When "optimization" was revitalized as a hot media topic around 2002, there was a flurry of interest in so-called "program optimization."  The goal was to improve a future plan by allocating plan weight and budget not just to dayparts, networks and other groupings, but all the way down to the level of specific individual programs. Unfortunately, a tremendous amount of that discussion back then misstated what such systems actually did, and the overall goal of such procedures was fraught with serious technical, practical, marketplace and speed problems.

 

Let's look at some of the problems and pitfalls of that approach.

 

Program Optimizations May Not Be Projectable to the Future

 

Is an "optimized" schedule, inherently based on past program audiences, projectable to future periods?  After all, over time programs are added, canceled, rescheduled, have their characters and plots revised, and are run against different competition, so no two periods are really alike.  Moreover, people change, and their program preferences change.

 

 

Program Optimizations May Not Be Projectable to the Real World

 

Due to the huge number of interactions involved, a set of programs that produces an "optimal" result against a Nielsen respondent panel may not produce optimal results against the population at large.  A program optimizer might well make placements on specific programs in order to reach specific Nielsen respondents so as to squeeze a few more of these individual people into the "reached" group; such "optimization" may have little practical implication for the population at large.

 

Optimizations that are instead based on program groups, networks, days and dayparts have less statistical wobble and can be more safely relied upon.  TView provides a very unique cross-validation capability to demonstrate the projectability of its estimates.

 

 

Program Optimizations Magnify Sampling Variation

 

Attempts to optimize down to specific programs thrive on high-water marks, thus distorting reach estimates.  No ratings figure is taken as absolute truth -- all ratings are estimates based on sampling and will have estimates of statistical error ranges.  But a program optimizer takes these ratings far more literally.  Suppose in the general population, either of two programs would add the same amount of reach to an existing schedule.  When a Nielsen survey is conducted, however, there will be a statistical error (quite normal) and it is unlikely that the two shows will yield the same reach improvement within the Nielsen panel.  Nonetheless, a program optimizer will pick the show that has the highest reach improvement, even if that is due only to choosing those high-water marks on statistical fluctuations.

 

Optimizers that are based on groups of programs, or days, or dayparts, have much more stable information to work with, and are not as susceptible to this kind of failure.

 

 

Channel Surfing Distorts Program Reach

 

Some portion of the audience to every program is produced merely by brief stops by viewers during channel surfing.  This pattern is even worse for lower-rated programs.  A program optimizer that is looking for specific shows that will reach specific members of the Nielsen respondent panel might choose a program simply because a certain specific Nielsen respondent once dallied on it for a moment while surfing.  This behavior of one person could hardly be expected to occur again or to be projectable to the population at large.

 

 

Program Optimizations Lead to Cherry-Picked Schedules at Unrealistic Boxcar Prices

 

Most television sellers will charge more for a highly customized, cherry-picked schedule.  This added cost is not factored into the a program optimizer's deliberations, and thus makes the "optimized" schedule look more alluring than it really is.  The real test here is whether a cherry-picked schedule can produce better reach results even after building in the somewhat higher cost.  Of course, even that may not be a good test in situations where a television seller simply will not sell a cherry-picked package.

 

 

Program Optimization Overemphasizes Reach

 

The reasons to choose one program over another are generally far more profound than whether one can add another specific Nielsen panel person or two.  Goosing up reach by a tenth of a point (or less!) is nowhere near as important as finding shows that provide the right viewing environment, good stewardship, desirable placement among and within commercial pods, and responsible merchandising and dealer network tie-ins.

 

 

Conclusion

 

If you'd like to learn more about these program-level approaches and their problems, we encourage you to read our white paper, "Television Optimizers -- Can You Do Better?" Just phone or e-mail us for more information.

 

Instead of fine-tuning based on meaningless decimals of reach for specific shows, TView looks at program groups, networks, days and dayparts to find the ideal regions of interest (after all, "media is a mesa, not a mountaintop") and then provides unique tools like Faves and Switchpitch to help planners make intelligent, guided decisions based on the factors that really matter.