Skip to content

Series · 7 parts Published August 11, 2026

Lead source attributor

A small system that captures where each enquiry came from, stitches together the touches that led to it, assigns credit under a stated rule, and — the part most attribution tools skip — reports how much of the total it could not attribute at all. Seven posts on the same system, one diagram at a time, with a cost breakdown and an engineering reference at the end.

  1. 01

    A lead source attributor on AWS for a few dollars a month

    The whole system on one page — capture, stitch, credit and report — and the single decision that determines whether the output is trustworthy.

  2. 02

    How a touch gets captured

    What is recorded on each arrival, why the referrer is less reliable than it looks, what direct traffic actually contains, and the offline touches nothing can see.

  3. 03

    How visits get stitched

    How repeat visits get linked, why the enquiry itself is the strongest signal, how confidence is recorded, and the identity work this system deliberately refuses.

  4. 04

    How credit gets assigned

    The four common credit rules and what each is actually good for, why reporting two beats picking one, and the case against the rule that looks most sophisticated.

  5. 05

    How the report shows its uncertainty

    Why the unattributed share belongs in the headline, how to present two rules without confusing the reader, and the standing caveat every report should carry.

  6. 06

    What the lead source attributor costs

    About $122 a month. One Bedrock read per touch is the only line that grows; the queue, the table, the mail and the storage are rounding errors. Plus the three ways the bill could surprise you.

  7. 07

    Engineering reference: the lead source attributor architecture

    Same system, drawn purely for engineers. Service names, region, Lambda inventory, IAM scopes, the schemas and the exact model id.

What is a lead source attributor?
A small serverless system that records where enquiries came from, links repeat visits from the same person, applies a stated credit rule, and reports the result with its unknown share visible.
Why not just use the last click?
You can, and for many businesses it is the right choice. The problem is not that last-click is wrong; it is that nobody says they are using it, so the report reads as truth rather than as one rule among several.
How much of a typical report is unattributable?
Commonly a quarter to a half, and the post on this is blunt about it. Direct traffic, word of mouth, blocked tracking and phone calls are all real sources that leave no trace.
Does this replace an analytics product?
No. It answers one question — where did this enquiry come from — and stores the answer next to the enquiry, which analytics products are surprisingly bad at.
What does it cost to run?
A couple of dollars a month. See part six.
All posts