Commentary
The Three Pillars Of Programmatic
- by Justin Evans , October 31, 2014
Programmatic advertising consists of:
1. Aggregated inventory. To run programmatic advertising, diverse inventory is yoked together in small batches of impressions to form a single ad campaign. In television, this means the purchase of multiple networks’ inventory, from multiple day parts, on specific days of the week that a target audience is watching. No guarantee is given that any particular network will comprise any particular portion of the campaign. However there will be guarantees on total audience delivery, with daypart guidelines.
2. Audience targeting. The exact network or day part composition of the buy is not of paramount importance, because the campaign is primarily focused on reaching an audience target. While traditional TV buys are also audience-targeted (based on demos), the quality, type, combination, and prominence of the programming is of equal value to the audience composition. In programmatic TV, the audience is the key. There are nuances here:
a. While programmatic TV may seek to target the same age/sex combinations available in traditional TV, a programmatic plan may also be based on purchase data, lifestyle segments, even CRM files — any other data that can be overlaid with TV viewing information through a database match.
b. Programmatic plans are based on concentrations of audiences — not to be confused with “addressable,” one-to-one targeting. For example, for a CPG campaign, ABC Family / Daytime will be purchased because it shows an index of 186 for Moms, Weather Channel / Overnight because it indexes at 171, and so forth (illustrative data). This is done so that the final plan — harnessing the concentration of all these high-indexing fragments — delivers a final audience composition that is much higher than anything that can be bought through traditional means (and at a lower price). By contrast, one-to-one targeting, which finds a TV set or household that exactly corresponds to the target characteristic, should be defined as “addressable.”
3. Systematic delivery. Aggregating inventory and audience data are complex processes. To deliver campaigns on a repeated basis, using these complex processes, requires technology, or at least focused, specialized operations. We’re talking Adam Smith, not Don Draper. That said, the technology supporting programmatic TV is in its early stages compared to the highly developed exchanges, bidders, and private marketplaces of digital. While technology plays a part in delivering these results, many participants in the ecosystem are achieving an efficient process through people operations. As a result, “systematic,” rather than “technology-based,” is a more accurate characterization of the state of play. As with aggregated inventory and audience targeting, the goal of these “systems” is to create efficient delivery: plans created, orders placed, orders delivered, reporting supplied, in a fashion designed to be frictionless and repeatable.
How can you use this framework? At the very least, having the terms clarified can help you identify just how “programmatic” any given solution is, and not be misled by hasty or broad characterizations.
Naturally, others will want to add to, adjust, and clarify the definition I have offered (after taking up the theme from Mr. Welch).
One area that is still very inchoate is the audience delivery proof of perfomance. You write that there "will be guarantees on total audience delivery, with daypart guidelines." This is the dark pool that is holding back broader embrace of programmatic. What exactly is the "total audience delivery" that we will be guaranteed? Households certainly are not an acceptable currency. Age/gender is where we are now. Until we have a strong agreed-to system for measuring how the "audience delivery" actually reflects the audience we have targeted there will be challenges for clients to release too many dollars to this platform.
Suppose that "The Jerry Springer Show" indexes at only 78 versus 112 for a broadcast network primetime show like "NCIS" for AT&T's target group, but the cost per viewer "impression" against said target group is three times higher for "NCIS". Doesn't this mean that the AT&T "trading desk" will want to buy impressions attained----or promised -----by the Springer selling desk, rather than the computer managing "NCIS" sales?In other words, how does the system handle the immense CPM pricing variations between platforms, shows and dayparts which usually far transcend the relatively smaller targeting variations between them? Isn't AT&T going to be obliged to set impression goals by daypart and, for that matter, between broadcast and cable within dayparts, to account, not only for CPM differences but also for total plan reach attainment? And what about more engaging "premium" inventory? Will sports, news and broadcast network prime be ruled out because their costs per viewer are so high, relative to alternative ways to garner target group eyeballs, that AT&T will shun these high priced program forms? Or, will they simply be "legislated" into the programmatic buys, regardless of what the numbers say? Just asking.
Another thing to consider is that these deals seem to be centred around two things - audience delivery 000s, and CPM. That is, spots are bundled up to deliver the highest 000s at the lowest CPMs. Remember the onset of 'optimisers'? The goal of an 'optimised' campaign was one of two things - the highest reach for the available budget, or the lowest cost for the required reach. While optimisers are a sophisticated algorithm that rely on past performance (the audience 000s and the duplications patterns) utilising respondent level data, they were firmly grounded in good marketing and media practices. It seems to me that the above is based on buying tonnage at the lowest cost and walking away from principles such as reach levels at a required frequency. Cheers.
That's how it looks to me, as well, John. Also, how do the programmatic models handle the viewer "engagement" aspect that seems to be so important to advertisers. Is this, yet another index to be factored into the equation. If you apply one index after the other, after a while you are likely to find that they all cancel each other out----in effect, pulling in different directions. Finally, even if the sellers agree to supply rate cards for all of their "inventory"----which is most unlikely----what's to stop them from hiking their rates on whatever shows are in greater demand, thereby nullifying the supposed benefit of "more efficient targeting".
Engagement is the tricky one. First, what is 'engagement'. I think Erwin came the closest when he said something along the lines of ... it's round, it's gold and you slip it on her finger. Second is, are we measuring the 'engagement' of the content or the context. This is a crucial question. You can derive a rough proxy for engagement of the context (i.e. the programme) by looking at the relationship between the reach of the programme and the average minute audience of the programme, For example, if the average audience is 5m but It reached 10m, of those that viewed ANY of the programme the average person viewed just half - i.e. low engagement. But measuring the engagement of the content (i.e. the ad) is much more problematic. Truth be known, the broadcaster doesn't 'own' the ad - it is really the advertiser's problem. A good programme and not make a bad ad good! As one sagacious US network pundit said many years ago (paraphrasing here) ... we spend tens of millions of dollars making a programme that is watched by 20m people, you put one lousy ad on and in 30-seconds lose 5m of them, and then you want a discount ... go make better ads!!!
Oops ... typo. Fifth line from the bottom should read "A good programme can not...". Joe any chance you could bring in a screen preview before submitting?
Also, Joe, can the format be modified to allow for separate paragraphs?
*** Prays for paragraphs to come back ***