|
|
|
|
|||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
MEDIA MALPRACTICE
Why Our Traditional Approach to Measuring Media
No Longer Tells Advertisers What They Need to Know.
By Erwin Ephron
|
|
This upfront, four Cable TV networks and the Cable Television Advertising Bureau presented data on measures like loyalty, viewer involvement and brand resonance to suggest their viewers are more likely to see and remember commercials carried by their channels than by others. In Print, eight publications, spearheaded by Britta Ware and the Digest, have been promoting measures like frequency of reading, reading time and one of my favorites to make the point that their readers are more “involved” and are therefore more likely to see remember the advertising carried.
On the buying side, the five largest TV agencies have each developed proprietary systems using less orthodox measures like duration and frequency of viewing, program involvement and commercial recall to better judge attention paid to commercials carried by different types of programs. The ratings are failing Why this research pandemic? I think we want more than our traditional ratings measures of exposure because we know they have become a poor proxy for people seeing or hearing advertising. This paper confronts that growing problem and suggests a few difficult remedies. It is a timely reminder of both how much and how little things have changed in the world of media and research. The medium isn’t the message Most of us are at least subconsciously aware that the audience counting systems that direct billions of advertising dollars are not working as well as they once did. We also know that many of the Nielsen, MRI and Arbitron viewers, readers and listeners that advertisers pay to reach are not likely to see or hear their ads.
To understand the roots of “Media Malpractice” we need to understand the severe limitations of the media measurement model, where the basic unit is the exposure. What is an exposure? A media exposure is someone seeing or hearing a media vehicle (program, title, website, station) carrying an ad. It is in many ways less than it appears to be. Exposure is what researchers grandly call a “stimulus” measure in that it requires no consumer response to prove it happened, just consumer proximity. [i]
Another problem. In the US, exposure typically refers to the medium, not to the advertising messages the medium carries. To bring advertising into the media planning model, researchers had to invent another name for a media exposure. They called it “an opportunity to see an ad”, or an “OTS” for short. [2] To nail the point, we need to realize that OTS for any medium, as the word “opportunity” makes clear, must overstate what we are really trying to measure, which is “people seeing advertising.” So at best our measurement system reports exposure to media carrying ads that people may or may not see. Outdoor: A Beta test We began with the statement that many of the reported viewers, readers and listeners that advertisers to pay to reach are not likely to see or hear their ads. That’s because advertisers pay based on “opportunity to see”, not on consumers seeing ads. How do we move our planning numbers from something as vague as “Opportunity to see” to something more relevant like “likely sees”?[ii] Among all media, Outdoor has the most serious OTS problem. Its audience estimates are based on traffic-counts of cars passing by Outdoor locations.
In the UK and overseas an adjustment to Outdoor exposure estimates based on secondary measurements has been introduced to make the exposure numbers a tighter fit. The total system is called Visibility Adjusted Impacts or VAI. VAI reduces the audience count for an Outdoor display based upon the probability that a driver or passenger passing by will see it. The measured variables that determine the probability of seeing the board include size, distance from road, angle to road, duration of exposure, etc. A major advance in research thinking This is a major advance in media research thinking. If Outdoor opportunity-to-see is too loose a measure, we can use other measurements to tighten it. We can adjust Opportunity to See by Probability of Seeing to make it a better measure of media value. And the idea is applicable not only to Outdoor, but to all media. Falling commercial recall scores and negative ROI’s reinforce the idea that the difference between OTS, our measurement, and sees my ad, our goal, is getting larger every year. This suggests we should apply Outdoor VAI-type discounts to all media, based on other measures like communication, persuasion and sales, to bring their audience counts closer to sees the advertising. [iii] Not a new idea They called this Ad Perception. The recommendation lost on a technicality. Ad Perception is not a pure measure of media effect, because you can only measure it as ad communication, which is also a function of the ad. [iv] Today using advertising communication to measure media performance is not a deal-breaker. We are practiced in the mathematics of isolating the effect of a variable in a mix of variables. There is no good reason why the exposure measure of all media should not be advanced from “Opportunity to see” to “Likely saw” the ad through the judicious use of attentiveness, communication and sales measures. Let’s start with Television. Adjusting TV Based on a wealth of relevant information, heavily weighted to commercial recall and self-identified attentiveness studies, I have used three simple adjustments to calculate the “likely sees the ad” exposure counts delivered to advertisers: These are: Let’s look at each in turn. Size of unit • 15-second audiences are discounted by 50%, compared to 30-second commercial audiences, which are not discounted at all. We know from research and brand experience that shorter units have a far lower probability of being seen, remembered and recalled. The CAB/Nielsen study showed recall of 15’s was half of that of 30’s. But opportunity to see, our TV exposure measurement, assumes different size ads have the same opportunity. This encourages the penny-wise to substitute shorter units for longer units in TV plans since there is no evident penalty in plan dimensions (TRPs or reach and frequency) at the lower cost. We think the advertiser is better served if the difference is made obvious by adjusting down the count of viewers for shorter form commercials. The unacceptable alternative is to consider all commercials as equal, regardless of duration. Pod length • Pod length. One-to-three-commercial pods are discounted by 10%, Four-to-six-commercial pods, by 20%, seven-and-more-commercial pods by 30%. Program promos are counted as commercial messages. [vii]
The adjustments are again based on ad recall studies, specifically those conducted by Nielsen for the Cable Television Advertising Bureau in the US and proprietary UK research, done by Billett’s, a leading media performance monitoring firm. [viii] The number of commercials in a pod strongly affects commercial recall. Part of the loss is in memory, but much of it is the result of commercial inattention and avoidance encouraged by long program interruptions with too many messages. For that same reason, program promos, when scheduled as part of predictable break pattern, should be scored as messages contributing to pod clutter. Day part and program type by demo • Daytime is discounted, on average, by 10%-to-20%, Prime by 10%-to-20%, late night by 15%, early news by 15%-to-30% and team sports, by 5%-to-10%. These are averages. Specific adjustments will vary by demo. The day part and program-type adjustments are based on a number of viewer attention measures buttressed by ad recall studies. For example, the Nielsen/CAB study (among others) found that viewer reported attention correlated directly with their ability to recall commercials, a strong indication that the commercials were seen.[ix] Data from the continuing Simmons Marketing Research Bureau study show younger viewers report themselves as less attentive than do older viewers (-10% to -15%). Men, although they view less TV, consistently report themselves as more attentive than women (+5 to +10%).
Reported attentiveness by program type divides into programming with a strong plot line, which better captures and retains the viewer’s attention and more loosely formatted programs like variety, talk and news magazines, which produce less focused viewing. Reality programs are in between. Live information such as News and Weather attract attentive viewing. Men report paying high attention to team sports.[x] Combining the adjustments This helps to compensate for the effects of double counting. For example a 30-second commercial in a 6-commercial pod running in a primetime drama program with a W 18-49 target, would be adjusted as follows: • Unadjusted W 18-49 commercial audience: 5,000,000 [xi] The “likely sees” the commercial audience is 3,600,000 or 72% of the reported audience. A straight sum of the adjustments (5,000,000 – 30% of 5,000,000) would put the adjusted audience at 3,500,000 or 70%. If it was a 15-second unit the adjusted audience would be 5,000,000 – 50% = 2,500,000 – 20% = 2,000,000 – 10% = 1,800,000, or 36% of the reported commercial audience. A more elaborate adjustment system would model the combined effects of the adjustment variables as is currently done in the Outdoor VAI calculation. Rethinking reach Recency theory shifted the planning balance from frequency to reach in the late 1990’s, at just the right time. With TV costs rising sharply, the old frequency planning goals were about to break the bank.[xii] Recency’s central idea that advertising needs moderate reach and continuity instead of the old high reach and effective frequency has kept TV cost-effective on paper. But if OTS-based frequency exaggerates sees-the-ad frequency, then reach isn’t reach anymore. And that is our situation. When reach isn’t reach anymore Let’s assume our adjustment for “likely sees the commercial” averages 60% across a television schedule. The exact number isn’t important, the size of the difference is. Adjusting schedule TRP’s and impressions for “likely sees the commercial” is a simple matter. If the average spot on the schedule has a 60% probability that a target consumer will see the message, then the adjusted impressions and target point totals are 60% of the Nielsen reported numbers. But, what does this do to reach and frequency? It’s obvious that our reach estimates are also too high, but adjusting reach is nastier than adjusting ratings, because the reach of a schedule is not a linear function of its TRP’s. Reducing TRP’s by 40% does not cut the reach by 40%. An easy way to recalculate reach for “likely sees the ad” is to move to the lower TRP value on a generalized reach curve, but that’s sloppy. The better way to is to understand that a Nielsen frequency of one isn’t reach any more and recalculate reach by using the adjustment factor as a probability and applying it to the frequency distribution of the schedule. [xiii] For example if the reported Male 18-to-34 viewers of a schedule, have a hypothetical “likely sees” adjustment of 60% (-40%) on the reported Nielsen data, the probability that the average young male viewer will see the average commercial is 0.6 for each exposure.
The schedule reach and frequency is adjusted for this by reducing the size of the 1x frequency group to 60% and reducing the TRP’s of the total schedule to 60% as shown in Schedule A. In this example, all of the 2x frequency group is considered reached because their “likely sees” probability is greater than one (2 x 0.6 = 1.2).
In Schedule A, 12% of the target was reached once. These 12 reach points are cut to seven (60%), reducing reach from a 36 to a 31. Frequency is recalculated by dividing the new reach into the adjusted TRP’s, and drops from 2.1 to 1.4. Thirty-one at a 1.4 is the schedule reach and frequency recalculated for the 0.6 probability that a commercial will be seen. A second, more dispersed 75 TRP schedule, (Schedule B), delivers a Nielsen reach of 42 and a frequency of 1.8.
Schedule B looses far more reach than Schedules A because it has a larger 1x frequency group. [xiv] Too big a problem? Most agencies consider the growing gap between media exposure and sees the ad too big a problem to meet head-on. No US agency that I’m aware of has attempted to adjust Television or any other whole medium. [xv] Their focus has been on figuring-out which programs have an edge in delivering attentive viewers to their client’s commercials, (using factors like pod-length, duration of viewing, commercial recall, etc.), mainly to keep ahead of the competition and to help inform their buyers.[xvi]
Accountability But let’s think about the importance of a better set of numbers in advertising planning as well as new business. Advertisers are demanding accountability, but how can Advertising be accountable until it uses more realistic estimates of people seeing advertising? Falling recall scores, weak sales response and disappointing ROI reinforce the idea that the difference between OTS, the media measurement, and sees my ad, the real goal, is getting larger every year. For this reason alone, I think even crudely adjusted data are far more responsible than the current numbers. But we continue to look at ways of improving the numbers we spend by as a competitive advantage, not as an urgent industry need. It is institutionalized Media Malpractice. I invite those who would pitch-in to do better to contact me.
[1] The thinking in this paper was encouraged by an assignment from Media Experts, an innovative Canadian media agency. I am also grateful to Roger Baron, Frank Harrison, Andrew Green, John Philip Jones, Mark Sherman and Jon Swallen for their helpful critique of the original draft. While there is strength in numbers, the prickly ideas are my own. [2] For the balance of this paper OTS, exposure and audience are used almost interchangeably. [3] I’m certain there will be disagreement in process; whether I have used the right variables. And there will be disagreement in nuance, i.e., whether a specific adjustment for a television program-type should be minus 20% or minus 10% of the reported audience number. But there can be no real disagreement in principle. The adjusted data will be an improvement. NOTES [i] When we do require a response to determine exposure, like asking whether a program was viewed as we do with the diary, it leads to biased counts. [ii] We might even move to a more definite measure, like “an 80% or greater probability that the ad will be seen.” [iii] Since VAI deals with physical dimensions of outdoor units, some equivalent TV measures might be location of set, presence of children in the room, time of telecast, etc. [iv] Attempts to carefully separate media effects from advertising communication effects do more harm than good, since communication effects, which should inform media planning, are less likely to. [v] Rating level comes immediately to mind as a marker for “likely sees the ad”, after all advertisers pay more-per-thousand for higher rated programs. But over the broad range of inventory available, there is no systematic evidence that higher rated TV is more effective TV. For an excellent analysis see Andrew Green, “Better Scheduling for Bigger Profits” OMD white paper, September 2003. [vi] Here pragmatism trumps the logic of research. We appreciate that shorter commercials may be seen by as many viewers as adjacent longer-length spots and have lower recall because they have less time to communicate. But that is a fact about all short-form commercials, and can be compensated for by simply adjusting the audience size, realizing the true loss may be in communication rather than exposure. [vii] First, second and last positions score higher in recall, but since pod position is randomly assigned, the distilled figure reflects the average score by number of messages in a pod. [viii] Nielsen telephone study for the Cable Television Advertising Bureau, February 2000. “Optimizing Media Investment through Better Placement,” A report from Media Performance Monitor America, Billett’s, August 2004. [ix] A reasonable marker for “more likely to see the commercial” is viewer “involvement” with the program. This is in turn reflected in viewer loyalty (measured as episodes viewed) and duration of viewing (measured as % minutes viewed). This segmentation is called “QUAD” analysis. [x] Estimated from Nielsen/CAB study, SMRB 2003, and data summaries in TV 2004, edited by Ed Papazian and published by Media Dynamics, Inc. [xi] I am assuming that Nielsen will soon supply commercial audience data. [xii] A typical frequency planning goal was expressed as: n 4-week reach at a frequency of three-or-more. [xiii] A similar method is used by POSTAR in the UK to calculate the reach and frequency of VAI-adjusted Outdoor schedules. [xiv]When OTS adjustments are made for all media, target points will go down and CPM’s will go up, but it is the high frequency media (Outdoor and Radio) that will retain more of their reach, because their one- and two-frequency groups are small. [xv] Again, Media Experts, an innovative Canadian Media Agency is attempting to adjust all media for “likely sees.” It is my view that Magazines, Newspapers and Radio are far simpler to adjust than Television is. [xvi] OMD, Starcom, Mindshare, Zenith, etc., all have proprietary systems for improving
the Nielsen national TV numbers. - September 8, 2004 -
|
|
|||||||||||||||
|
|||||||||||||||
|
|||||||||||||||
|
|||||||||||||||
Have you noticed the
increasing emphasis on “qualitative” measures in media? 
In Radio, a major study sponsored by the RAB and its many members reported
that consumers considered Radio their own “personal” medium and
are more ready to react to radio advertising at an emotional level.
Now there has always been “extra” research to go
beyond the ratings, but the current volume of “extras” has become
overwhelming. And all of it attempts to show that the bond between the consumer
and the medium helps gets the advertising seen (or heard), and to work better. 
The erosion of our measurements is the result of many things. More media being
measured, poorer respondent cooperation and record-keeping, growing clutter
and commercial avoidance. But the conclusion is inescapable; our media measurement
systems need refitting, badly. 
Media exposures share this ambiguity with medical exposures. You may be exposed to a billboard and not see it, just as you may be exposed to the measles and
not catch them. 
These counts are much higher than actual ad exposures would be. Higher enough,
in fact, to get the Industry to do something to improve the numbers.
An adjustment for unit size, works for TV and Magazines, not
just Outdoor, why not use it? Most of our TV measurements do not factor-in
attentiveness,
commercial avoidance or leaving the room. Magazine readership assumes issue
readers read all the ads. Why not adjust for these over-estimates also? 
Improving on OTS is not a new idea. The original 1961 ARF media
model Toward Better Media Comparisons, rejected Opportunity
to See as the proper
measure
for media because of its limitations and instead recommended persons seeing
the ad. 
How might we go about adjusting-down TV audience counts to better reflect
people actually seeing commercials?
The length and predictability of commercial pods is an important trigger of
commercial avoidance. 
Early evening viewers (when dinner is being prepared, served, eaten) report
the lowest attention levels, especially for women. Prime and Late Night show
higher levels of reported attention. Daytime, in particular when children are
likely to be present, show somewhat lower attention levels. 
Since there is overlap in adjustments, for example 15-second commercials
tend to appear in pods containing more commercials; live sports and news tend
to have shorter commercial pods and to attract attentive male viewers,
etc.,
adjustments are applied in sequence by size (high to low) to that portion
of the audience remaining after the prior adjustment. 
Calculating the affects of OTS adjustments on reach is an equally important,
but more subtle assignment. 
That means it takes more than one Nielsen exposure for the average young male
viewer to be counted as reached with a commercial message. 
Here the 1x frequency group is 30 reach points; the >1x group is the remaining
12. In this case the 60% adjustment reduces the 1x frequency group from 30
to 18. This reduces schedule reach from 42 to 30 and frequency from 1.8 to
1.5. 
