Every bidder proposes on their own terms.
We put them on yours.
Benchify prices every term in every bidder's contract against your client's real claims, then ranks the bids by what the client would actually have paid.
For benefit brokers, consultants, and the self-funded employers they advise.
Contracts are not written to be compared.
They are written to govern one relationship, not to line up beside a competitor's. So the comparison falls to the rate card. One carve-out can outweigh every discount difference in it.
“Plan shall receive a rebate credit for all brand claims adjudicated through the retail network.”
“...excluding claims for products within a manufacturer-designated therapeutic family subject to a separate pricing arrangement.”
The discount was real and the credit was real. Nothing here is hidden. But this is one sentence in one bidder's contract, and every bidder writes its own — so there is nothing the bids can be lined up against. Whoever recommends one will be asked why.
Where the ranking changes.
So the comparison has to run on the one basis no bidder chose: what your client would actually have paid. Bidder A leads on the rate card, and on that basis the order comes out the other way.
Four contract lines and four clauses put a different bidder first.
Bring one client's RFP and we'll build this for it.
Start with one contractEveryone else benchmarks the rates.
We execute the contract.
Rate data tells you what a price is. Contract language decides which price applies, to what, and with what excluded.
Precise, and none of it decides the rows below.
4 of these 5 were settled by contract language, not by the rate exhibit.
Every obligation in the agreement becomes a rule with the clause it came from attached. Your claims run against it, and each determination is recorded: same inputs, same output, every time.
The contract becomes a program, not a prompt.
We verify contract language the way critical software is verified. Each obligation compiles to a rule with its clause attached, and the answer comes from executing that rule against your claims.
A number, and the clause that produced it.
Every valid reading, enumerated and priced separately.
The field the contract computes off, which the extract does not carry.
Compile the contract once.
Ask it anything after that.
The contract only has to be compiled once. After that the ambiguity, the redline and next year's market check all run off the same rules — including which rates and which terms are worth the most to negotiate, against your own claims.
Price the ambiguity
Where language does not resolve to a single rule, we enumerate every defensible reading and price each one. Failure to compute is the finding, not a gap to paper over.
“Rebate credits shall exclude claims for low-cost brand products.”
Rank the bidders on one basis
We execute every bid in the field against the same claims file, not the three there was time to model by hand.
Draft the redlines
Counter-language written into the contract text, with the math attached to every change.
Closes the §2.14 exposure priced above.
Rank the rates worth pushing
We execute the contract against your own claims, then rank every rate tier by what a point of movement repays on this book — not by whether it beats a market table.
Priced on your 27,400 claims. A different book ranks these differently.
Rank the language worth pushing
The same execution ranks definitional scope, control rights and remedies. They are not rate tiers, and they routinely outrank every rate on the page.
Priced on your 27,400 claims. A different book ranks this differently.
Project next year's rates
Proposed discounts and rebate rates run against your own prior-year utilization.
Flag what your data cannot answer
Where a clause computes off a field your extract does not carry, we report it as a finding about the agreement rather than approximating it silently.
Sometimes a substitute field can be validated against the clause. When it cannot, that is the finding.
Four questions. Four different reasons comparison fails.
Three contracts, except medical asks two separate questions: what the providers get paid, and what the administrator charges on top. Each fails comparison for its own reason, and each is answered from the contract text.
The deeper discount can apply to half as many claims.
A's deeper discount is measured on 8.4K claims. B's shallower one is measured on 14.9K.
Every number opens.
Click any figure and descend: from the projection, to the clause that governs it, to the claims it was computed from.
Re-runnable, and identical every time.
The same contract and the same claims produce the same number a year later, which is what makes the trail worth keeping.
| claim_id | drug_class | channel | credit_applied |
|---|---|---|---|
| RX-48291 | Autoimmune biologic | $118.40 | |
| RX-49877 | Autoimmune biologic | Specialty | $124.10 |
| RX-51203 | Oncology support | $96.75 |
Two files: the claims extract and the contracts.
The plan already has both. Nothing has to be requested from the vendor being measured.
The record assembles itself.
Every figure traces to a clause and a claim line, so the documentation of how the decision was reached is a byproduct of making it.
When a plan's payments to a service provider are questioned, the documentation of how that vendor was selected is what gets examined.
The work gets cheaper to run and harder to lose.
Your client gets a better contract. Here is what you get.
Every figure in it traces to a clause and a claim line, so the questions it invites are ones you can answer in the room.
The analyst hours go away.
Repricing a bid by hand takes days. Here the marginal cost of one more bidder is effectively zero, so candidate lists stop being limited by reading time.
More clients, same team.
Senior analyst capacity is what caps RFP volume. Removing manual repricing moves that ceiling without adding headcount.
A reason to be in the room at renewal.
A clause-level comparison and a set of redlines is a concrete, recurring deliverable.
You stay the advisor of record.
A tool you run, not a firm you bring in. Every alternative puts someone else's experts in front of your client.
And after the award.
The same compiled contract verifies whether the vendor delivered what they bid: guarantees reconciled against real claims, not against their own reporting.
Rate benchmarking is table stakes in a finalist meeting. A priced ambiguity report is not.
One contract is enough.
Send what you already have. You get modeled output back before you get a sales call.
59% of employers are making cost-cutting plan changes in 2026, up from 44% two years ago (Mercer, 2025). Vendor selection is the one lever that does not touch the benefit.
You have an RFP in flight
Send one client's claims file and the bidder contracts. You get back the net cost bridge, ranked by what your client would actually have paid.
You have a contract in force
Send the agreement and a claims file. You get back the priced ambiguities, the disadvantageous terms ranked by dollar impact, and redlines you can use. No RFP.
Send one file and one contract.
BAA executed before any data moves. Market check, mid-term renegotiation, or full bid: the mechanic is the same.