash flow.
- Not modeled: avoided medical spend, productivity, or turnover effects. Those are real but arrive later and are far harder to attribute honestly.