What is triangulation in marketing measurement?

Triangulation is the idea that no single method measures marketing completely, so three should be used together: a marketing mix model for the whole picture, attribution for the day-to-day inside each platform, and experiments for causal readings of single channels. Google's measurement framework describes the same triangle. The idea is sound. What separates vendors is how the three are combined.

The first way runs three parallel analyses and reconciles their outputs afterwards, in a dashboard, with multipliers to make attribution agree with the model. It sounds complete on paper and is often hard to use in practice, because three answers to the same question have to be argued into one.

The second way folds the other two methods into the model before it is trained. Platform data, together with what the team knows, informs the priors. Experiment results, a lift and its uncertainty, go in as priors on that channel's return. The model still estimates the final answer across every channel, but it starts from the causal evidence where it exists. Calibrating a model this way is mainstream; Google Meridian and Meta's guidance both do it. What differs is the discipline of setting the priors.

For how much media budget is needed and how to split it between channels, the marketing mix model is the source. Those two questions do not need a separate triangulation step, because the model already takes what it can from the other two when it is built. For optimizing inside a platform, the platform's own tracking is the best tool there is; it is just not a currency that works across platforms. And where the model's range is too wide to act on, an experiment is the way to narrow it, and its result goes back into the model.

One caution on granularity: a platform that promises ad-group or ad-level recommendations from a triangulated model is promising more than aggregated data can carry. Odins is clear and actionable at the level the data supports.

• Calibration: Anchoring a model to outside evidence, usually experiment results. In a Bayesian model, calibration is a prior.
• Attribution: Crediting conversions to the touchpoints that preceded them. Great inside a platform, not a currency across them.
• Incrementality test: A controlled experiment that measures what one channel causes, by comparing exposed and unexposed groups or periods.
• Prior: What the model believes about a channel before it sees the data. The part of MMM that decides how much the data can teach.
• Marketing mix modeling (MMM): The statistical method behind everything else on this list: spend in, sales out, every channel on equal terms.

Go one level deeper: Triangulation: MMM, attribution and experiments. Recommendations and testing. Or go back to all terms in the glossary.