Commentary
Fixing TV's Demo Obsession
- by Dave Morgan , Featured Contributor, August 10, 2017
This must change. Here's why:
Demographics are a lousy surrogate for purchase. As CBS’s president of research Dave Poltrack has been telling us for years: “Age/sex demographics do not have a good relationship with advertising value…No one has ever proved that age/sex is a good surrogate for any product usage."
Audience fragmentation makes broad demographic brackets practically meaningless for TV planning and buying. Today, 75% of national TV audience time occurs on shows with an average rating under 0.5. Yes, the vast majority of TV viewing occurs on shows whose audience scale would have been considered insufficient 15 years ago. However, since total TV viewing hasn’t declined much, this viewing is now spread among thousands and thousands of different shows and all dayparts.
advertisement
advertisement
Categorizing dozens of shows by a handful of broad descriptors worked well in the past. Trying to meaningfully describe the 250,000 different national TV spots that run every day by the same few descriptors adds practically no value to the planner or buyer trying to truly differentiate one from the other. It would be like trying to sell all of the hundreds of different colors of Crayola Crayons in opaque packages labeled only by descriptors “red,” “yellow,” “blue” and “other.”
Pairing demographics with GRPs only makes it worse. The gross rating point (GRP) used to work together splendidly with demographics as its "other half." However, the GRP today has become as dysfunctional an effective tool of marketing objectives as broad demographics.
GRPs measure the “weight” of TV media delivered by one or more spots, nothing more. When we only had three TV networks to worry about, buying weight meant -- by default -- buying reach. Viewers didn’t have many choices, so if you bought a lot of TV you bought a lot of viewers.
No more. In a fragmented world, buying TV according to GRP weight is now a great proxy for buying low reach and high frequency -- and, without meaningfully granular audience descriptors far beyond basic demographics, it’s hard to fix that problem.
The answer?
TV planning and buying needs to shift to people and purchase-based descriptors as its primary currency. Digital has taught marketers the power of granular audience segmentation, defined by thousands of different and valued characteristics, from media behavior to purchase-based definitions to life stages.
Unlike broad demographic descriptors, people and purchase-based descriptors can scale efficiently across TV's fragmented landscape. In fact, the more audiences fragment, the more valuable and helpful granular segmentations become.
Most critically, people and purchase descriptor are the most effective at helping marketers make efficient media buys. Who wants to buy GRPs of “women 18-45” for a laundry detergent campaign when they could buy campaigns with a known reach and frequency of “heavy detergent buyers” or “lapsed Tide buyers” or “new, first-home purchasers”?
I used to think that demographics would remain TV’s primary currency for the next seven to 10 years, and that purchase and people metrics would only be used as “secondary” currencies. I no longer believe that, largely because it's becoming more and more apparent to TV networks that they are leaving massive amounts of money on the table pricing and selling their inventory on demographics and GRPs instead of deep target audiences who represent real, self-evident value to marketers.
To borrow a quote again from CBS’s Dave Poltrack: “The fact is that if I’m an advertiser with a particular product and service, the current pricing of the television business has no relationship whatsoever with the value of that inventory.”
That is about to change. The folks with the most power over that decision -- the TV networks themselves -- have already started the ball rolling with their recent audience-based selling initiatives, from NBCU’s billion dollar allocation to audience to FOX, Turner and Viacom’s OpenAP.
Get ready for the “after-demographic” world of TV advertising.
Dave, the way forward for those advertiswers who care about consumer targeting above all other considerations is to break down the corporate buying systems which account for about 70-80% of national TV time buys. This poses many problems---like small brands getting lost in the shuffle and a huge increase in person power to make so many individual brand buys as well as increased selling costs for the same reason----but if advertisers are serious about change then it's up to them, not the sellers, to do something about it. So far, I see only very limited movement in this direction, but I hope that this begins to change---- and soon.
Sounds great in theory. Waiting to see if the granular data is as good as we wishfully think. Or will we just whistle past this problem and pretend it's ok. How good are we at identifying buyers of any category or product? Why do Experian and Axciom data look so different for simple things like identifying HH with kids.
It works great when the messages and subsequent purchases or actions are observed on digital platforms. No so much when we blindly have faith in third party data for other media. This needs fixing.
You're right Jack. The key will be in execution. Data quality and verification of its quality becomes critical. The industry hasn't done such a great job in that area with digital audience targeting data.
Dave- agree that national TV needs to move beyond age/sex based buys, but how addicted are marketers to Audience guarantees from the networks and the make goods that are almost always sure to follow...coupled with Jack's comment about inaccuracies of even broad demographics like HH's w/children, who will lead the way?
Jack, regarding the "granular data" that is such a big part of what is termed "advanced audience buying", for TV, I remember a time when there was great concern about using single source data as "the gold standard" for determining whether media audiences were also product or brand buyers, rather than mixing info from various sources, with all of the issues that such melding raise. It seems that in our rush to embrace better targeting and our assumption that large numbers of homes in a panel is better than smaller samples, that we have forgotten the lessons of the past. The fact is that "big data" set usage ratings, blended statistically with "third party product buyer/user demos, then applied on top of Nielsen viewer ratings is exactly what we were told was a no no. Keeping an open mind on this---as things sometimes change----I, too, have yet to be convinced that all of this mixing and matching, plus the use of set usage as a surrogate for viewer targeting, gives us a better set of metrics for targeting TV buys. But let's see....perhaps they will demonstrate that this can, in fact, be done on a reliable and meaningful basis.
I totally agree Ed. The market participants will have to prove that they can do this.
Ed, this discipline doesn't have to get to one hundred percent accuracy in order to be valuable. If the granular targeting is correct in identifying audience members eighty percent of the tiem, 4 out of 5, isn't that substantially more reliable and accurate than the metrics broadcast/cable reliy on now? In technology like this, it has been my experience, at least, that attaining 80 percent accuracy is sufficient to achieve results for the client. If we wait till we get closer to 100, we'll never getb these systems out. In addition, the development cost to get beyond 80 increases substantially, so reutrns diminish to the point it isn't worth the spending to achieve incremental improvements, at least all at once. Better, I think, to release software that adds a lot of value (the 80 percent) and make improvements from there. The old line "don't let the perfect be the enemy of the good" applies.
Very good point Henry. Totally agree. I suspect that the quality of the data will be measured by the results it drives - the measurable value it delivers - rather than by an abrstract analysis of how it was acquired, cleansed, balanced, etc. Thus, with the focus more on the results than the process, all improvements over current practices will be valued and the pursuit of perfect won't get in the way of better.
Henry, where did you get the 80% accuracy figure? Just curious. Here are a few facts. When Nielsen tallies the average amount of TV set usage per household it gets something like 60 hours per week---I don't have the exact figure but I'm close. When Nielsen does the same thing with an average resident in its TV peoplemeter panel, the corresponding figure is roughly half of the household number--around 30 hours of claimed and recorded viewing. Hence my point about not knowing who is watching half the time when set usage is assumed to reflect people's viewing patterns.
A second and equally important point concerns demography. Aside from a relatively small number of older slanted product categories, most marketers will find that their product user profiles---especially frequent users---is skewed towards adults aged 18-49 and those with above average incomes. As it happens such households are considerably more likely to use one or more TV sets per average minute than their low income and/or older counterparts. Why? Because younger-middle aged homes have about 75% more residents than older homes and a similar, but less pronounced difference is noted between upper and lower income households. Also, younger/affluent homes tend to have more receivers. The result is that the disparity between TV home ratings and viewer ratings in younger/affluent households is considerably larger than the overall average. These homes may be above average TV users but the adults who live in them are, on average, lighter than the norm when it comes to viewing. Result: the numbers favor the sellers over the buyers, sometimes by a significant margin.
As I said, I'm perfectably willing to await solid evidence that shows that my reservations don't really make a difference an that the "big data" profiling system works as claimed. But, if anyone is interested and subscribes to Nielsen, just take a random selection of nationally-aired shows and tally their household ratings, indexed against the programs' national rating by age of head of house---or by income level--- then compare the findings with those for adults in those homes by the same breaks and see for yourself.
Ed, Very true. I think that in the near term, most of the focus for more granular targeting will be rolled up at the household level, since that's how the purchasing is likely to be tracked and matched. Of course, with the fast growth of connected TV's, and the ability to link those TV's to personal media/communications tools like mobile phones, tablets, laptops and PC's, we'll probably see the development of a pretty robust identity graph that will work well at the individual level as well.
Ed, my reference to 80 percent is based on the degree of accuracy, or reliability, most business and tracking systems in this industry achieve in order to be viable. Certainly, if this was the critical care diagnostic and treatment field, four out of five would not be good enough for a product to be released; we'd want higher relaibility to avoid mistakes with life threatening consequences. But for most industries, including advertising, getting to eighty percent has value. In the past several years working with a number of software firms serving the buy and sell sides, I've made this point. It follows the software development path for other business verticals, and I've seen it hold true for decades. Eigthy percent gives you a viable product to launch, and the extra cost incurred trying to do substantially better is not worth the wait. Better to launch and improve as you go forward. That's what software release cycles are about.
Personal example - Years ago, Google enabled individuals to see how we were profiled based on their attributes. I looked at mine (also my wife, since we used the same pc at home, so our search history from which data was drawn was combined, much like your TV set / household case.) I saw they had our preferences right at a little more than 80 percent. Close enough, considering before this tech came out no one would have a clue what we liked and didn't at the level of detail provided.
I'm with you, Ed. Testing these types of thing is what is needed. Maybe and ARF or CRE project that let's all see what happens would be great.
Thanks, Jack. By all means let some industry body take a look at both types of TV audience data and see how closely they correlate for defining who in the home is watching.
Is there a place where we could see, or has demonstrated, 80% agreement between two sources (let's say Axciom and Experian) for an average consumer product category. The data I have seen shows far less agreement then that for even simple household characteristics like presence of children. If these two services cannot match on something as simple as this, we need to fix the data. The models will work fine if the data is good. And it doesn't need to be perfect, just worth the cost and time.
Henry, I see what you mean, however we are not discussing theoretical correlations or even real ones where the metrics are comparable. My basic point is that set usage ratings are not---repeat not---comparable to viewer ratings, especially for younger/middle aged and/or upscale consumers. When it comes to older consumers who, typically, reside in one or two member homes, there I would accept an 80% correlation----but most TV advertisers know that such audiences are easily reached and readily available no matter what types of TV they buy. It's the others---generally lighter than normal viewers who live in heavy TV using homes where the real discrepancies are found. So why isn't the industry trying to get to the bottom of this? I'm a reasonable guy and would welcome being proven wrong---if the data shows that. After all, who is against true proigress---not I.
In an attempt to see where we are now, do we have any estimtate of what percentage of all TV ads are delivered on the basis of targeting that is more granular than age/sex and what percentage of total TV ad revenue these ads generate? Thanks.
James, a lot of agencies and marketers use deeper data for targeting, and deeper data is used for large trading decisions, but when it comes tactical campaign buying, my bet is that less than 5% of TV campaigns are bought - and posted (meaning the currency paid on) - today on highly granular non-Nielsen targets.
I agree with Dave regarding the degree to which many advertisers define the targets of their ads and use such insights in their brand positioning strategies and the ways they create their commercials. It usually goes far, far beyond adults 18-49 or women 25-54. Unfortunately, because so many marketing directors pay scant attention to the media function, TV time buying is usually done on a corporate basis---about 75-85% of it for national buys----and this leads to umbrella "demos" being used as GRP "currency" as this is the only way for the buyers to protect their fannys and obtain audience guarantees from the sellers. I doubt that even 1% of national buys use anything but Nielsen ratings in dealing with the sellers, however for quite a few years so-called engagement factors have been added to the ratings via indices as a supposed refinement. As for NBC's much touted allocation of $1 billion in upfront time to "advanced targeting" buys, this turned out to be a very much smaller operation than was assumed by some, mostly involving cable and probably was based on Nielsen ratings with some product user indexing added to the mix.
Despite the unwillingness of many advertisers to allow their brands more freedom of action in buying TV time, there is some progress in this area and also in the ways that the corporate spoils are doled out to individual brands via "allocation models". What's next? I forsee a time in the not so distant future where we have two upfronts followed by the quarterly scatter sales. The first upfront involving, perhaps, half of national TV spending, would be where the major CPM-driven buying as well as the major sports and big event sales are made. Later, a second upfront with about 25% of the dollars would be a brand by brand affair using appropriate metrics for each buy. Finally, about 25% would involve scatter time purchases and a large part of this could also be brand by brand.
With this new system, advertisers who opt for CPM efficiency and/or obtaining high visibility sponsorship positions in major TV events, spots, etc. can play this game---some allocating all of their dollars in this manner. However, others might invest much less in the corporate buying upfront but, in doing so, they could hedge their bets by garnering low CPMs for portions of their budgets, then play the more selective targeting card with each brand going its own way. As a result, brands that split their investments will use the first upfront to build an efficient CPM audience tonnage base, then complement it with better targeted impressions which may involve higher CPMs. Overall, this would result in major targeting improvements,without the risks of being totally at the mercy of the sellers in the brand by brand buying negotiations..
Ed, I really think that you're on to something with your idea for a two-step Upfront. I could absolutely see the industry going that way.
Dave, I think that there is a way forward on this---but it would take a few years to bear fruit. What if a number of advertisers----perhaps your clients as well as others---simply reduced their corporate up front buys by 10-20%. When the fourth quarter scatter sales begin---or in advance of them----these advertisers would notify the sellers that they want proposals for all four upcoming quarterly scatter periods---in effect, creating a second upfront for brand by brand buying. The cable channels are the most likely to cooperate, which is fine as they are far more selective targeting-wise, but certain types of broadcast network genres---early AM or late night, as well as schedules on Fox or The CW might also be considered. Once these initial buys are made the results would be promoted----just as the programmatic guys have been doing---via "opinion pieces", "conferences",etc. and this will make other advertisers sit up and take notice. Also, I think that there is a solution to my reservstions about big data set usage vs. viewing that could be added to the mix.
So smart, Dave. It is thinking like this, on your clients' behalf and not just based on "the way things are done," that is driving Simul to be so successful and such a leader in the business.
Thanks Bill. We've been fortuante to have a medium like television that many have underestimated and some very progressive partners like Hallmark willing to try new things. Lots more to do.