Insurance or cash: which is cheaper for this prescription?
Every prescription can be paid two ways, through insurance or with a cash discount card, and nobody tells you which is cheaper before you reach the counter. FillSmart compares both for a real Medicare drug plan and always shows the math, not just a verdict.
Problem
Patients can’t tell before the counter whether insurance or a discount card is cheaper, and every tool has a reason not to say.
Insight
It’s not a data gap, it’s an incentive gap. The neutral answer is the product.
Decision
Start with Medicare Part D, where plan data is public and real, and never let revenue depend on the answer.
Result
Working prototype on real government data: drug lookup, plan formularies, cash benchmarks, and a comparison UI.
Both sides know half the answer, and each is paid not to finish it
Cash side
- GoodRx, SingleCare, Cost Plus, TrumpRx
- Know: the cash price
- Won’t show: your plan’s cost
- Earn: when you pay cash
FillSmart
- Your drug, dose, and quantity
- Your plan’s tier and copay
- Where you are in your deductible
- One answer, with the math
Plan side
- Insurer and pharmacy-benefit portals
- Know: your exact copay
- Won’t show: a cheaper cash price
- Earn: when you use the plan
01Who it’s for
02The problem
The most consequential math in a patient’s drug spending is left to them, at the counter, under time pressure.
Pharmacists and comparison articles already tell people to check the coupon price against their copay before using a discount card. That advice is this product’s entire feature set, automated.
03The insight
It’s not a data gap. It’s an incentive gap.
Coupon apps are paid when you pay cash, so they never tell you insurance wins. Insurer portals are paid when you use the plan, so they never mention cash. A tool that tells the truth either way is the product, and that only holds if its own revenue never depends on the answer.
04How I measured success
What I optimized
Correct deductible math on Part D checks, verified against government plan dataPlus a trust metric to publish: how often FillSmart recommends the option that earns it less.The obvious metric I rejected
Revenue per comparisonThat’s the referral model. It rewards steering, the exact behavior that breaks the product.05How it decides
The answer always carries its math
The result layout, with placeholders where a real check fills in numbers.
Pay cash
cash cheaper · your plan · tier n
Cash saves $X on this fill, but paying cash forfeits about $Y of deductible credit. For a maintenance drug, that can change later-year insurance costs.
06Decisions
- Why
- It’s the one place where plan formularies and pricing are public and real, so the math can be checked. Commercial plans come later, clearly labeled as synthetic.
- Instead of
- Covering all four personas from day one on made-up data.
- Cost
- A smaller market at first.
- Why
- Trust comes from seeing both prices, their sources, and their dates.
- Instead of
- One “best price” number, like the coupon apps.
- Cost
- A busier result screen.
- Why
- Charge for convenience like household tracking and price alerts. Any referral fee is disclosed and applied after the recommendation, never in it.
- Instead of
- The referral model, which is easier and pays sooner.
- Cost
- Slower revenue.
- Why
- This app isn’t covered by HIPAA, so the FTC’s health-breach rule is the one that applies. That rule’s first enforcement case was a prescription-pricing app sharing data with ad platforms. No ad tracking anywhere near a drug name, and no account needed to check a price.
- Instead of
- Assuming no HIPAA means no rules.
- Cost
- No ad revenue, ever.
- Why
- A partner API can be cut off without notice. The cash side falls back to the government drug-cost benchmark.
- Instead of
- Building on a single coupon provider.
- Cost
- The benchmark isn’t a retail price, and the app says so.
07What I left out
| Left out | Why |
|---|---|
| Real-time claims | Needs enterprise pharmacy-benefit data access. |
| Live commercial plans | No public equivalent of the Medicare files exists. |
| Prior-authorization status | That data lives inside benefit managers. |
| Drug or dose suggestions | Out by design. It compares payment paths for a prescription a doctor already wrote. |
08Where it stands
Three of four build phases are done. I wrote the product and technical specs, then built each phase with AI coding agents against them: drug lookup on the federal drug-name standard, weekly cash-cost benchmarks, Medicare Part D plan formularies and pricing, the comparison engine, a price-check UI, and nearby pharmacies from the national provider registry. All of it runs on real public data.
Not done: saved history, caregiver profiles, and labeled synthetic commercial plans. Whether a paid version needs an insurance license in some states is a question for a lawyer, not one I’d guess at.
09What’s next
- Test it with Part D enrolleesDo people trust a recommendation that sometimes says “use your insurance”?first
- Add the annual viewFor maintenance drugs, the right answer is a strategy for the year, not one fill.then
- Get legal clarity before chargingState insurance licensing could apply to paid recommendations.before revenue