I spent years on the performance side and treated brand measurement as a way of justifying spend that could not be justified otherwise. Some of that scepticism was earned.
Having since had to run brand activity and defend it, I would separate the parts that are genuinely measurable from the parts that are not.
The fair criticism
Stating it properly first.
A lot of brand reporting consists of metrics with no established link to commercial outcomes, measured inconsistently, presented without a control, and interpreted generously.
Awareness rising during a period when spend occurred is not evidence that spend caused it, and it is routinely presented as such.
Tracking studies with small samples produce movements within the margin of error that get reported as trends.
And the absence of a falsifiable prediction means brand activity can rarely be shown to have failed, which is a legitimate thing to be suspicious of.
What can actually be measured
Several things, with proper method.
Brand lift studies with a control group. Exposed and unexposed populations, surveyed on the same questions, with the difference attributable to exposure. This is a genuine experiment and the major platforms support it.
The caveats are real — self-selection into the exposed group, survey response bias, small effect sizes requiring large samples — and it is a controlled comparison rather than a correlation.
Branded search volume, which is my preferred indicator because it is behavioural rather than stated. Somebody searching for your name has been prompted by something, and the series is available daily, free, at no survey cost.
It is contaminated by other activity, and it responds visibly to substantial brand campaigns and is a reasonable leading indicator.
Direct traffic, similarly, with similar caveats.
And geographic holdout tests, which are the strongest method available. Run the activity in some regions and not others, matched on relevant characteristics, and compare commercial outcomes rather than survey responses.
The geo test deserves emphasis
Because it answers the question performance marketers actually want answered.
It measures effect on sales, not on stated recall. It has a genuine control. It is not dependent on tracking individuals, so it is unaffected by signal loss.
The requirements are that you have enough regions to construct matched groups, enough time for effects to emerge, and the organisational willingness to deliberately not advertise somewhere.
That last one is the obstacle, and it is a political obstacle rather than a methodological one.
The cost is the foregone activity in the control regions, which is real and is considerably less than the cost of running ineffective activity everywhere for years.
What genuinely cannot be measured well
Being honest about the limits.
Long-term effects beyond a year or two are extremely difficult to attribute to specific activity, because too much else changes.
The value of brand strength as an asset — pricing power, resilience, negotiating position with retailers — is real, well evidenced in aggregate across many businesses, and hard to quantify for one business in one period.
And the counterfactual over a long horizon is unknowable. What would have happened to a business that stopped brand activity for five years cannot be established except by doing it.
The evidence that persuaded me
Not any single study.
It was the accumulated body of work analysing large numbers of campaigns across many categories, consistently finding that activity with broad reach and emotional content produces larger long-term business effects than narrowly targeted rational activity, and that the ratio between the two matters.
That work has its critics and its methodological limitations. It is a substantial body of evidence pointing in a consistent direction, which is more than most propositions in marketing have.
The practical position
What I would actually recommend.
Use branded search and direct traffic as continuous, cheap indicators.
Run brand lift studies where the platform supports them, and read them cautiously.
Run a geographic holdout at least once, properly, on the largest brand investment you make. It is the only method that produces an answer a sceptic will accept.
And be willing to state in advance what you expect to happen, so that afterwards there is a basis for saying whether it did. The refusal to make falsifiable predictions is what earns the scepticism, and it is entirely within the discipline's control to stop.
Setting the expectation in advance
The practice that has done most to make these conversations productive.
Before any significant brand activity, we write down what we expect to see, on which measures, by when, and what magnitude would count as success.
That document is circulated and agreed, and it is read afterwards.
It is uncomfortable, because it creates the possibility of being demonstrably wrong, which is precisely why it earns credibility with the people who are otherwise sceptical.
The share of voice relationship
One further measure worth knowing about.
There is a well-documented relationship between a brand's share of advertising spend in its category and its subsequent share of market, with the gap between the two predicting growth or decline.
It is estimated across many categories rather than proven for any single business, and it has the advantage of being calculable from data you can obtain.
As a planning input it is useful, because it frames budget as a competitive question rather than an efficiency one, which is a more honest description of what brand spend is doing.