FidelityGrid - Sell Project

FidelityGrid — I'm Building an Independent Reconciliation Engine for the Gap Between "Sent" and "Recorded" The Insight Every system that moves money has the same blind spot: the sender says they sent it, the receiver says they got it, but nobody independently verifies the record matches the event. I call this the Acknowledgment-to-Record Delta (A2R). It's the gap between a system confirming receipt and that event actually showing up correctly in the database. It sounds like it should be zero. It's not. I'm building the independent measurement layer for this — starting with two markets where billions flow through unverified pipelines. Market 1: Affiliate Attribution ($20–28B economy) Affiliate tracking vendors (Everflow, Impact, Partnerize) self-report their own accuracy. Advertisers trust the dashboard. Nobody checks. Here's the problem: a conversion event fires → the vendor ACKs receipt → but somewhere between acknowledgment and the final commission record, data diverges. Commissions vanish. The advertiser assumes it's "tracking waste." The publisher's payouts don't match clicks. The vendor says systems are working as intended. I built a tool that runs synthetic conversion events through real vendor infrastructure and measures what comes back vs. what was sent. But the key innovation isn't catching discrepancies — every monitoring tool does that. The breakthrough is classifying why the data diverged. We sort every gap into four buckets: Policy Loss — Signal destroyed by platform privacy changes (ITP, Privacy Sandbox). Vendor's handling it correctly. Unfixable at the vendor level. Technical Loss — Postback failures, dropped webhooks, sync errors. Vendor's responsibility. Fixable. Suspicious Loss — Unexplained divergence that survives a 24-hour audit window. Can't be attributed to policy or technical causes. This is where the money questions live. Legitimate Loss — Async processing delays, batch reporting, sampling. Expected behavior within vendor policy. This distinction is the entire value proposition. Existing tools give you a raw discrepancy score. That's useless — it conflates Apple privacy updates with genuine vendor failures with normal batch processing. Our taxonomy tells advertisers exactly where to focus and what's actually recoverable. Why now: The industry-wide transition from client-side to server-side tracking (driven by ITP, cookie deprecation, Privacy Sandbox) has created an attribution fog that benefits opaque incumbents. The vendors processing your data have zero incentive to make this transparent. DoubleVerify ($4B valuation) and IAS ($1.9B acquisition) proved the market for independent ad verification — but they cover programmatic. Affiliate has no equivalent. The lane is empty. Market 2: Webhook Processing Fidelity (The Bigger Opportunity) While building the affiliate product, I discovered the same A2R gap exists in a much larger market — and nobody's even looking at it. Webhook delivery networks (Hookdeck, Svix, Convoy) tell you the webhook was delivered. HTTP 200, done. But delivered ≠ processed. Did the receiving application actually parse the payload, execute the business logic, and write the correct record to the database? Nobody checks. Every webhook dashboard in production shows 99.9% delivery success. That number measures HTTP acknowledgment, not processing integrity. The gap between those two things is where payments fail silently, orders get lost, and subscriptions don't activate. This is the untouched observability blind spot. Datadog monitors infrastructure. Sentry catches errors. Hookdeck ensures delivery. But the question "did this Stripe webhook actually result in the correct subscription record in my database?" — nobody answers that independently. Our approach: a deployed agent model running in the customer's environment with read-only database access. Not an external audit — an internal verification layer. It watches webhook events arrive and confirms the expected database state materialized within the processing window. When it doesn't, it flags the divergence with the same loss taxonomy framework. Starting with Stripe webhooks (most developer-familiar endpoint, highest payment-critical stakes) and expanding from there. Buyer persona: CTOs and SREs responsible for mission-critical webhook infrastructure who've experienced silent delivery failures their existing tools didn't catch. What's Built The affiliate product is live with a working A2R pipeline: Postback ingestion endpoint accepting vendor webhook events Two-phase verification worker (ACK vs. record match on transactionid) 24-hour audit window before Suspicious Loss classification Internal control self-test running synthetic events every 5 minutes for measurement calibration Component-level health monitoring dashboard Bloomberg-terminal aesthetic UI with 30-day trend visualization, loss taxonomy breakdown, vendor reliability scores Full auth system with admin panel Live at fidelitygrid.polsia.app The webhook produ...