This is a case-study skeleton — a template we use with new enterprise customers who want to replicate the playbook. Swap in your own ICP, vertical, and platforms, and the same structure produces predictable results. We have published the anonymized version here because the template itself is more useful than any single customer story.
The starting point
A Series B SaaS company in Southeast Asia, selling workflow automation to mid-market logistics firms. Average contract value around forty thousand USD. Sales cycle ninety days. Before Anvil: one outbound SDR sending four hundred cold emails a week, producing roughly eight discovery calls per month. The CEO wanted to triple that without tripling headcount.
Day 0-7: ICP and signal definition
The first week is not about collecting leads. It is about defining what a lead looks like. With the customer, we built a written ICP covering: industry sub-segment (cold-chain logistics), company size (200-2000 employees), geography (Indonesia, Vietnam, Thailand), tech-stack fingerprints (using spreadsheets or legacy ERP, not already on a modern automation stack), and trigger events (hiring a "logistics coordinator" or publicly complaining about manual processes).
This became the rubric for everything downstream. Leads that missed any two criteria would be scored below the threshold and never reach the SDR inbox.
Day 7-30: Collection and calibration
We connected three platforms in week two: LinkedIn for hiring signals, a Vietnamese logistics forum for workflow-pain complaints, and Xiaohongshu for broader supply-chain sentiment. The crawler found ninety-two thousand public posts in the first fortnight. The AI scoring engine filtered that down to eight hundred candidates matching the ICP, and then the intent model ranked them from highest to lowest probability of conversion.
The customer spent week three manually reviewing the top one hundred leads and marking them as "hot", "warm", or "cold". Those labels fed back into the scoring model, which retrained overnight. By end of week four the false-positive rate had dropped from thirty-one percent to eight percent.
Day 30-60: Outreach choreography
The SDR no longer sends cold emails. Instead, every morning he gets a ranked queue of twelve prospects, each annotated with the specific signal that flagged them (e.g. "posted a complaint about missing shipments three days ago" or "just hired a logistics coordinator, likely replacing manual processes"). The opening message references that exact context.
Reply rates climbed from 1.2 percent in the old regime to 9.4 percent by day fifty. Meeting-booked rates roughly quadrupled. Crucially, the SDR actually worked fewer hours per day because the research burden moved from humans to the platform.
Day 60-90: Pipeline and scale
By day ninety the customer had 1,047 qualified leads in the CRM (defined as "matches ICP and has a verified contact channel"), eighty-one discovery calls booked, twenty-two opportunities in pipeline, and two closed-won deals worth combined ninety-thousand USD. The pipeline coverage improved so much that the team paused outbound for two weeks to catch up on inbound conversations.
What made it work (and what did not)
Three things mattered more than the rest. First, a disciplined ICP written down before any scraping started. Second, weekly calibration sessions where the sales team labeled leads and the model retrained. Third, respecting the guardrails: the team treated Anvil as a signal-provider, not a spam cannon. When they tried sending bulk generic messages to the middle of the queue, reply rates collapsed and they quickly reverted.
If you want the full playbook adapted to your business, our customer success team runs a free ninety-day kickoff program for new Pro and Enterprise customers. It is the fastest way to replicate these numbers.