I help Amazon agencies and ad-tech teams use Claude with AMC, so the numbers they hand clients are right, not just confident.
After a decade in corporate data, I picked one layer and stayed there.
Not "Amazon marketing." Not "ads." Just the layer that tells you whether your spend actually grew a client, or only moved a number around.
Agencies and ad-tech teams are wiring Claude straight into their Amazon data now. It works — and it will hand them a wrong number with total confidence. I build the layer that catches it before the client does, so the number they present is one they can defend as their own.
Four that turn up again and again. None of them throws an error, which is exactly why they reach clients.
Spend that reads 100,000,000× too high
AMC reports some fields in microcents. Sponsored divides by
1e8, DSP by 1e5 — same report, two constants.
"No data" that is really suppressed data
Rows AMC withheld for privacy, not zero. The two are indistinguishable in the output.
A month that drifts while you read it
Re-queried across the restatement window, the same month can move from ~9,000 customers down to ~1,200. The day you run it changes the answer.
"Incremental" where no holdout was ever run
The word doing work the data never did.
From a free series through to sitting inside a platform team. Details and pricing on request.
AMC 0 to 1 with Claude — free
Video series and checklists. Start here.
AMC Setup Check
Ninety minutes on your own AMC account.
Verified Measurement Read
One AMC question answered properly, on your data, with the checks written down.
AMC with Claude — Founding Group
Four weeks, live, for agency teams building on their own accounts.
Embedded AMC Advisor
For ad-tech platforms building AMC features. I check the number before your clients see it.
I'm a data scientist who came up through energy, pricing jet fuel for airlines, where a wrong number moved money on a live contract. Senior data science roles at BP and Shell. I was building machine learning models before ChatGPT existed.
Now I only do AMC.