No fee is disclosed in the Doximity-UpToDate announcement. A precise claim that the agreement is worth a particular number would therefore be unsupported. A useful alternative is to show how different commercial assumptions produce different costs, and identify which missing variables matter most.
Every price, population and usage figure in the model below is invented for sensitivity analysis. These are not reported deal terms, market-rate quotations, forecasts or estimates of the actual payment. They illustrate the mechanics of a possible contract, not the contents of this one.
Define the cost being modelled
The model concerns an annual supplier charge for an embedded service. It excludes any pre-existing hospital subscription, the platform's own engineering payroll, taxes and the rest of the clinical AI product. That boundary prevents a content-service estimate from being mistaken for the total cost of a deployment.
The UpToDate Connect product description makes entitlement-based integration a relevant architectural possibility. It does not supply a public rate card for the Doximity agreement. The broader deal-structure analysis explains why existing content licences and incremental delivery charges should be separated.
Here, a deliberately hypothetical hybrid contract includes a fixed fee, an active-user fee and a billable-query fee. A real contract might use only one of those components, bundle them or credit one against another.
The equation and its assumptions
Define annual charge as:
Annual charge = fixed annual fee + (active users x annual user fee) + (billable queries x query fee)
For all three examples, assume a fixed annual fee of $150,000, an annual fee of $36 per active user and $0.05 per billable query. Assume a separate, one-time implementation charge of $150,000.
The active-user count is a stable average over the model year, not the number reached on the final day. Queries are eligible requests to the hypothetical supplier service, not all prompts, all app openings or all activity across the platform. Constant averages are used to make the arithmetic transparent; actual adoption would require a monthly model.
Three hypothetical deployments
Illustrative inputs only; none of these populations or prices describes the announced rollout. All amounts are USD.
| Model input or result | Smaller deployment | Mid-scale deployment | Larger deployment |
|---|---|---|---|
| Eligible clinicians | 5,000 | 20,000 | 100,000 |
| Average active proportion | 40% | 50% | 60% |
| Average active users | 2,000 | 10,000 | 60,000 |
| Billable queries per active user per month | 12 | 20 | 30 |
| Annual billable queries | 288,000 | 2,400,000 | 21,600,000 |
| Fixed annual fee | $150,000 | $150,000 | $150,000 |
| Annual user-fee component | $72,000 | $360,000 | $2,160,000 |
| Annual query-fee component | $14,400 | $120,000 | $1,080,000 |
| Total recurring annual charge | $236,400 | $630,000 | $3,390,000 |
| First-year charge including implementation | $386,400 | $780,000 | $3,540,000 |
The lesson is the sensitivity to scale, not the apparent precision of the totals. They are arithmetic consequences of chosen assumptions. They do not establish a credible lower or upper bound for the actual agreement.
Which assumptions drive the result?
In the mid-scale example, increasing the annual active-user fee by $10 adds $100,000. Increasing the query price by one cent adds $24,000. One additional billable query per active user each month adds $6,000 a year.
Those sensitivities show where diligence should focus. If the contract charges for all enabled clinicians rather than active users, changing the denominator can matter more than negotiating a small reduction in the query rate. If every follow-up is separately billable, conversational use can produce a different bill from counting only initial questions.
The fixed fee also changes effective unit economics. In these invented examples, recurring cost per average active user is $118.20, $63.00 and $56.50 respectively. The falling figure reflects fixed-cost spreading despite higher assumed usage in the larger deployment.
Minimum guarantees must not be counted twice
Suppose a contract instead has a minimum annual usage commitment. If that minimum is creditable against metered charges, the relevant component is the greater of the minimum and actual eligible usage charges.
For example, the formula could be usage charge = max(minimum commitment, metered usage). Adding both in full would overstate the bill. A separate platform fee might still apply, but that would need to be an additional contractual component.
An advance payment creates another distinction. It can affect cash timing without representing an extra economic charge. Similarly, a three-year contractual commitment is not the same as annual revenue, first-year cash expenditure or the present value of the agreement.
Why retail subscription arithmetic is misleading
Multiplying a public individual subscription price by Doximity's total membership would ignore eligibility, existing institutional licences, negotiated rights and actual usage. It would also assume that every member receives the same product and that wholesale pricing equals retail pricing.
The relevant population is the covered contractual population. That might be licensed seats, activated accounts, monthly active users or a defined enterprise cohort. Each produces a different interpretation of the same platform's reach.
The same problem affects query counts. A provider who uses a communication or scheduling tool is not automatically a billable user of an embedded reference service. Broad engagement figures cannot substitute for the missing contract definition.
Model-service prices are not content-licensing prices
A low technical cost for generating text does not establish what maintained clinical content should cost. The negotiated service might include editorial maintenance, access permissions, source delivery, support and contractual obligations.
Conversely, a premium content brand does not prove that the supplier receives a large payment for every generated answer. Existing licensing relationships, strategic distribution value and reciprocal benefits may reduce the incremental charge.
Without the rights schedule and commercial terms, neither inference is reliable. The content-rights analysis explains why retrieval, display, storage and model-training permissions are different commercial questions.
A disciplined conclusion is more useful than a rumoured number
The agreement could support modest incremental payments or substantial recurring charges under different rights and deployment assumptions. The public record does not establish which scenario applies. The right questions concern the payer, denominator, metering rules, commitment and scope.
For clinicians, the commercial model matters when it affects access or continuity. For a focused platform such as iatroX, it also illustrates why product value should not be judged solely by the size of a content contract. Useful regional reference and learning can be assessed directly, rather than inferred from an undisclosed deal price.
Frequently asked questions
Is $630,000 an estimate of the Doximity-UpToDate annual fee?
No. It is the result of one deliberately invented scenario. None of the user counts, prices or fee components has been established as a term of the actual agreement.
Could the actual cost be outside the modelled figures?
Yes. Different rights, minimum commitments, reciprocal arrangements, volumes and pricing structures could produce very different amounts. The scenarios are not a valuation range or a negotiation benchmark.
What information would make a real estimate more defensible?
The covered population, billing unit, access rights, usage levels, guaranteed commitments and treatment of existing licences would be essential. A disclosed comparable would also need genuinely similar scope.
Compare the practical value of iatroX's reference and learning tools against the tasks you actually need to complete.