Back to home
All four

Creator research · precise screening · personalised outreach

Creator research and outreach system

Finds creators worth approaching, writes each of them a tailored message, and sends approved outreach.

Outreach sent

Operational review

Signal → cause → change

NO EVIDENCE ≠ NO FIT
  1. NO EVIDENCE ≠ NO FIT
    Signal

    Some platforms block verification, while small creators may leave little searchable evidence.

    Root cause

    The tool's evidence ceiling says nothing negative about the creator.

    Change

    I added a pending-verification state instead of turning missing data into rejection.

What I built

I built a system that can research thousands of creators, find the small group that genuinely fits the product, and write outreach based on each person's real work. A person reviews the shortlist and message before anything is sent.

Complete workflow

the run of 2026-07-29

Define fit before looking at reachcriteria

Target, and what is excluded

Independent creators in crypto
Not agencies, not people who sell courses or shill coins

The run begins with a clear target and explicit exclusions.

The funnel comes from one dated Sippa run; screening accuracy and outreach outcomes are confirmed results from the live outreach.

System design

Critical controls

  1. 01

    Product fit over reach

    Product fitPrimary
    Follower countReference

    Follower count is not a retention signal. I designed the system around a harder question: would this creator's own audience pay them for what they make?

  2. 02

    Rules belong in scripts

    ScriptMetrics and bandsAIContent contextAIProduct fit

    A model once left 27 of 49 score fields blank and miscounted reposts as original work. I moved objective metrics and banding into code, where the same evidence always yields the same result.

  3. 03

    Uncertainty is not rejection

    VerifiedShortlistMissingPendingEvidenceDo not shortlist
    Person reviews list and message

    When a platform blocks verification, the candidate moves to a separate review state. Tool limitations are recorded as tool limitations and never converted into negative evidence about a person.

In operation

Operating results

  • 2,116→26from source audience to evidence-backed shortlist
  • 100%screening accuracy in the manually reviewed sample
  • 80%reply and partnership conversion on live outreach

The funnel comes from one dated Sippa run; screening accuracy and outreach outcomes are confirmed results from the live outreach.