Multi-Touch Attribution Lite
Stitch every contact's touch history into a clear journey, then credit your channels with simple first-touch, last-touch, and linear models you can actually explain - no black-box platform required.
A private internal tool where you import touches and conversions, see each converter's real journey, compute first/last/linear channel credit, approve the model and output, and export a clean channel-credit report - all with login, audit trail, and a human approval gate.
Before you start
- A touchpoints export (CSV): contact, channel/source, timestamp, type
- A conversions export (CSV): contact, conversion date, value
- Free accounts: Vercel, Supabase, Resend
- Claude Code installed on a Linux machine
The problem this kills
You know your leads touch a dozen things before they buy - an ad, a webinar, an email, a demo. But when someone asks "which channels actually drove revenue?", you're stuck. Enterprise attribution platforms cost a fortune, take months to set up, and spit out a "credit score" nobody can explain. So you fall back to last-click in your ad dashboard and quietly hope the budget meeting goes okay.
The truth is, you don't need a black box. You need to see each converter's real path, apply a model you understand, and stand behind the numbers. This plan builds exactly that.
What you'll build
A small, private web app - just for your team - that:
- Imports your touchpoints CSV (who touched what channel, and when) and your conversions CSV (who converted, the date, and the value).
- Stitches every contact's touches into a time-ordered journey, then attaches their conversion.
- Computes channel credit three explainable ways: first-touch, last-touch, and linear (evenly split across every touch in the path).
- Shows you the actual journey for each converter so you can sanity-check the math and trust it.
- Makes an analyst review and approve the model and the output before it's used for any budget or reporting decision.
- Exports a clean channel-credit report CSV and journey CSV in the columns your stakeholders expect.
What's inside the Implementation Plan
The plan opens by interviewing you about your business - your channels, your CSV column names, your conversion definition, your typical and peak volumes, and your messy edge cases (contacts with no touches, duplicate touches, touches that happen after the conversion). It reflects a short tailored spec back to you and waits for your thumbs-up before building a single thing. You get a tool shaped around your data, not a generic template.
From there it's a step-by-step build: data model, secure CSV import with duplicate guards, journey stitching, the three attribution models, the converter journey view, the human approval gate, the audit trail, and the CSV exports. Every step ends with a ready-to-paste prompt.
The governance it includes (this is the point)
- Login so only your team can open the tool.
- Row-level security so each organization only ever sees its own data.
- A full audit trail - who imported what, who approved which model run, and when.
- A hard human-in-the-loop approval gate - the tool computes the attribution, but a person reviews the journeys and the model and approves before the output is final.
- Duplicate guards so the same touch or the same conversion can't be counted twice.
Who it's for
Marketing ops pros and analysts who need a credible answer to "which channels helped?" without buying or learning enterprise attribution software. If you can export two CSVs and read a journey, you can run this.
You've got this - paste the first prompt and let's build it.