Explainable, or it doesn't ship.
Two mechanisms carry the whole platform: a 0–100 score that always shows its work, and research that is not allowed to make a claim without a source. This page explains both.
One number, three dimensions.
Every provider receives a 0–100 composite score against your specific targeting, summarized into Hot / Warm / Cool / Cold tiers. The composite is built from three dimensions, each answering a different question — and each visible in the score's breakdown, so a rep can always see why a provider ranked where they did. The signals listed under each dimension are representative of what it weighs, not an exhaustive feature specification.
“How well does this provider match the targeting you described?”
- Specialty and taxonomy from the NPI Registry (NPPES)
- Geography and practice addresses
- Procedure and prescribing activity relevant to your query (CMS Part B / Part D)
- Practice characteristics named in your description — ownership, size, setting
“Does their clinical behavior suggest they would actually buy?”
- Procedure volumes and trends in the relevant codes (CMS Part B)
- Prescribing patterns by drug and class (CMS Part D)
- Practice-growth signals and quality-program participation trends (e.g. MIPS Promoting Interoperability)
“Can your rep actually reach them — and are they the decision-maker?”
- Evidence-graded contact points with reach labels and confidence grades
- Practice ownership and decision-authority signals
- Licensure status from state medical boards
Scoring is not static. Models learn from your team's recorded outcomes — which leads convert, which deals close — so prioritization comes to reflect what predicts closed deals for your product specifically. Dimension weights are adjustable per campaign.
Tier labels are capped by evidence, not just cut by percentile. Tiers are drawn from each pool's score distribution with absolute floors underneath, so a weak pool cannot promote mediocre matches into Hot. Two evidence caps bound what a label may claim: a match that fails a state you explicitly named is capped below Warm, and a match our deeper ranking model never evaluated is capped below Hot. When a cap fires, the reason is stored with the score — the label is as inspectable as the number.
Confidence is not fit. Every provider also carries a separate Low / Medium / High data-confidence label — built from identity verification, practice data, enrichment coverage, and dataset freshness — that is deliberately never blended into the 0–100 score. Complete data earns no free points, and a sparse profile is labeled sparse rather than quietly penalized or quietly promoted.
What the score does not claim.
It is not an outcome guarantee.
A 92 means strong evidence of fit, readiness, and reachability — it is not a promise the deal closes, and we publish no win-rate statistics until we have referenceable customer measurements to stand behind.
It is not a black box.
Every score ships with its breakdown. If a provider scores Hot, you can see which signals drove it; if the explanation looks wrong for your market, you can adjust dimension weights per campaign.
It is not a judgment of clinical quality.
The score measures commercial fit for what you sell — never whether someone is a good doctor. Where quality-program data informs a score, it is read as an operational signal (for example, how a practice is trending in its technology adoption), never as a verdict on the quality of care a provider delivers.
It does not invent data.
Fields the platform cannot source are marked unknown rather than guessed. An honest gap beats a confident guess your rep dials.
When research disagrees, the lead says so.
A completed research dossier is the most detailed evidence the platform holds about a provider — so it is not allowed to sit beside the score without being compared to it. Whenever a dossier completes, its verified findings are checked against the score's verdict on the signals your targeting asked for. Where the dossier corroborates the score, the lead is marked Research confirms; where it contradicts it, the lead carries a visible Research disagrees flag — on the lead's detail view in the app, in CSV exports, and to a connected AI assistant — before your rep invests in the call.
The same restraint applies here as everywhere else: when a dossier does not speak to a signal, the platform says nothing rather than guessing, and the flag annotates the score — it never silently rescores the lead. Verdicts come from the dossier's verified, structured findings, and each one records which version of the cross-check rules produced it.
No source, no claim.
Research agents work under one hard rule: a finding that makes a claim about a provider must cite a source. Publications cite PubMed; trial involvement cites ClinicalTrials.gov; speaking and news cite the venue; disciplinary status cites the state board — the agents are not allowed to assert a clean record without an official source to point to. Findings that cannot be sourced are dropped, not asserted.
The same provenance discipline applies to every stored field: verified, inferred, or unknown, with the source and a checked-at date. How each grade is assigned — and what each dataset contributes — is documented on our data page. To see the output this produces, read an example dossier.
Let the agents
do the research.
Your team does
what it does best.
Want proof first? Read an example dossier — sample data, real mechanics.