Long-term attribution: measuring the marketing that doesn't convert today

Long-term attribution works in two stages: first establish which brand metric actually drives your commercial outcome and with what lag, then model what moves that brand metric. The result is a chain from media spend through brand to revenue, which lets you value activity whose effect arrives quarters later.

It is the layer most measurement programmes skip, and the reason brand spend usually gets defended with theory rather than numbers.

Stage one: find the metric that matters

Most organisations track several brand measures: unaided awareness, prompted awareness, consideration, preference, net promoter score. They are usually treated as a set to be reported together, with an unstated assumption that all of them are good and more is better.

They are not equally predictive. In any given business, typically one has a consistent, lagged, statistically significant relationship with the commercial outcome, and the rest move without moving anything else. Establishing which one is the first job, and the answer varies by category more than people expect.

The output of this stage answers two questions precisely. If we shift that metric by a given amount, what happens to our sales base? And how long until it shows up? That lag is the single most useful number in the exercise, because it tells you how far ahead your planning horizon needs to sit.

Stage two: model what moves it

Once you know the metric, you can treat it as the dependent variable and run something structurally similar to marketing mix modelling against it. Which channels shift it, at what spend levels, with what diminishing returns, nationally or state by state.

Now brand spend has a modelled return. Not a philosophical argument about salience, and not a benchmark borrowed from another industry. A coefficient estimated on your own data.

Lead metrics, and why they matter more than they sound

A two-year lag between spend and confirmed effect is commercially intolerable. No board approves a plan whose first evidence arrives after the next two budget cycles.

Lead metrics solve this. They sit between the long-term brand metric and the commercial outcome, and they respond faster. Identifying them gives you in-flight tracking: a long campaign can be diagnosed at three months rather than validated at eighteen.

What it takes

You need a brand tracking series with enough history, plus the same aggregate media and outcome data an MMM requires. Three years of brand tracking is comfortable; less is workable with wider intervals. If you have no brand tracking at all, Share of Search can stand in as a proxy for brand demand while you establish one.

The harder requirement is organisational. Long-term attribution produces recommendations that cost money now for returns later, which means it only survives if the finance function is in the room while the model is being built rather than being presented with its conclusions afterward. In our experience that sequencing decides whether the work changes anything.

Where it fits

Long-term attribution on its own tells you about brand and leaves the short term unmanaged. Marketing mix modelling on its own is accurate about the quarter and silent about the compounding. Direct attribution optimises the last step well and explains nothing else.

The value is in the combination, which is what our full funnel framework is for. Any single layer will be more precise about its own slice than the framework is. None of them will answer the question the CFO is actually asking.