Advertising on television is both dynamic and expensive. For a month’s worth of digital campaigns, you get one spot, and the context around that spot can shift dramatically at any given time. Media practitioners can plan the channel, program, and timing of their ad, but not the day’s political happenings, international events, or weather. Marketing, especially on TV, depends on psychology – and psychology depends on immediate circumstances.
Success in this environment demands a meticulous understanding of which stations, programs, and creative are working at any given time, and why. All systems must align to get the right message to the right people at the right time: spray and pray is no longer an option. But testing is a challenge. Due to limited inventory and short time frames, there is often little or no opportunity to fine-tune your placements in the traditional way.
Standard attribution solutions struggle to keep up as well. Some tools can take up to months to return optimization recommendations, which are then essentially moot given how quickly market psychology changes. Even a lag of one or two weeks no longer cuts it; while regression algorithms may work in less time-sensitive scenarios, there simply isn’t enough runway for them to be effective in today’s DR television landscape.
Which brings us to real-time attribution. “Real time” suffers severely from buzzword-itis, with varying and flexible definitions depending on delivery. Some attribution providers claim to operate holistically in real time, but only measure digital day-to-day while TV lags far behind. Others claim that the mere idea of real-time attribution is either impossible and/or without value.
This argument typically stems from being overly literal. Is to-the-second attribution an option? Perhaps not. Is it possible to analyze and act on attribution data hourly or daily in order to keep up with today’s media climate? Absolutely. With a blended algorithm model, such as the Ensemble Method, media planners are no longer beholden to outdated time-reliant data analysis.
“Real” real-time attribution accurately assumes what is happening in the marketplace based on a subset of data. Instead of relying on post-log data from two weeks ago, it analyzes daily pre-logs against up-to-the-minute web traffic. Say you aired a spot on Fox News during prime time. Based on the data modeling, we can determine that it worked – site traffic spiked. As a result, the media planner can go back to the station and make sure that particular inventory, in those particular spots, are clearing, and potentially add or subtract inventory in other locations based on performance. If a spot is working well in prime time, you shouldn’t pay as much for morning spots, or vice versa. You now have the ability to negotiate based on actual information.
These real-time optimizations carry significant weight given the size and importance of TV ad spend. Even a two degree improvement – and when done right, real-time attribution is capable of much more – can confer tremendous benefit to a million dollar per week campaign. This easily offsets the cost of the tool, especially when compared to less actionable, more long-term data.
Real-time attribution is not a myth, a fad, or a marketing scheme. It’s very real, and very important to any advertiser dedicated to optimizing their marketing programs as quickly and effectively as possibly. Market psychology changes daily. Make sure your media strategy can keep up.