ITV recently released an article detailing their findings around DRTV and what type of airtime drives performance. ITV accounts for over 40% of the linear market. They still have the biggest commercial channel, and despite recent declines, attract the biggest audiences in the commercial space. So, what do they have to prove? Let’s start at the beginning.

The birth of DRTV 

Performance TV is usually called direct response TV (DRTV). It was a technique first coined by Channel 4, selling remnant airtime to brands that had a strong call to action within their copy, (at that time, with no internet, that CTA was usually a telephone number), and a whole industry of TV buying launched. Since then, brands have sought to understand where the most effective airtime can be found, and which spots drive the most response. To this day, the definition of DRTV in the industry is daytime TV, Monday to Friday, usually the least expensive airtime.  

This is an issue for ITV. They are usually the most expensive in the market (for good reason). Some of their channels offer enticing environments for DRTV advertisers, ITV3 for example, but overall, their main channel, ITV1, lacks investment from DRTV advertisers.  

Why DRTV advertisers steered clear of ITV  

If we look at the TV marketplace, there are clear characteristics of a DRTV campaign. DRTV brands invest c. 30% less into ITV1 than All Time campaigns, in fact, they down weight all terrestrial, investing 20% more share into multichannel stations compared to All Time campaigns. They tend to favour weekdays rather than weekends and mostly stick to the daytime dayparts rather than creep into peak. Some advertisers that still use phone numbers as their CTA choose airtime aligned with the opening times of their call centres. But with the age of the internet, this doesn’t explain why the characteristics of a DRTV strategy remain consistent.  

When we again reflect on the origins of DRTV, where the CTA was a phone number, airtime that at first seemed less attentive, became worthwhile to those advertisers who wanted their audience to pick up the phone. They didn’t want engaged viewers; they wanted distracted ones. We started analysing which spots drove the most phone calls and through basic attribution, could find the hidden value of airtime that would drive the most efficient response. As we moved into the digital age, attribution had to become more sophisticated.  

The viewer no longer had to leave their sofa; they could now respond via their handheld device. Brands wanted the equivalent of Google Analytics for TV, which led to the development of the TV attribution models that we see today. 

Unlike online channels, the difficulty with TV attribution is that the viewer is one step removed from the behavioural process; there is no click-through or post-impression view.  

Probabilistic TV attribution models (of which the majority are) work by placing a window over each spot and capturing the response that occurs during that window. Should multiple spots transmit during the attribution window then response is divided according to the share of impacts each spot has accounted for. The general result from these models shows that the smallest airtime is the most efficient, i.e., provides the best response rate and, as a cost per spot in the UK is usually calculated by impacts (viewing) multiplied by CPT, the smallest spots tend to be the cheapest. As this plays out across the schedule, environments with less investment provide the most efficient response rate (similar to an ROI model). The result is that DRTV advertisers, across a variety of sectors, targeting a variety of audiences, all tend to have similar TV schedules that do not favour ITV. 

Attribution for the digital age 

In ITV’s article, Viewers Logic provides the equivalent of the post impression tracker that we see in online media; an example of the deterministic model in practice. In some respect, they correlate data from the TV to any device connected via the same IP address as said TV. This period of ‘post impression’ is indefinite, the Viewers Logic model uses 3 months in the quoted article. Whatever we determine is an acceptable period, the fact remains that this approach has a far longer attribution window than the probabilistic model, which is usually an analysis of the immediate response. 

What Viewers Logic found was that peak airtime, across a 3-month period, generates a 1.4x larger return than daytime airtime. The report goes further to explain that “focusing on short term response inevitably leads to underestimating the cumulative response to advertising and this is true of daytime spots as well as peak.” 

A probabilistic model is matching spots to response. The more spots a channel transmits the more likely it is that at some stage a mis-attribution will occur. This compounds over time and very small channels can start to appear efficient, driving response beyond their reach. A deterministic model matches reach to response; the higher reaching the channel the greater the chance of misattribution and, as such, this reveals the pros and cons in both models. 

The challenge is that when probabilistic models are compared to deterministic models, the two seem to convey an opposite viewpoint of what is and isn’t efficient.  

Which attribution model is right? 

The answer is neither and both. Both models will have an element of truth, but both will also have an element of misattribution. The power of any attribution model is the analyst rather than the analysis. The analyst should consider the following: 

1. Attribution models should rarely be used as a business effectiveness tool, TV attribution tools should be used as channel optimisers, and the business effect should be seen or proved in the macro data, not in the micro.
2. TV attribution should be about identifying trends. A good example is channel attribution. The human (viewer) doesn’t watch channels or platforms, they consume programmes. A channel like Channel 4 can attract a 75-year-old at 15:00 to watch Countdown, and then 3 hours later attract a high proportion of 16-24s to watch Hollyoaks, so calling a channel responsive or efficient misses the point. Identifying why that channel or programme is driving response, and then finding a correlation within the data that supports that trend, should be the goal.
3. Lastly, use group think analysis rather than a single view. An attribution model is incredibly valuable, but viewing TV’s performance through one lens restricts the results and leads the schedule to smaller and smaller environments. Ensemble analysis is key: combine the micro with the macro. Compare the attribution model alongside a regression model, continually testing theories and generating experiments. 

TV’s power is in its mass, and now, its diversity of content. Balanced schedules that reflect real human viewing will generate response in the short and long term far more efficiently than schedules which rely too heavily on machine-led metrics.  

The ITV/Viewers Logic analysis is right and wrong. It’s correct to assert that peak airtime can produce just as efficient response as any other airtime in the schedule, but it’s wrong to assume that peak is always the answer. Analysing all airtime, using different techniques, will ultimately provide the most insight to the planner and the brand.