Most prospecting tools flood a freight team with domestic ecommerce stores that look busy but ship parcels, not express air. Freight Sora is the opposite: it separates international-air potential from parcel volume, and refuses to inflate a score without genuine evidence.
It is built validation-first. Before any production scraper exists, the method is tested against known customers and plausible look-alikes. Only a passing gate earns the build. That discipline is the product.
A signal is only as good as the proof behind it. We rank on freight-relevant evidence, not parcel counts.
Confidence is shown separately from TPR. We would rather say “we’re not sure” than fake a precise number.
Every score is an explainable set of reason codes — no black box a rep has to trust blindly.
We prove the method recovers real customers before we build scrapers. The gate decides, not optimism.
No company reaches the board before clearing suppression. Wrong-fit and known-customer noise stays out.
Sensitive carrier data never leaves its safe environment. Only anonymised aggregates ever enter the product.
Freight Sora never ingests identified customer data. Where carrier intelligence informs the model, it is assessed inside the carrier's approved environment and exported only as anonymised, aggregated benchmarks — industry × region × company-size × spend-band — each backed by at least five companies.
A fail-closed PII guard re-checks every imported row and rejects anything carrying a company name, ABN, email, domain, street address, exact spend figure, or a cohort of fewer than five. Nothing partial is written.