Mailer per-lead research v2 — Exa Websets + Apollo enrichment fusion

Cold email lives and dies on the first sentence. Mailer v1 generated that first sentence from whatever LinkedIn + company-website snippets were available at extraction time. It worked, but the personalisation ceiling was limited by the depth of the available context. Mailer v2 raises the ceiling by running a real research step on every lead before the email is written.

The new pipeline

• Apollo enrichment provides the structured signal — title, seniority, tenure, team size, funding stage.
• Exa Websets runs a per-lead web search scoped to that person + their company, ranked by recency and source authority.
• A fusion step deduplicates and reconciles the two streams, producing a 4-paragraph dossier per lead: who they are, what their company does, what changed recently, what an outreach hook might be.
• The email-writing model consumes the dossier (not raw search snippets) and produces the first touch. Mailer logs the dossier alongside the sent email so a human can audit what the model saw.

Measured impact

On our internal book of business, reply rate on the first touch lifted from 8.2% to 14.6% — a 78% relative improvement. Open rate barely moved (the subject-line generator did not change), but reply quality went up sharply: more inbound questions, fewer one-word "unsubscribe" replies.

Cost

Each research step costs roughly $0.018 per lead at current Exa + Apollo pricing. Mailer v2 caches dossiers for 30 days so the same prospect across multiple campaigns is researched once. The cost is rolled into the existing per-lead Mailer credit price — no new line item.

How to enable

Already on. v2 became the default for all new campaigns on 2026-02-08. Existing campaigns mid-flight kept v1 to avoid altering their measured baseline; new campaigns and reruns automatically pick up v2.