Why last-click attribution quietly killed your brand budget

Last-click attribution credits a sale to the final touchpoint before conversion, which systematically over-values the channels that appear late in a purchase journey and under-values everything that created the demand in the first place. Optimise against it for long enough and you end up with a media plan that harvests demand efficiently and generates almost none.

This is not a controversial position any more. It is worth revisiting because the damage is slow, the symptoms arrive years after the cause, and the remedy has to be argued in front of a CFO who was told a decade ago that digital had solved measurement.

How the trap closes

Nobody sets out to defund brand. The sequence is almost always the same, and it looks rational at every individual step.

A team adopts digital measurement and can suddenly see conversions attributed to specific placements. Brand channels such as television, out of home, sponsorship and radio produce no such trace. Under pressure to demonstrate return, spend shifts toward the channels that can prove themselves. Performance improves on the measured metric. The improvement justifies moving more budget the same way.

Then the medium term arrives. Cost per acquisition begins to rise despite continual optimisation. The pool of people already looking for the category has been harvested thoroughly, and nothing upstream is refilling it. Bidding gets more expensive because everyone else in the category has run the same playbook. Volume flattens or falls even as efficiency metrics look defensible.

At that point the marketing team is in a difficult position. They have several years of reporting that says the current mix works, and a deteriorating business result that says it doesn't. The measurement system cannot see the cause because the cause is the thing it was never built to measure.

Why it isn't a technology problem

Every few years a platform arrives promising to fix attribution, and the underlying issue survives it. That is because the issue is a measurement design question, not a tooling one.

Any attribution model has to make an assumption about how credit is distributed across a journey. Last-click makes the most convenient assumption available: all of it goes to the end. Multi-touch models spread it more plausibly, but they still only see the touchpoints they can observe, which means they still cannot see the television advertisement that made someone search your brand name three weeks earlier.

Buying a better attribution platform does not resolve this. It changes which unobservable effects get quietly assigned to zero.

What actually resolves it

The measurement has to work at more than one time horizon, and it has to include the things a tracking pixel cannot see.

• Marketing mix modelling works on aggregate data, so it can include offline channels, competitor activity, price, seasonality and macroeconomic conditions. It does not require user-level tracking, which also makes it durable against signal loss.
• Long-term attribution establishes which brand metric actually drives your commercial metric, how long the lag is, and what shifts that brand metric. Without this layer, brand spend remains an article of faith.
• Direct attribution still matters for the short-term conversion decisions. The point is not to discard it. It is to stop using it to answer questions about the whole funnel.

Combining all three is what our Holistic Analytics Marketing Model does, and the reason we build it that way is unglamorous: it is the only arrangement we have found that lets a CMO answer the CFO's question in the CFO's own units.

The conversation to have with your finance team

Do not open by arguing that brand is undervalued. That is a position, and positions get debated. Open by asking a question the current reporting cannot answer: what happens to our cost per acquisition over the next eighteen months if we hold this mix?

If nobody can answer, you have established the gap without attacking anyone's numbers. That is usually enough to fund the work that answers it properly.