Real-Time Bidding (RTB) is an integral aspect of digital programmatic, so it’s only natural to think about applying auction-based approaches to Programmatic TV.
But the rationale of “TV should work like online video” completely ignores the structural differences between the two media. It ignores the concerns that many premium online video publishers have with buyer bidding.
Auctions can provide definite benefits in appropriate circumstances, but TV media owners need to clearly understand these circumstances if they are to avoid outcomes detrimental to their businesses and the TV industry at large.
Auction approaches, in which ad opportunities are offered up to buyers one or a few at a time, all use decisioning algorithms that mathematicians refer to as greedy. They assume that making a locally optimum decision on each opportunity will yield the best total solution (or at least one that is close enough to the best).
For a media owner, that translates to “if I get as high a price as I can for each spot, I’ll end up with as much total revenue as I can.” That only works when one decision doesn’t affect any other, such as when there is little to no perceived contention for a given opportunity, or when the opportunities are clearly known to be exceptional and rare (more on this later).
That’s not how TV, or any truly premium content, really works.
There’s generally not enough inventory to satisfy all buyers; the media owner has to decide which orders to take and how to best fulfill them. In these cases, the optimal solution to the whole problem (assigning orders to inventory) can’t be achieved by short-sighted aggregating optimal solutions to the individual pieces of the problem: which ad to run in a given avail.
The optimal solution can only be found by considering the big picture.
Here’s an example: you have a shoe box containing 37 baseball cards and a bunch of potential buyers. You start pulling out one card at a time and asking the buyers to bid. Let’s say the first card (Andy Merchant) goes for $2, the second (Steve Dillard) for $1, and by the time you’re done, you’ve sold 25 of the cards for a total of $45.
And that’s when one of the buyers says: “You had a complete matching set of the 1975 Red Sox? I would’ve given you $500 for the whole set!”
The last example highlights another reason media owners should be very selective when consigning inventory to auctions – information asymmetry. You didn’t know that buyers would’ve paid you more for a complete set. A similar problem often occurs in auction-based markets: the media owner has no idea why a buyer bought an opportunity and thus no way to understand the value of the rest of his inventory.
Looking at it from the other side, buyers love bidding because they can cherry-pick — which is why you’ll see people with buy-side backgrounds pushing such approaches.
That’s not to say that all auctions are bad.
Currently, there exists both auction-based and order-based marketplaces on behalf of TV media owners. Auctions tend to generate upward price pressure when a) the ad opportunities are exceptionally differentiated because of program (such as the Super Bowl or directly within a highly-rated program) or the ability to leverage specific context (such as within a news program), and b) the auction is price-transparent (all the bidders have an understanding of the current highest bid).
The moral of the story is that TV media owners should be highly selective in the inventory they put up for bid and they should do so in a manner that more resembles an art auction than a flea market.
Some very good points, Joel. Nice piece.
The need to bundle is also a reason why "premium" content cannot be sold on a cherry picking basis as most programmatic trading desks will want to buy time in the same shows---probably about a third of them--- leaving the remainder, unsold. No network, cable channel, station or syndicator can survive if this is allowed to happen.
Some may think that the alleged precision and "granularity" of "big data"/third party data targeting will allow sellers to move everything they've got, because different advertisers will target variable sets of consumers. Unfortunately, this is not the case----especially when household data is used. If you look at the huge amount of data that has long been available--MRI, Simmons, etc.---you will see that affluent homes with household heads in the 24-54 age group and one or more youngsters index above average for the majority of products and services. Any computerized buying system that can not deal with reach but only uses targeted consumer CPMs as its main "currency" will want only those shows that index high on the demo profiles mentioned above----CPMs held constant. That isn't going to work for most advertisers due to reach issues as well as engagement, merchandising and other factors. Nor will it work for sellers, for obvious reasons.
Very good points here.
Ed's point about packaging inventory is exactly right. Planners have always written buying guidelines that identified the most desirable programming using product usage, psychographic and other qualifiers using research (sorry, data) from sources such as MRI & SMRB. As Ed correctly points out, no TV network will sell a package of only the most desireable inventory, so the buyer always winds up with a mix, some of which they may not really want at all. This is further complicated when buying on a corporate basis for a company with multiple brands/products that may have entirely different consumers. I do think that there is a role for programmatic buying in TV, I'm just not sure anyone has quite figured it out yet.
This is the article I've been wanting to read about this subject! Thanks!
Thanks, Ed, Neil, and Kathy. Bundling is a manual form of inventory optimization, which is a beneficial aspect of programmatic selling. I'll address this in a subsequent piece - stay tuned.
Everyone here missed a major point- the concept of serving relevant ads on a technology platform. Sorry Ed.
P.S. Lack of transparency is not a friend but it is for the buyer or advertiser. Let's get some dialogue on this.