Television measurement: for better and worse
The current approach to quantifying television consumption was most accurate when there was a limited choice of channels, there was no other viewing choice and viewers watched on TV sets. In this context, a sample of a few thousand viewers was an accurate guide to how many people watched each program, and the share for each channel at each point of time could be estimated with a high degree of certainty.
While watching television has remained a firm feature in the lives of billions of people around the world, where, when and what we watch has evolved, necessitating a move to hybrid measurement. However hybrid measurement is likely to be a work in progress in 2014, with significant iteration required to get the best out of the additional data sources. In the long‑run this new approach should be more accurate; in the near‑term it may introduce some distortions.
For example adding in broadcasters’ video‑on‑demand server data has the potential to make measurement more accurate; server logs can tell exactly how many programs have been requested and, for streamed content, how long they have been watched for. However as of 2014, in the majority of cases, these logs do not measure how many people watched each program; while it is likely that a program streamed to a smartphone is being watched by one person, that content may be mirrored on to a television set and watched by a household. Further, if programs are downloaded to be watched later, the service may not measure if, or for how long, the content is watched. Including VOD data requires all entities that provide viewing data to have the same parameters.
The key advantage of incorporating STBs into measurement is their quantity: there are hundreds of millions of units around the world which can log which channel they are tuned to. But STB data has three principal deficiencies. It cannot tell who in the home is watching each program. It may not even know if the TV set is on: a STB may remain on, and tuned in, when the TV set has been off for many hours. And finally, the platform owners collecting STB data may not know the membership of each customer’s household.
Analysis of STB data along with measurement data enables the development of algorithms that can interpret STB patterns better. For example a STB switched to the same channel for two hours after midnight, with no zapping between ad breaks, is likely to be connected to a TV that has been switched off.
Measurement of television viewing is getting more complex, and as a result may get more expensive. Adding additional devices and measuring viewing of foreign TV schedules are technically possible, but add to costs, possibly significantly. In some regards, fully comprehensive measurement, which includes a range of foreign TV schedules, may not be worth the effort or the cost.