Can a CFO rely on marketing attribution for a budget decision?

Credit for what was recorded

Google Analytics defines attribution as "the act of assigning credit for important user actions to different ads, clicks, and factors along the user's path to completing the action." The model that assigns the credit, in Google's description, can be a rule, a set of rules, or a data-driven algorithm.¹

Attribution begins with outcomes the measurement system has recorded and assigns credit for them across eligible interactions. How that credit is estimated can differ substantially between models. How well a particular approach does so is a technical question, and it can have a good answer.

Not one kind of estimate

It is tempting to file every attribution model under "credit rules" and leave causation to some other discipline. The platforms' own descriptions don't support that.

Google describes its data-driven model as evaluating both converting and non-converting paths and, "using a counterfactual approach", contrasting what happened with what could have occurred, to determine which touchpoints are most likely to drive key events.¹ Meta describes an incremental attribution model that uses machine learning to predict whether a conversion was caused by an ad.² These are attempts to estimate what advertising contributed, not just where it appeared along the way.

The familiar objection that attribution "isn't causal" is therefore too blunt to be useful. Some attribution is built on causal reasoning. The limit that matters for a budget lies somewhere else.

Last year's return and the next pound

Google's open-source marketing mix modelling framework, Meridian, draws a distinction that is useful well beyond that tool. Its return on investment for a channel is built on the incremental outcome the model estimates for that channel's spend, and the documentation describes it as "a historical, channel-wide average". Marginal return on investment is a different figure: "the return on the next dollar spent". Where marginal return is much lower than the average, the documentation notes, the channel is beginning to saturate at its historical level of spend.³

The distinction matters beyond the terminology used in Meridian. An estimate derived from spend already observed—even where it estimates causal effect—does not automatically tell the business what an additional unit of spend will return. A proposal to add budget concerns money not yet spent, possibly at a level the business has never operated at. The two figures can be close. They can also diverge, and the historical figure does not, by itself, show which.

Experiments share the boundary. Google describes incrementality experiments, Conversion Lift among them, as its way of measuring the causal impact of ads, and as showing how effective ads are at driving an action "at a certain point in time".⁴ An experiment answers its question for the period and conditions in which it ran.

Where the budget question sits

None of this makes an attribution figure wrong, or a budget request weak. It places them. A report on past performance answers a question about the past, and can answer it well. A request for more money rests on an expectation about what happens next, at a different scale.

When the one is offered in support of the other, the decision sits in the space between them, and the quality of the figure on its own terms does not close that space. Where that distinction is material to a consequential budget commitment, the attribution figure on its own does not resolve it.

A second view before the money moves

When a larger budget commitment is approaching and an outside view would help, the MICHVI Evidence Assessment provides one: an independent assessment of the evidence relied upon for a defined consequential decision. It is standalone and starts from £5,000 in the UK or €5,000 in the Eurozone, with the final fee confirmed at scoping. MICHVI does not manage media budgets, buy advertising or optimise campaigns.

To discuss a specific decision, scope an Evidence Assessment.

Sources

1. Google Analytics Help, Get started with attribution
2. Meta Business Help Centre, About incremental attribution
3. Google for Developers, Meridian, Incremental Outcome, ROI, mROI & Response Curves
4. Google Ads Help, About Conversion Lift