idea-003 · desirability deck
Every slide states a condition that would have to be true, then reports where it stands using the candidate file's own claim and its own confidence tag. A deck never upgrades a tag, invents a number, or recommends anything, and there is no ask slide.
The chair: the assay-development scientist trying to prove a limit of detection at 0.001% VAF, the lab director who signs the validation packet, and the inspector who reads it.
How to read this: every slide is a condition, not a conclusion. A bracketed claim reference points into knowledge-base/candidates/idea-003.md — the number is the claim's position in its ## Claims list, the tag is copied from it and never adjusted here. A [no claim] marker means nothing in the file speaks to the condition at all. See docs/deck-spec.md.
What changed since the 2026-08-24 build: the research lane characterised the field this product would sell into — what the incumbent reference materials actually offer [claim 45: unconfirmed], [claim 46: unconfirmed], what the MRD assays claim as their limits of detection [claim 34: unconfirmed] through [claim 37: unconfirmed], and how reproducibly laboratories can work at low allele fractions today [claim 39: unconfirmed], [claim 40: unconfirmed]. That last pair is the most important thing on this deck and it cuts both ways. Rebuilt from the file, not patched.
Would have to be true: A laboratory validating a tumor-informed MRD assay would have to be struggling with its own contrived controls at ultra-low allele fractions — and would have to be doing this often enough for a panel to be a recurring purchase rather than a one-off.
Where it stands: The demand driver is still the strongest verified fact in the file: CLIA requires laboratories to run control procedures monitoring accuracy and precision across the complete analytic process [claim 7: verified]. That establishes controls are mandatory; it does not establish that in-house controls are inadequate, which is the actual premise and remains an assumption [claim 23: unconfirmed], nor that purchasing recurs [claim 27: unconfirmed]. What is new is evidence that the market is growing quickly: Natera reported MRD unit volume growing by 34,000 tests in Q2 2026 over Q1, the largest sequential increase in its history [claim 43: unconfirmed], and Guardant reported Reveal volume up more than 100% year-over-year for a third consecutive quarter [claim 44: unconfirmed]. Both are vendor press and earnings sources with no connector able to verify them, and the file says so. The segment this candidate would sell to remains the Generator's own narrowing [claim 12: unconfirmed].
What would settle it: Interviews with MRD assay-development leads at named laboratories; CAP proficiency-testing and validation documentation practices for LDT MRD assays.
If it's false: A regulation that mandates controls in general is being read as demand for this control, and the candidate has confused an obligation with a purchase order. Note that growing test volume is demand for the assay, not for the reference material — the inference from one to the other is exactly the step that needs a customer to confirm it.
Would have to be true: A laboratory would have to stop building its own contrived controls and anchor a published LoD claim to somebody else's lot.
Where it stands: The candidate names this itself and names it as an assumption: the idea depends on MRD laboratories preferring a third-party commutable panel over the in-house controls they already build, and on their being willing to anchor a published LoD claim to an external lot — with no connector for laboratory purchasing preference and no primary research behind it [claim 21: unconfirmed]. The second half is the harder ask and is separately on record [claim 25: unconfirmed], because anchoring creates a dependency: when the lot changes, the claim has to be re-established. That dependency is also, on this candidate's own account, the moat [claim 19: unconfirmed] — which means the moat is a switching cost the customer has to agree to bear. One precedent is now on record and it is encouraging for the general behaviour if not for the specific ask: a published tumor-informed monitoring method benchmarked itself on three commercial cfDNA reference standards rather than on in-house material [claim 38: unconfirmed].
What would settle it: Interviews on how labs currently establish and re-establish LoD claims; a look at whether the existing vendors' customers re-buy the same lot lineage.
If it's false: The product is a convenience purchase, priced against in-house preparation cost rather than against the value of a validated claim, and the moat is gone with it.
Would have to be true: The inspector would have to accept it, and the buyer would have to be able to buy it without a capital process.
Where it stands: Inspector acceptance is unbacked [claim 24: unconfirmed] and is the pivot of the whole value proposition — a panel that a CAP assessor does not credit is a reagent, not a validation asset. Purchasing mechanics are likewise unbacked [claim 26: unconfirmed]. There is one adjacent precedent worth weighing: an 89-laboratory external quality assurance programme for cfDNA variant testing ran on reference material with predicted allele fractions of 0.5%-2.5% and found more than 92% of laboratories concordant [claim 40: unconfirmed] — third- party reference material is already the substrate of formal proficiency testing in this field, though at allele fractions two orders of magnitude above this candidate's target. There is also a regulatory hazard sitting on this slide rather than on the feasibility deck, because it is created by who the customer is: marketing a Research Use Only product specifically to CLIA laboratories for validating clinical tests is claimed to be exactly the conduct that defeats RUO labelling [claim 3: unconfirmed], under a 2013 FDA guidance no connector here can read [claim 2: unconfirmed]. The customer this deck is about is the customer that creates the risk.
What would settle it: A regulatory-counsel read of the 2013 RUO guidance against the intended marketing conduct; CAP checklist language on third-party reference materials.
If it's false: Either the packet is not accepted — and the purchase has no rationale — or the marketing that would sell it is the marketing that voids the product's regulatory position.
Would have to be true: The panel would have to cost less than building and characterising controls in-house, and the lot dependency would have to cost less than the validation time it saves.
Where it stands: No price appears anywhere in the file [no claim]; the nearest input is a QC-spend ratio the market connector reports as unverifiable [claim 11: unconfirmed]. The non-monetary cost is the one this deck can state precisely: anchoring a published LoD claim to an external lot [claim 25: unconfirmed] transfers a re-validation obligation to the customer every time the supplier's lot lineage moves.
What would settle it: A labour-and-materials estimate for in-house contrived control preparation at these allele fractions; the existing vendors' list pricing as a comparator — the two vendor product lines are now specified in the file [claim 45: unconfirmed], [claim 46: unconfirmed], though neither carries a price.
If it's false: The lab builds its own, as it does today, and the candidate is selling against free.
Would have to be true: The orthogonal characterisation would have to be credible at the allele fractions that matter — and commutability would have to be demonstrated rather than asserted.
Where it stands: This slide changed the most, and the new evidence is genuinely two-sided. The differentiation is now documented rather than asserted: the two named incumbent product lines both bottom out at 0.1% VAF, LGC/SeraCare's in a vendor-described "plasma like matrix" [claim 45: unconfirmed] and Horizon/Revvity's manufactured from fragmented cell-line genomic DNA spiked into a synthetic plasma matrix [claim 46: unconfirmed] — neither pooled human donor plasma, and neither reaching the 0.001%-0.01% VAF this candidate targets. Meanwhile the assays that would use it claim limits of detection well below that floor [claim 36: unconfirmed], [claim 37: unconfirmed], [claim 34: unconfirmed], [claim 35: unconfirmed]. There is a real, documented gap between what the assays claim and what the reference materials cover, and this product is aimed exactly at it. Against that sits the hardest fact in the file. A multi-manufacturer study across eight NGS platforms found the coefficient of variation increasing as mutant allele fraction fell, with poor repetition below 0.5% [claim 39: unconfirmed] — and the 89-lab EQA programme worked at 0.5%-2.5% [claim 40: unconfirmed]. The field's demonstrated reproducibility floor is therefore roughly 50 to 500 times higher than this panel's target levels. The candidate already names the physical reason: producing material with a stated uncertainty budget at 0.001%-0.01% VAF is Poisson-limited at typical 10-30 ng cfDNA inputs [claim 16: unconfirmed]. Commutability itself remains this deck's own write-back and is nowhere evidenced [claim 31: unconfirmed].
What would settle it: A commutability study against clinical patient samples across the major MRD chemistries; a published uncertainty budget with the Poisson floor stated explicitly rather than worked around — and stated against [claim 39: unconfirmed], which is the number a sceptical scientist will raise first.
If it's false: The panel's lowest levels — the only ones that distinguish it from what is already sold [claim 45: unconfirmed], [claim 46: unconfirmed], [claim 17: unconfirmed] — cannot carry a claim, and the product collapses into the existing category.
Would have to be true: Ten conversations with MRD laboratory leads would have to return the same account: in-house controls inadequate, willingness to anchor to an external lot, and a recurring budget line for it.
Where it stands: No primary research sits behind this candidate; the file is explicit that laboratory purchasing preference has no connector [claim 21: unconfirmed]. The segment's small size is an advantage here that few candidates in this repo have [claim 12: unconfirmed] — a customer set that can be enumerated and called, not sampled. The research lane has made the enumeration easier by naming the assays and their published performance [claim 34: unconfirmed] through [claim 37: unconfirmed].
What would settle it: The calls. Name the labs from the published MRD assay literature and from MolDX-covered test lists, and ask.
If it's false: The product was specified from a regulation [claim 7: verified] rather than from a customer, which is the recurring failure shape this repo's rubric is blind to.
Every condition above with nothing verified behind it.
The five verified claims in this file are all regulatory or definitional [claim 4: verified], [claim 5: verified], [claim 6: verified], [claim 7: verified], [claim 22: verified]. Every one describes the rules the product would live under. None describes a customer, and thirteen new claims since the last build did not change that — the research lane characterised the market and the competition thoroughly, and every line of it landed as unconfirmed, because vendor pages, press releases and earnings calls have no connector in this repo and literature scans cannot appraise a finding.
If only one thing from this chair could be checked: whether laboratories will anchor a published limit-of-detection claim to an external lot [claim 25: unconfirmed].
It survives the week's evidence unchanged. It is simultaneously the purchase decision, the recurring-revenue mechanism, and the moat [claim 19: unconfirmed] — and it is an unusual ask, because it hands a laboratory's own published performance claim a dependency on a supplier's lot lineage. If labs will do it, the switching cost the candidate calls a moat is real. If they will not, the panel is a development-phase convenience with no repeat purchase and no defensibility, whatever the material's quality.
One thing the new evidence does sharpen: the ask is harder than it looked. A laboratory anchoring a claim at 0.001%-0.01% VAF would be anchoring it in a range where the published multi-manufacturer record shows poor repetition below 0.5% [claim 39: unconfirmed] — so the customer is being asked to take a dependency precisely where the field's own reproducibility is weakest.
Naming it is not a recommendation, a gate, or a kill.