Show HN: Built an attribution tool that uses Bayesian inference to track ROI
I’ve spent the last few quarters building utm.one – a privacy-focused link management and revenue attribution platform.
The Problem: I was frustrated that Google Analytics attributes most B2B conversions to "Direct" or "Organic" because it loses track of the user journey across devices (e.g., mobile click -> desktop purchase) or offline events.
The Solution: instead of just counting clicks, I built an identity graph that stitches sessions using a first-party pixel and uses Bayesian inference to calculate the probability of a conversion coming from a specific link.
Key Features:
Probabilistic Identity Graph: Links anonymous mobile clicks to desktop purchases based on IP/Time clusters (no third-party cookies).
"One-Tap" Badge Scanner: A PWA that uses OCR (via LLM) to scan event badges that standard QR readers can't parse, then syncs them to the attribution model with event halo which let you know how much lift has taken place at particular location. This solve for event attribution.
Revenue Lift: Calculates incremental lift using a control-group logic rather than just last-click.
The stack is React, Supabase, and Edge Functions for the redirection engine (latency was a huge focus).
I’d love you to roast my logic, the UI (went for a Jony Ive-inspired monochrome look), or the attribution math.
Live demo/Link: https://utm.one/auth
PS. Working with Stripe to get access; they only support if you are doing upwards of $100K. Any suggestions on payment gateways