Benchmarks
What a verified buyer costs.
Credits per verified buyer, contacts per verified buyer, and the share of accounts that pass each gate. No sprint has completed yet, so every figure below is a design estimate from the pipeline's routing, labelled as such. As sprints complete, each estimate is replaced by the recorded figure from the pipeline's own ledger and the date of the change is noted here.
| Measure | Estimate | How it is counted |
|---|---|---|
| Model calls per account researched | 1 research + 1 copy + 1 review (estimate) | Small model; a strong model only after a reviewer rejection |
| Credits per verified buyer | 3 to 5 (estimate) | Realised credit spend divided by rows that reached verified; shown on your Settings page |
| Contacts per verified buyer | 1 (by design) | One verification per business e-mail |
| Accounts that pass company fit | no estimate; recorded after the first sprint | Official-site evidence with a sourced product fact and a labelled use case |
| Vendor enrichment cost per contact (industry) | $0.01 to $0.50 (published vendor prices) | Hunter at the low end, a fully enriched Clay lead at the high end |
Methodology: a recorded figure is a ledger row (realised credit spend divided by rows that reached verified, and so on), never a projection; until one exists the row says estimate. Counting units never mix: screened, identified, verified, prepared, approved, sent and replied are reported separately.
What we learn from your own ledger
Every first touch is classified by code, at draft time, into one of five opening shapes: a question, a dated trigger (hiring, launch, funding, press, leadership), an observation about the company, a referral that names a mutual context, or a direct offer. When you record a send and later log what came back, GenTool counts, per ideal customer profile, how many first touches of each shape went out and how many got a reply. The source is your receipts and your logged replies, never a model's guess and never a projection.
A reply rate is shown only once a shape has ten sent first touches for that profile; below that the counts are listed and the rate is left blank, because a rate from three sends is noise. Profiles are never summed into one number. Once a profile has a best shape, the copy model receives it as a soft preference for the next draft; every copy rule and the independent review still apply. The figures appear only inside your workspace (Today, get_metrics); this page publishes no customer numbers.