METHODOLOGICAL NOTE

There are a myriad of difficulties that must be recognized and planned for when deriving sample from cable-TV billing data.

Billing data represents a cost efficient means of sampling cable-TV subscribers.  However, there are potential pitfalls and limitations when using this information.  Depending on the sample target, these difficulties can increase data collection costs, and cloud or even invalidate survey results.  The use of this data for tracking subscriber activity, analyzing promotion effectiveness, etc. can result in similar problems.

Problems typically fall into one of three areas: basic data problems, currency and timing issues, and interpretational difficulties.  It is not our intent, nor is it feasible, to detail every potential problem or circumstance that may give rise to difficulties.  Rather, the objective here is to simply increase awareness of these problems, while providing some background which may aid in developing research plans, communicating with clients, and dealing with the billing services themselves.

Basic Data Problems

The most important thing to keep in mind is that data and coding conventions may be different for each cable system - even when the data comes from the same billing service.  Even though each billing service (e.g., CSG, CableData) imposes a basic record layout and functionality, the actual coding scheme and responsibility for consistency and timely updating remains with the individual cable system.  Consequently, the processing requirements may be different for each system - a complication when dealing with data from multiple systems and billing services.

Outlined below are some basic problems that need to be addressed:

Data Structures. Tapes are most often requested in ASCII format.  Depending upon the format used when the data was originally created, translation may eliminate certain vital codes.  If these codes are critical to your project, additional programming efforts may be required to reconstruct the missing information.

Level of Aggregation. Information is maintained at both a converter and net household level.  Households with different service combinations on different converters will appear quite differently at the household level.

"Non-Standard" Subscribers. This is our term for accounts other than paying household subscribers.  Look closely enough at the data supplied abd you will most likely find "comp", hotel/motel, bulk, and possibly even some TVRO reseller accounts.  The account types you want and do not want in a sample must be clearly defined.

Pay-Per-View (PPV) Activity. If your study requires reference to PPV activity, it is usually stored in a separate file.  The characteristics of this data vary by billing system.  Be aware that movie/event title information is subject to misspellings and/or truncation and that there are no standards for this information.

Again, it's crucial to understand that even though each billing service imposes certain minimal standards, the service coding, packages, promotions, and additional information will probably be coded differently for every cable system.

Currency and Timing Issues

This is the most misunderstood issue related to using cable billing data.  Billing services provide their database as of a point-in-time near the end of each calendar month.  The usual assumption is that this represents and "end-of-month: picture of each household's subscription status, activity, and billing record.  In reality, it provides the status as of the last time the household was billed - the timing of which depends upon a household's billing cycle. 

Cable systme typically have two to four billing cycles, staggered throughout the month.  An end-of-month database is a cumulation of these separate cycles, some of which were closed three to four weeks prior.  This means that, on average, the data is two weeks old when it is cut, and a minimum of four weeks old by the time data collection begins.  Obviously, this has implications for both survey and analytic efforts:

• Some "current subscribers" will have disconnected (possibly as long as two months ago).
• Some recent disconnects will have reconnected.
• "New connects," are a problem if defined as "past month" - almost no one contacted could have subscribed within the past thirty days.  If targeting those who have responded to a promotion, be sure the billing data covers all cycles spanning the promotional period.  If the project involves data manipulation (i.e., tracking of connect or upgrade activity) be aware that many June connects/disconnects and upgrades/downgrades will not appear until July's end-of-month tape.
• PPV activity tracking suffers similarly - even major "events" may not appear on the database for two months following day and date exhibitions/

The most serious problems arise when dealing with promotions during specific time periods, or day and date events.  Currency issues translate to additional screening costs, and/or skewed samples of househoolds with recent connect/disconnect activity.

Interpretational Difficulties

If not handled correcly, the issues discussed above can lead to problems throughout the research process:

• Validation of results against "known" system statistics can be problematic; past month connect/disconnect activity on a household level won't match.
• Sample selection is complicated by multiple outlets.  If you are sampling HBO disconnects, you may get households that just dropped HBO on one outlet, but still subscribe to HBO on other outlets.
• Sample selection is usually at the household level; this can complicate the analysis of questions about "the monthly cable bill" - the respondent may be paying for additional outlets, or be receiving a promotional or introductory price.

The above examples represent some elementary problems that will be encountered with cable billing data.  The billing systems were developed to simply get the bills out correctly and on time.  Their use for marketing, database and sampling applications were not part of the original design considerations.

One final caution: get it right the first time.  Billing systems retain data for just a short time (some for just one month).  Get as much information as possible about each subscriber.  The original data may be long gone by the time you are in the field.