Are we measuring the longer-term benefits of response advertising right?

Last month, ITV and media investment analysts ViewersLogic published a cross-market study that proves the value of peak airtime of responsive advertising.

They say that peak TV spots deliver direct responses in the short term, but their hidden value is in their longevity, generating responses long after other TV spots advertising have finished working.

Omar Oakes sat down with ITV’s Measurement and Modelling senior manager Neil Charles and ViewersLogic co-founder and CEO Ronny Golan to discuss the faults in media measurement today, and how a longer-term view can be accomplished.

Omar Oakes: Maybe it’s a media owner problem. Maybe media owners in general don’t have the capacity to be able to prove the longer-term benefits of response advertising in the way that this research tries to.

Ronny Golan: Oh, totally. I think that in general, if you look at measurement of media today, most of it is wrong.

Even if you look at the thing that is the most basic, like online ads. Online ads is very simple to measure, presumably, because all you need to see is whether someone saw the ad and clicked on it.

But in fact this way of measurement of online ads is very problematic, because what we can see in our data, once you look at single-source data, you can see, for example, that people who click on an online ad were primed by TV to do it.

So in fact when Google or Facebook or any other online company says, We give you ROI of X, this is not true. Because that ROI is based on TV advertising.

What we see many times with companies is that they stop TV advertising because they say everything comes from online advertising, and suddenly they see how their cost-per-response starts to increase online.

In general, single-source is important for measurement for any medium, because nothing happens alone. TV effects online, online effects mobile, et cetera. And only if you measure the entire exposure can you really understand the effectiveness of media.

OO: Neil, is that something that bears out in the numbers that you see internally at ITV? To what extent are direct-response advertisers ignoring peak time?

Neil Charles: I’m not sure it’s an ignoring. It’s a difficulty proving the value of, potentially, which is what this study speaks to.

One of the things we did in terms of that difficulty of measurement — I think it’s a fair challenge to ask if we’re measuring well, even up to the gold standard of things like full econometric modelling — and one of the things we wanted to do with this value of peak study is to say that actually some things are very difficult to measure just for one advertiser or one brand.

Even if you have single-source data, some things are quite slow mechanisms, quite shallow, small, but long-lasting mechanisms. And they’re quite difficult to measure for a single brand.

When we built this study, we picked off the top 50 largest advertisers on ITV1, just to get a big dataset of brands. Within that you’ll end up with brands like Specsavers and eBay within this study. That’s the joy of the ViewersLogic dataset is we can see people visiting those websites without having to work with the individual advertisers to do it.

But the models don’t work, even on individual level data coming from ViewersLogic, at a single brand one at a time. No one of those brands could have run this model on their own; it only works because you’re rolling up 50 brands for a year and asking: this quite small but very long-lasting effect from peak, can we see it?

There would be a few individual brands in there that might be able to do it, but in general you could be the best analyst in the world on a single advertiser working on the best data, and you still wouldn’t be able to get it.

You have to roll up at least tens of brands to make this kind of study work, and then you can start to say we’ve got this lovely short-term spike that everybody’s good at measuring — and when you’re working on a single advertiser, that’s what you end up measuring by default because it’s what you can see, what your models can see is a day’s response or a week’s response. What this study is saying is there’s a smaller response that we’re measuring out to three months — it probably goes further — which is very difficult to measure on a single advertiser.

You’re contrasting this short-term big fat spike that’s nice and easy to see within a day or seven days with something that’s much smaller and harder to detect, but when you add up that whole three months, it’s worth more.

That’s why we’re saying you’ll undercook the peak effect by two-thirds versus undercooking the daytime effect by a half; it’s because there’s that more longer-term effect.