The Causal Ledger
The subject now is the connection among the advertising journey, the measurement regime, and the budget decision. The generator platform, trusted marketplace, and brand are not being compared as isolated businesses; they are being linked through the path a purchase takes from created intent to booked transaction to reported evidence. The practical judgment is an audit: before accepting a ROAS or lift claim, identify where demand was created, where the outcome was booked, what was observable, what was randomized, and which threshold the evidence is supposed to clear.
Follow the purchase, not just the receipt
Take one customer who sees a product on the generator, becomes more likely to buy, and later checks out on the trusted marketplace. The commercial event is a purchase. The causal event is the change in the customer's likelihood of purchasing because of the advertising opportunity. The reporting event is the part of that journey that the brand can connect to a touchpoint. The allocation event is what the brand does after seeing the report. A useful ledger keeps all four events in view rather than allowing the last one recorded to stand in for the first.
The first ledger field is demand created. The generator's spend may produce incremental purchase intent with diminishing returns. This is the contribution the brand wants to value when it asks whether more investment in G will create more purchases than the no-ad alternative. It is not yet an order on any particular storefront.
The second field is checkout location. Trust can send the customer to G's storefront or to R. That choice changes the commercial record's location without necessarily changing the amount of intent created. A marketplace checkout can be essential to completing the transaction while remaining downstream of the original demand-generation effect.
The third field is channel-complete outcome. If the brand's first-party records include purchases on G, R, and other relevant destinations, they can represent the total outcome of the assigned advertising opportunity. If they include only G's orders, they represent a narrower surface. A narrower outcome is not automatically useless, but it cannot support a claim about the whole journey without an additional assumption.
The fourth field is observable match. Attribution sees conversions connected to a touchpoint through the signals and rules available to it. The observability factor κ summarizes how much of G's caused demand survives when some customers transact on R and only part of that off-platform activity is recovered. This field answers “what can be linked?” It does not answer “what would have happened without the ad?”
The fifth field is counterfactual comparison. Randomized incrementality adds a treatment assignment and a control condition. It compares channel-complete outcomes among comparable assigned groups rather than relying on the last visible interaction. This is the field a causal budget claim cannot do without. A transaction record tells the brand that a purchase occurred; a randomized contrast helps establish whether the advertising opportunity changed the number of purchases.
The sixth field is decision threshold. A brand may have a return hurdle, a minimum incremental conversion value, or a requirement for a segment-level claim.