MediaWave Actionable Insights and Industry News for Media Professionals

Machine Learning: The Future of Automated TV Buying

July 29th, 2019   ||    by Gary Milner

There is no doubt that the business of TV advertising and the viewing landscape overall is changing. Recent subscriber losses total 946k at AT&T, 224k at Comcast and 141k at Charter, with some of these subscribers moving to IP-based, connected TV. Exactly who is switching is not clear, but Hulu now has 28 million subscribers, with 50% growth in viewers last year and up 12% in 2019, according to Variety. In time, this will affect all TV planning and buying – measurement systems will fundamentally change from generic reach and frequency to more defined audience targets and specific outcomes.

With Disney, NBCU, AT&T, HBO, Apple and others launching streaming services, visibility of cable and satellite alternatives will increase to consumers, as well as accelerate the migration to a digital world and new planning and measurement systems, while leveraging artificial intelligence.

In the meantime, many of the existing human-based TV advertising planning and buying processes and systems still exist between stations and agencies / brands.

One of the more time-consuming parts of the whole TV advertising process is forecasting ratings. You can see research from Videa in the transformation of the TV advertising space at: which proves this is the case. When you look at stats like this, you can see that applying technology to this process has a significant benefit to not only the speed of the process but also costs. With 20% of a buyer or analyst’s time spent forecasting ratings points (their #1 most time-consuming task), any automation of this will drive efficiencies and save time and money for stations and agencies.

Forecasting ratings is the process where advertisers, media buyers and marketers evaluate ad campaigns by looking at both how much reach the medium offers, and the frequency at which the viewer sees the ad. The metric they use is gross rating points or GRPs, which is calculated by multiplying the audience reached by the frequency of its exposure to the message during a given period. A rating point is one percent of the potential audience, meaning a show that has a rating of 10 points gets 10 percent of the viewers. So, if a TV ad has a reach of 30 percent of its target audience, and the ad shows four times, the ad campaign has 120 gross ratings points.

So, how could this process be automated to drive efficiencies in the TV planning system?

Videa’s platform, which is an end-to-end solution that automates local broadcast TV ad buying and selling, is using machine learning— an application of AI— to speed this planning process up with no evidence of any loss of accuracy or precision.

This benefit does not just apply to shows that have a solid history of ratings which are easier to forecast. The system can provide estimates of new shows by leveraging past data on show type, timing —really any external factors (holidays, special events, sales days, etc.) — and forecast a rating to be used in planning.

Accurate forecasts mean less over and under delivery of campaigns, leading to less administration time with clients and ultimately saving money.

The use of machine learning in Videa will drive other new innovations as automation further develops. Clearly, there is no better time to start the journey toward automating your local TV ad buy.

Visit to view the full study results comparing agency derived ratings compared to Videa’s forecasted ratings and the actual ratings delivered.


Tags: , , , ,