Making online campaign measurement more productive

Undoubtedly, advertisers and agencies would spend more on digital advertising if they could only be sure that their placements were working. Similarly, if publishers could only demonstrate the relative value of different placements and formats to their buying partners then a productive discussion would ensue. Inventory with proven performance would command higher prices and those prices would find their natural level as demand and supply equalised.

But what’s been preventing this mutual exchange of value? The chief culprit has been the data upon which our increasingly online-centric (as opposed to platform-neutral) and complicated (as opposed to sophisticated) digital trading environment is based.

We have seen tremendous innovation in real-time bidding, behaviour-specific precision marketing, automated optimisation – innovation that has provided online with some clear advantages and will continue to do so. However that rapid-fire innovation has come at a cost. The quality of online datasets is, to be generous, highly variable. Those within the industry have inferred behaviours, inferred the needs driving those behaviours, inferred demographics and in some cases actively processed people out of datasets.

The result is the commoditisation of both data and inventory – and advertisers are left still asking three fundamental questions:

1. Who actually saw my ad?
2. What is the relative value of my online activity compared to TV?
3. What is the link between my digital media activity and consumer purchases?

Publishers also have their own questions: How do I place a value on my inventory, and how do I differentiate my audience? To provide evaluation of media effectiveness and how that leads directly to a client’s profitability, then the bedrock of that work lies in the humble Gross Rating Point (GRP). Put simply, that is the total percentage on-target reach multiplied by the average frequency. The GRP is a TV-standard metric and provides a measure of the weight of a campaign and its effectiveness.

We have been plugging GRP data from multiple media into our marketing mix models for a long time, but online has remained a stubbornly elusive medium to measure in such a way. It lacks a robust (high sample coverage) and scalable (every campaign measured, regardless of size) Gross Rating Point that can be compared directly to a television rating. The availability of such a metric would allow advertisers to answer those three questions above. Up until now approaches to GRP measurement have been limited to yesterday’s logic: estimates based on traditional panel-only approaches or to cookie/browser approaches.

Panels are difficult to build and they suffer from low sample coverage, meaning that – on average – they represent only one percent or fewer of the people we wish to estimate (therefore estimates are based on 99% projection). The resulting GRP could hardly be described as robust. Further, due to the restrictive size of panel-only approaches, there is a significant limit to the granularity that can be reported rendering the approach unscalable.

If the GRP is not robust and does not represent all activity, how could we plug it into an econometric model and ask clients to base rational business decisions on the basis of such data? That is the question we’ve been asking of ourselves over the last five years.

At Nielsen we’ve done two things to support buyers and sellers of media:

1. Developed a marketing mix model that allows clients to conduct multivariate analysis with optimisation capabilities to help them understand how to build the right media mix for their business objectives.
2. Developed an online GRP worthy of input into those marketing mix models – essentially a tradable set of data that supports media evaluation for both buyers and sellers and has the ability to optimise campaigns of any size in flight by day, by publisher and by placement.

Years ago, Capital Radio ran a piece of research – the Ironing Board study – shortly after the introduction of Rajar. It assessed the relative effectiveness and efficiency of TV and Radio GRPs. Whatever the outcome, it was a piece of work which only became possible once a robust, transparent and scalable GRP was available for radio.

Online has indeed delivered at the bleeding edge of innovation with specialist digital metrics. We believe these metrics can only deliver true value to buyers and sellers once they can be evaluated in the context of people, not just machines, and within the wider media context – not just online. To do that, we cannot rely on traditional panel-only approaches or cookie-based inferences. We must innovate. The only thing worse than failing to predict the future is acting with yesterday’s logic.