The caused-vs-anticipated ledger
Holdouts and matched-market experiments on your actual spend — which conversions your money caused, and which were coming anyway, with confidence intervals printed on the slide.
§01 · The asymmetry
Other tools optimize the number your ad platform reports.
Mizoki optimizes the number your bank account reports.
Enterprise platforms have quant departments. You have dashboards that grade their own homework — every ad platform reports the conversions that happened after its own ads, and none of them tell you which orders were coming anyway, or what an order actually nets after components, packing, shipping, fees, and returns. Mizoki closes that gap with experiments and a ledger, not a prettier report.
§02 · The bundle that lost money — composite scenario · illustrative numbers
Revenue up. Cash down. The hero product hid its own costs.
Their $180 bundle was the best seller, and cash kept shrinking. Signal priced every order at what it truly nets — components, packing, dimensional shipping, payment fees, and the returns coming back.
The ledger — what each order truly netted
composite · illustrative
Hero bundle · a third of them boomeranged
−$2 to $16 per order
The "boring" single SKU · almost never returned
$31 per order
When net-yield bidding ships, the bidding signal switches from checkout revenue to net contribution, and the ad platforms start hunting buyers who keep what they buy. Revenue flattens. Cash grows.
Stated plainly: net-yield pricing and bidding are Preview · in development — nothing here claims they are deployed, and the label flips only after a real pilot writes verified numbers. Returns of 25–40% in apparel are an industry statistic, not a customer result (composite scenario).
§03 · The doorman — composite scenario · illustrative numbers
How much of your retargeting was already walking through the door?
Your dashboard says retargeting returns 8x (illustrative). The question it can't answer: how many of those people were coming back anyway? One team let us quietly hold the ads back from a small, identical slice of shoppers. That slice bought almost as often as everyone else — the real return was about 1.4x (illustrative). They cut the channel by a third, moved the money to net-new customers, and revenue held. You were paying a doorman to open a door people were already walking through.
§04 · What you get today
Proof, not attribution theatre.
One canonical event stream
Shopify orders, Google Ads, Meta, and lifecycle email emit the same governed business event — evidence to reason over, not eight vendor exports to reconcile.
A governed loop that earns trust
Starts in observe-only and earns spend authority the way you'd let a new hire earn it. Never a guaranteed outcome — always a decision with a receipt and a rollback token.
§05 · On the roadmap
Prove → Profit → Anticipate Preview · in development
Prove is live in the ledger above. The next two acts are built in the open and labeled honestly until a pilot verifies them.
Net contribution, not top line
Your real economics — landed component costs, pick-and-pack, shipping, fees, and the returns that quietly turn winning campaigns into losing ones — priced into every order from your own cost inputs, never invented defaults. After your pilot proves it, those net-contribution values feed the ad platforms' own bidders, recommend-only first.
Anticipatory intent
Reading where a shopper is heading before they raise their hand — anticipatory modeling of consented behavior, not mind-reading. By house rule, acting on a prediction always runs against a held-out comparison: the prediction never grades itself.
Stated plainly: both cards above are Preview · in development, not deployed. Signal uses no microphone or audio signals of any kind — not now, not on the roadmap.
§06 · See it run
Watch a real decision earn its receipt.
The live demo runs the actual pipeline — including the veto most demos hide. No pressure, no forms first: run it, replay it, then talk to us if it's your Tuesday.