From impressions to ratings – not such a giant leap
Impressions continue to be the default way in which publishers sell online inventory, but this misses a trick. Nielsen’s Phil Sumner argues – with new evidence – why it’s not such a giant leap to sell online inventory based on TV-equivalent Rating Points.
In 2012, Nielsen launched a service in the UK aiming to provide currency-quality daily ratings for online ads that could be compared to ratings for TV ads. Called Online Campaign Ratings (OCR) one of the things we expected it to change was the way digital media is bought and sold, namely, moving from sales models based on impressions to those based on Gross Ratings Points (GRPs).
Whilst we’re starting to see some online publishers offering guarantees based on achieving a TV-type audience benchmark, there is still hesitancy to adopt a GRP-based model. This means there is a large opportunity for those that do and here’s why it’s not the giant leap one may expect.
Throughout 2014, we started to see various publishers adopt guarantees, underpinned by an understanding and transparency of their audience provided by OCR. The most notable example in the UK is AOL, which now uses it to guarantee audiences of its pre-rolls. But adoption of this model is still relatively limited here. In France, the likes of MSN, CCM, Weborama and Prisma Pub have all quickly jumped on the opportunity.
Thus, there’s still a huge opportunity for sellers of media in the UK to seize the agenda and adopt this way of working. Why? Because there is a clear (and unmet) demand from both advertisers and agencies to move towards basic guarantees and, ultimately, a GRP-based guarantee model similar to what’s used in TV.
Publishers that stick their necks out now stand to steal a march on their competitors. Furthermore, adopting a fully-fledged GRP-based model is, in fact, the least risky way of offering guarantees. For instance, if a publisher guarantees hitting an audience index of, say, 150 but then delivers an index of 145, then there is little a publisher can do, other than offer some form of compensation. (What’s more, this may leave the buyer with a negative perception of the publisher).
With a GRP-based model, if a campaign has under-delivered, to “make good” merely requires running the campaign longer to achieve its goal.
There is a perception that moving from an impression-based model to a GRP-based model is a fundamental change, but I’d have to disagree. A sound grasp of how a publisher delivers against a particular demographic can be achieved quickly using OCR, and the leap to guaranteeing can be made with very limited risk.
GRPs and impressions are linear
First and foremost, it’s important to outline the important relationship between impressions and GRPs. In this context, the definition of a GRP is Reach x Frequency, for the overall population.
The relationship between GRPs and digital impressions is linear.
What does this mean? In the UK, 1 GRP equates to exactly 620,928 impressions. In effect, a block of 620,928 impressions gives up to 1% of the UK population the opportunity to be exposed once to an advertising campaign.
In practice though, it is highly improbable that 620,928 impressions will reach 620,928 people because of multiple exposure (frequency). If 620,928 impressions are served, then it’s much more likely to generate a result such as 206,976 people exposed three times, on average (206,976 X 3 = 620,928).
But GRPs are blunt and show scant regard for frequency, thus they form an entirely linear relationship against impressions. A brand advertiser can say with 100% certainty that if they buy 62.09m impressions, they will definitely achieve 100 GRPs. Figure 1 demonstrates this relationship drawing on 1,107 observations from UK campaigns measured by OCR.
This relationship shows that it’s possible for all digital publishers today to sell against GRPs, as opposed to impressions, using a very simple calculation. But this is just the first step towards a Target GRP-based model, where specific age and genders (targets) are bought and advertisers pay for only those eyeballs within their specified age group (as per TV). To move to this model, publishers need to have built up a robust data set from which they can have confidence to offer guarantees. Our data can help give them this assurance.
What does this mean for publishers?
Using actual data from 334 UK campaigns, Table 1 shows the performance of six randomly selected large media sellers. The data was run across one of the most popular demographic groups we measure, ‘People 25-54’.
For each media seller, we’ve built up a stable and robust picture of how well they can deliver against this group; it’s a consistently positive story.
For example, Trading Desk #1, on average, can deliver a Targeted Rating Point (TRP) for people 25-54 with only 457,331 impressions. In practice, they can guarantee 20 TRPs and know that this will roughly require them serving 9.14m impressions (20 x 457,331). The story for Publisher #1 is even more positive. They can deliver a single TRP for People 25-54 with only 274,818 impressions, i.e. more than twice as efficiently as natural delivery to the population.
Media sellers can report how they deliver to certain demographic groups and how many impressions it will take them to achieve this. Once a suitably robust pool of data has been built up, publishers have the power to be bullish when providing guarantees across a range of demographics, with minimal risk of under-delivering and having to provide rebates.
In the example given in Table 2, there are various common age/gender targets laid out with their average impressions-per-TRP. The table demonstrates that the trading desk is capable of delivering efficiently against all of these age breaks, compared to natural distribution. For example, it would be possible to guarantee 20 TRPs for men 25+ with roughly 8.3m impressions, whereas it would take around 11.3 million impressions to deliver the same TRPs against Women 25-49.
Seize the day
It is understandable that publishers are nervous about changing their commercial models overnight. Moving from an impression-based sales model to a TRP-based sales model seems like a huge change. But it shouldn’t, and here’s why.
Firstly, I don’t think the models are mutually exclusive; it’s possible to continue operating existing commercial models and to overlay a TRP selling model. Secondly, the risk of over-promising and under-delivering against an audience will largely evaporate after only a few campaigns have been measured, as we see stable data emerge relatively quickly.
No publisher wants to be in a situation where they have to offer significant rebates or run millions of additional impressions to reach a target, but this risk should be low and publishers can easily work within tolerance levels to reduce it further.
What is evident, though, is that those who do adopt these models now have an opportunity to steal a march on their competition and seize the day.
Phil Sumner is UK media director of Nielsen