Find out what your device would take to prove, before you hire the team or commission the study.
A costed evidence plan, the assumptions it rests on, and a range that admits what it does not know. Every priced line opens to a quantity, a rate, a base year and a source.
Or read a worked example → Assess existing evidence
No account· No card· No regulatory class required to start
A costed plan that says which parts it cannot price
Cash to the milestone you pick, with staff time counted separately and unpriced work named rather than dropped. A plan with a gap in it says “modeled subtotal” and lists what is missing; it never quietly reads as complete.
How much of it is still open, named rather than implied. The plan says how wide its range is and which line makes it that wide, and every priced row says what it rests on — a measurement, a borrowed figure, or our own assumption — so you can see which numbers would move if somebody checked them.
What a different choice would cost. Change the evidence route and see the arithmetic openly, including when the clever route costs more — model development, credibility work and compute can exceed what the physical testing saved, and this says so.
Change the claim, the population or the market instead and you still get the comparison — labeled a different development path, with what you gave up named. What you do not get is a saving, because a cheaper different product is not the same product for less.
Three people read the same report and need different things
The founder before the team. You know what the device does. You do not yet know what proving it costs, or which of your choices is the expensive one. Arrive with no class, no study design and no vocabulary.
The investor reading the deck. There is a number on a slide. This shows what it rests on, which parts are sourced and which are the founder's own assumption. It supports a diligence conversation; it is not completed diligence and it does not score the company.
The specialist being asked in. A frozen version and a named question, instead of a blank intro call.
Short list
Free, permanently. The planner, the full number, every comparison, and the report and its assumptions appendix, exported and yours. No card.
Evidence-strategy review, quoted per question. A specialist answers the specific uncertainty your own report identified, against a frozen version of it. You see the question, the scope and the fee before anything is committed. Sending an inquiry is not a purchase.
Why would I trust a number an AI produced?
You should not, so it does not produce one. A model reads your description and structures it, and retrieves reviewed cases. Every total, difference and schedule is arithmetic over parameters you can open. The model cannot author a rate, a quantity or a rule.
What deterministic arithmetic does not do is validate a guessed rate. That is why every figure carries its own basis — measured from a named source, borrowed from a comparable case, our planning assumption, or your own input — and why a total is never reported as stronger than its softest input.
What if my device is not one you cover?
You get the qualitative roadmap and the work we can support, with the rest visibly unpriced. Where a whole family has no credible basis we stop publishing totals for it until one exists. A partial plan is a valid thing to take; a complete-looking invented one is not.
Does simulation really cut my evidence costs?
Sometimes, and the tool shows you when it does not. We never transfer a percentage from someone else's device to yours.
Does this replace a regulatory consultant?
No, and it is designed to tell you precisely when you need one. It produces directional planning for founder and investor conversations. It does not predict clearance, produce a submission, or clear any review obligation.
Then you want the review workbench, not the planner: a governed assessment where every factor gets a human verdict against an anchor in the source, or a recorded reason it did not.