Case studies Updated 18 September 2026

Running an acquisition agent
in public.

Three studies from one operation. Each publishes the denominator rather than the conversion rate, names the things that went wrong, and shows the receipt behind every figure. No prospect is named in any of them.

Every number traced to a ledger row Failures included, on purpose

The series

One operation, seen from three sides

One sales story, one sourcing story, one economics story. Read in any order.

Part 1 of 3The Pepper Ledger1,146 cold emails, six replies, one meeting held — published with the full denominator. Part 2 of 3Disallow: SlashTrustpilot refuses AI agents by name. We obeyed it, and it cost us the whole market. Part 3 of 3The Failure Bill$648.69 in six months, and 44% of every model call came back an error.

The standard

Why publish the failures

A case study that hides its denominator is a claim, not evidence. These exist so a prospective client can check our arithmetic before trusting us with theirs — the same standard the outbound itself runs on: the composer refuses to render any figure that has no source behind it.

Rick Acquisition — $800/month