Methodology

In sequencing, confidence comes from coverage. A single read is a guess; thirty reads is a fact. The base call you trust isn't the one read once but the one read many times.

ReadDepth applies the same principle to the question of who's running which tools across the genomics ecosystem. No vendor knows the whole field. No analyst has access to enough labs. The map that would tell you what's actually deployed where doesn't currently exist.

We build it from contributor reads. Each is a single observation about a single institution: what it has, what it's looking at, where on the buying journey it sits. None of them, alone, identify anyone. Together, the picture resolves. We call the published artifact the industry genotype.

Each contributor fills a short adaptive form. The form captures: role and institution; geography; technology categories the contributor works with or follows; for each platform, depth of engagement and access mode (lab-owned, institutional core, external service, vendor program, collaborator); for high-stakes capital platforms, purchase trajectory (already installed, hands-on early access, budgeted, actively considering, watching, ruled out) and time horizon; and optional secondhand intelligence about other institutions.

The full taxonomy is canonical and editable — new platforms get added to the data model in seconds, not weeks.

Every claim in the dataset carries one of four confidence labels:

• Observed — directly stated in publicly visible sources: press releases, methods sections, conference abstracts, SEC filings, FDA decisions.
• Inferred — derived from publicly available signals weighted by source quality: institutional core webpages, S10 grant awards, citation networks.
• Human-validated — reported directly by contributors with subject-matter knowledge. May not be publicly citable; updates the model internally.
• Vendor-asserted — claims sourced from vendor marketing or commercial channels. Useful, but bias-tagged.

At signup, every contributor sets a citability default for their own contributions: full attribution, aggregate-only (the default), or confidential. Redaction requests are honored without question.

Secondhand claims about other institutions default to confidential and never appear with public attribution. Anyone reporting on another institution rather than their own is reporting on someone else's plans, and the model treats that with appropriate care.

Individual poll responses on LinkedIn are visible only to the poll creator (per LinkedIn's product). They are aggregated into archetype- and region-level summaries; individual votes are never identified in any output, free or paid.

Aggregate findings published in the free tier require minimum cell density before they appear. We do not show statistics for groups with fewer than five contributors. This protects individual contributors and keeps reported numbers from being spuriously precise.

Anonymous contributions are accepted but carry lower evidence weight than identified ones, because they cannot be triangulated with archetype priors or follow-up clarification.

The dataset is biased toward English-language, US/EU, sequencing- adjacent professionals — those most likely to engage with the founder's existing audience. Coverage in other geographies and adjacent fields will improve as the contributor base grows. Honest acknowledgment of this bias is a feature of the methodology, not a flaw to hide.

Questions, corrections, or redaction requests: alex@readepth.com.