Three axes, no black box

Every cohort scored on Scale, Cost, Quality -- independently. No composite rank. You decide the weighting. Every score traces to a source quote.

Why three independent axes

Brokers return a single ranked list. The ranking embeds the broker's weighting -- usually price. The buyer cannot inspect it. Cannot override it. Cannot see why candidate #3 ranked below candidate #2.

viCRO returns three transparent axes. The buyer can filter on scale alone (need n≥500 regardless of cost). Filter on cost alone (under $50K regardless of scale). Filter on quality alone (provenance depth >0.6 regardless of size or price). Or weight any combination.

The axes

Scale

The number that matters is usable_n_for_request, not headline N.

"ADNI has 2,000 subjects" is headline N. "202 subjects have matched blood EPIC methylation + CSF biomarkers" is usable N. Headline N is the cohort's published size. Usable N is the count of subjects that survive filtering by the buyer's criteria -- sample type, longitudinal requirement, disease subset, treatment naivety.

The score includes multi_site_potential -- whether this cohort can be aggregated with others in the wiki using the same platform and protocol.

Cost

Cost decomposes into the links of the procurement chain. Each link that carries a cost gets its own entry. Link types: specimen_source, specimen_fitness, provider, logistics, assay_execution, prior_art, data_delivery. Only include links relevant to the query.

Each link has an estimate, currency, evidence tier (PUBLISHED, DERIVED, or open_question), and a note. If any link is open_question, the total stays open. No invented composites. The buyer sees the known links and the gaps.

Quality

Four sub-axes:

Pre-analytical -- will this sample produce signal? Fasting status, tube type, freeze-thaw cycles, fixation time, cold chain integrity. Domain-specific.
Confounders -- documented exposures that alter the readout. Lipid-modifying drugs, neoadjuvant chemo, recent antibiotics. Also domain-specific.
Platform validation -- has this exact sample + platform combination produced reproducible results elsewhere?
Provenance depth -- fraction of 21 intelligence dimensions covered in the entity. A mechanical metric: covered dimensions / 21.

Each sub-axis gets a one-word verdict: good, partial, weak, or missing. Plus an evidence reference.

Per-axis confidence

Each axis has independent confidence -- high, medium, or low. They are independent. A cohort can have high scale confidence (usable N directly quoted) but low cost confidence (only 1 of 3 legs has a figure).

Scale: high if usable_n_for_request is quoted directly from the source. Medium if inferred from the cohort body. Low if estimated from analogues.
Cost: high if all chain links with costs have PUBLISHED or DERIVED figures. Medium if most do. Low if majority are open_question.
Quality: high if provenance depth ≥ 0.6 and pre-analytical facts are documented. Medium if depth ≥ 0.4. Low otherwise.

Intent changes the interpretation

The three axes stay the same. What changes is the evidence that populates them.

Access intent

The buyer wants existing data. Scale = existing data points matching the request. Cost source leg = data access fee. Quality pre-analytical = existing data QC and batch effects.

Commission intent

The buyer wants specimens for running new assays. Scale = banked specimens of the requested type. Cost source leg = specimen acquisition fee. Cost assay leg = provider quote for the intended assay. Quality pre-analytical = specimen fitness for the intended assay -- freeze-thaw history, aliquot volume, expected yield, storage conditions.

The structural rule -- three axes, never composite, per-axis confidence -- does not change. The content rule -- what evidence populates each axis -- flexes with intent.