SaaS · affiliate & creator marketing
Affiliate recruitment engine: evidence-based discovery and explainable scoring of creators
An engine that finds creators and publishers who already recommend comparable software, proves the fit with technical evidence and an explained score, then runs a personalized three-step sequence.
- Client
- Client confidential
- Industry
- SaaS · affiliate & creator marketing
- Year
- 2026
Challenge
To grow the same affiliate program, the client wanted creators and publishers who already earn by recommending comparable software: reviewers, tutorial channels and comparison sites. A list built from audience size alone says nothing about whether someone runs affiliate partnerships. The team needed proof of fit, a clear reason behind every score and a contact address with a known origin before anyone received an email.
Solution
We built an engine that answers four questions for each candidate. Who they are: identities, contacts and language, each with its source. Why they fit: evidence of existing partnerships and relevant content. How good they are: evidence, fit, reach and contactability combined into an explained opportunity score. How to contact them: which address, where it came from, whether it was verified and which message goes out. Qualified candidates receive a personalized three-step sequence; replies stop it, are classified and become manager tasks, and registrations in the affiliate program after contact are tracked as conversions.
How it connects
- 01 Discover candidates
- 02 Resolve links, detect affiliate fingerprints
- 03 Merge identities
- 04 Score with explanations
- 05 Enrich & verify contact
- 06 Three-step personal sequence
- 07 Replies & conversion tracking
What we delivered
- Pluggable discovery adapters for a video-platform API, web search, backlink indexes, partner directories and SaaS footprints
- SSRF-safe crawler and a caching redirect resolver
- Deterministic affiliate-link fingerprint detector
- Entity resolution across email, domain, channel and social profile, with an audit trail of every merge
- Explained opportunity score with contact provenance
- Personalized three-step sequence, reply classification and conversion tracking
- Dashboard and REST API with four roles: viewer, reviewer, affiliate manager, admin
- About 90 automated tests run on SQLite and PostgreSQL, plus a local demo that runs the real pipeline against a simulated internet
Functionality
- Discovery adapters that are restartable, idempotent and quota-aware
- Crawler that allows only public IPs, checked before the request, after DNS resolution and on every redirect, and pins the connection to the verified IP
- Size, MIME-type and time limits on every fetch; robots.txt is honored
- Redirect resolution for short links, cloaked /go/ links and meta-refresh pages
- New fingerprints re-index stored links without re-crawling
- Pre-score gate before any paid enrichment, plus per-provider usage and budget tracking
- LLM provenance: model, prompt version, hash, tokens and cost stored with every classification and generated opener
- Generated text validated for no links and no invented numbers, with a template fallback; the LLM is never the source of truth for scores
- Job queue with retries, back-off and a dead-letter queue, plus an audit log
- Read-only mailbox polling, because the inbox is shared with people, and address-level suppression across the client's outreach systems
Integrations
- Video-platform data API
- Web-search and backlink data APIs
- Contact-enrichment and email-verification APIs
- Google Gemini API
- SMTP sending and read-only IMAP polling
- Affiliate-tracking platform data for conversion tracking
Technologies
Result
Built and tested. The engine runs in dry-run mode, and no message goes out until the client gives an explicit go-ahead. Discovery adapters that depend on paid data providers switch on once their API keys are in place.
The engine shares its suppression list with the startup discovery & outreach platform and the client’s other outreach instances, so an opt-out recorded anywhere is honored here too.
Discuss your project
Tell us about your project