By Barry Flynn, Contributing Editor
TVSquared, which claims to have the most accurate TV advertising attribution solution on the market, says its new Predict platform – which allows agencies to forecast what ROI a TV plan will deliver – represents “the first step towards programmatic TV”.
‘Attribution’ in this context means establishing what impact a particular TV campaign or spot is having in terms of actual consumer response, such as website hits, call-centre contacts and online purchases.
This is far from straightforward, says Fiona Smith, Managing Director EMEA and UK at TVSquared, since accurate attribution requires the ‘noise’ from all other marketing and TV advertising activity to be filtered out of the data, a reliable baseline to be established, and the peak caused by the ad to be isolated, before the response that derived from it can be accurately measured.
Her company’s existing ADvantage product is able to achieve this on-the-fly by applying machine learning techniques to the flow of data, enabling airtime buyers to “optimise their TV campaigns in near real-time,” says Smith.
She adds that ADvantage “differs from anyone else in that we also take into account the whole consumer journey. If they are encouraged to visit the website at the back of the TV commercial but then come back four days later on the back of a Facebook ad, we attribute that accordingly. So we attribute some of that to Facebook, and some of that to TV, depending on the length of time before they returned – which is a really unique thing that no other tools do.”
TVSquared’s evidence shows that use of the system typically results in a 25% decrease in ‘cost per response’, and a 30% increase in either sales or registrations for the response-based advertisers using the system, who number around 300 across 42 markets.
What the new Predict product does is to take the accumulated historical attribution data from the ADvantage platform, and leverage the same machine learning techniques to develop algorithms able to precisely calculate what the response rates will be from a particular airtime plan.
TVSquared’s analysis of response data indicates that the best 5% of advertising spend typically generates 20-40% of the audience response, while the worst 40-50% of spend only accounts for 10% of the responses.
“It’s looking at your historical data to tell you exactly where and how you should be buying your TV,” explains Smith. “So it will look over months of data and tell you the day-parts, TV channels, programmes, creatives, that are best at driving conversions. What it does is to allow the planners and buyers for the first time ever to put a media plan together and go to the client and say, ‘this TV plan is driving this many site visits, this many conversions, this many sales, at an overall cost per ad of X.’ It’s a totally different way of looking at how TV has been planned and bought.”
Smith observes that while “everybody is going on about programmatic TV, programmatic TV is not going to be here any time soon. What we should be leveraging is real-time TV data,” she argues. “And that’s Step One in programmatic TV, I think.”
When programmatic TV buying does eventually become a reality, Smith sees the Predict product being integrated into real-time exchanges and agency trading desks, such that agencies would be able to access real-time data about what spots they should be buying and at what price to maximise ROI.
“I get really excited about this,” says Smith, “because it’s holding TV accountable by digital metrics: TV is effectively becoming more digitalised and being made to work harder.”
“There will be a real need for TV buyers to start thinking more like digital buyers,” she concludes.”There’s no reason why TV should be inefficient.”