I wrote the following article on MediaPost in March 2008 and offer it here as a launch point to a fuller discussion of set top box data that will occur in the 4th Quarter 2009 with Mitch Oscar's set top box data initiative.
The challenge with set top box data has been, and continues to be, lack of standardization across data providers and platforms. Different definitions for the same term, different calculations for the same metric, etc all contribute to industry confusion and delay in acceptance. This article briefly outlines some of the concerns:
What Do you Mean Set Top Box Data?
By Charlene Weisler, SVP Research Rainbow Networks and Services
With all the recent Mediapost articles on set top box data, I thought I would add my two cents on set top box data measurement. The more I researched the issue, the more I found different methodologies for the same piece of data.
The media marketplace is in a constant state of change. Off platform, non-linear, broadband, DVRs and HD all create a vibrant yet unsettling advertising environment. The current standard of measurement needs to keep pace.
Enter set top box data. Described by media experts as a new standard of measurement, set top box data resides in household cable boxes and tracks user activity. Tracking varies by system and by processor. Currently various MSOs and Satellite providers are partnering with measurement services like Nielsen, TNS and Rentrak. In this rush to market, it may be helpful to explore definition of terms of measurement and address issues. Here is a short list to start the conversation:
The issue of “Polling” vs “Census” data where polling data is a subset of the full footprint. Data processors have no control over the amount of data they can receive from an operator. Therefore some data are census and some are a subset of the full footprint. Since sampling methodologies vary from operator to operator, the data must be used directionally until there is a census (or a consensus) for all data sets.
Appears to vary by system. Some pull data on a second by second basis, others during an interval of time (like quarter hour intervals) and others pull at activity points like channel change.
Rating or Delivery:
In order to gauge viewership, the boxes are “pulled” at a certain point in time. The channel that the box is turned to at that moment is the channel that gets credit for the viewership. Box pulls might be regarded as ratings (in the case of a partial footprint), delivery or perhaps a new metric?
Smallest viewing increment:
If someone is channel surfing, then a pull to that box for that second would not represent actual viewing. Most processors say that the standard length of time should be in 5 second increments. However Rentrak uses a formula to calculate a standard length that varies by programming genre and changes as the amount of data increases.
DVR metrics need to be decided. According to TNS this data is not currently available in the U.S. but is available in the U.K. Rentrak says that it depends on the operator and the device. Some operators have trick mode data available and Rentrak says that it is in the developmental stage.
Lag time as the box changes channels or uploads. How is this viewing ascribed, if at all? The standard here appears to be at 5 seconds but this could vary based on the operator and their platforms.
Some processors are using channel change activity to denote tuning events. Others say they are using 5 second intervals. A combination of both might indicate actual viewing.
Nielsen uses the term “dwell times” to describe tuning event intervals. Dwell times are impacted by “latency”.
Picture in a picture as well as simultaneous viewing/ net usage:
Cannot currently be measured. Therefore no standards are in place yet.
Differences in set top boxes:
Can impact measurement insofar as the output data looks different. DVR capability (or not) impacts data because of behavioral difference.
“Set Top Box On TV Off”:
Current processors offer varying solutions. Nielsen match streams to the people meter boxes. TNS and Rentrak use a series of algorithms. Another company will use an entirely different data source that is currently in development.
”Lack of Demographics”:
Nielsen will match to its sample. Other techniques include data fusion with lifestyle, segmentation and geodemographic databases such as PRIZM and Cohorts and purchasing behavior databases such as Experion.
Box availability within the home (“Not all homes have boxes” or “Not all televisions have boxes” or “Not all televisions in the homes have digital boxes”):
The general position is that the data reflects the digital footprint and makes no claims to measure beyond that. Further, the 2009 retransmission makes the issue irrelevant.
A way to get data back to the operator. Cable systems have a back channel but satellite operators can only send data back via the phone lines. Since not all satellite set top boxes connect to phone lines, this must be taken into account when receiving set top box data from satellite homes. TNS says there is no significant difference in the satellite homes with and without the back channel. Rentrak says boxes are connected in different ways – some need to back channel to get scheduling data, others connect via broadband. It all varies by operator. More examination may be in order.
One last point:
All this is emerging and changing – what an operator uses today can evolve or change tomorrow. I am sure that there are other issues and terms and would appreciate your input. Please send it on.