Just a quick post today, to steer readers of this page to AdExchanger for some thoughts on programmatic reserve.
Today’s AdExchanger roundup highlighted a great presentation by Bob Arnold of Kellogg Company. His talk was on brand marketing via programmatic buying. We obviously hear a lot about programmatic buying (primarily RTB) for DR campaigns, but this is a great reminder that we’ve still just scratched the surface of what programmatic buying can offer.
First of all, there’s much more that can be done with RTB to help brands increase efficiency. Given the size of Brand budgets, this is a big (huge) area for growth. Bob shares some great case examples from Kellogg’s experience:
1) Having the right measurement framework is critical. Bob’s not talking about CPA and conversion attribution here. He’s driving long-term changes in purchase behavior through an offline channel. So he divides metrics into short-term (brand safety, viewability, composition and frequency), mid-term (attitudinal lifts like awareness and purchase intent) and long-term (offline sales lift as measured by Marketing Mix Models). This is all stuff I’ve discussed before, but we in the tech community can’t hear it too often and it’s particularly great to hear it from a big marketer explicitly in the context of RTB. Nielsen for one is listening. They already had a big presence in the first and last buckets, and they recently doubled down on the middle with their Vizu acquisition. Vizu and Nielsen work great together, so this is a very smart deal.
2) Creative is fundamental. In Bob’s words: “Creative quality is paramount…the media plan simply amplifies the creative message. Studies from comScore and Dynamic Logic [show that] 50-80% of the value of a branding campaign comes from the creative…it’s imperative you get that right.” Great creative is altogether too rare online, but there are some great examples out there, like this one. This ad made me stop and think. And the more I thought, the more it took me away. I will remember it and it will make me more likely to buy Legos for my kids. Bob’s comments and creative like this are a reminder that while advertising – the activity – will be an increasingly technical pursuit for the foreseeable future, the best advertising – the thing – will continue to be driven by great creative that leverages this technology to connect people with products in a way that creates a lasting impression.
3) Doing it right drives great results. Bob claimed that programmatic buying for Kellogg doubled targeting accuracy while increasing viewability to >70% (higher than direct buys). It also cut eCPM (CPM adjusted by comp & viewability) by >50% from their starting reference point. The combination of these factors (assuming creative was comparable before and after) drove improvements in ROI (via MMM) by ~5X. Just a stunning impact.
These examples demonstrate the power of RTB, in the hands of a savvy marketer, to drive Brand results. Good news for all of us.
But I also want to point out that there’s more to programmatic buying than RTB. There’s a whole set of programmatic premium, automated reserve (or whatever else you want to call it) capabilities possible on which the ecosystem as a whole as has barely scratched the surface. A couple of the pioneers in this area include Brand.net and isocket, but I assure you there will be other, larger players down this path after them. These next-gen programmatic capabilities are a great complement to the current set of RTB capabilities and they are tailor-made for branding.
It’s exciting to see a large marketer smartly leveraging the available technology in a new way to drive such great results. It will be even more exciting to see what’s possible as the market begins to deliver more technology focused on this important class of use cases.
Great article by John Ebbert himself on Ad Exchanger today.
John draws very insightful parallels between the data ownership issues that are an important factor in the current trade dispute between American Airlines and several online travel agencies and a similar data ownership conflict brewing in the online ad space.
I should have referenced this example in my recent post on the latter topic. It’s a great example. I was aware of the American story, but didn’t make the connection.
Nice job, John!
A quick post to direct readers to today’s guest article for AdExchanger. It will be pushed to the broader AdExchanger audience tomorrow in John’s roundup, but I wanted our readers to have a “sneak peek” to get the dialog started.
As always, I am very interested in your thoughts and comments.
I agree with those who indicated it is way too early to lock down a narrow definition of DSP. Arguably anything that’s really a “platform” should never need a description that’s as detailed as the list offered in the first post, but regardless its definitely too soon in this particular market.
At this stage I think all are better served by a more general definition. Fundamentally, I think any entity that meets the following criteria with sufficient breadth of capabilities is a DSP:
- Technology that interfaces directly with demand-side entities
- “Interface” does not necessarily mean GUI. An API could be even more useful if it meets the customer requirements
- Demand-side entities may include agencies and/or advertisers
- Technology that adds significant value in the process of buying and/or management of media
- Value could originate from data integration, forecasting, buy automation or other operational efficiency gains, supply source integration, delivery and/or pricing risk management, increased ad effectiveness through optimization, impression filtering/categorization
- Etc., etc., etc…
- Technology that operates as directed by the demand side entity (i.e., the customer)
- The technology can be used flexibly and transparently by the customer in a way that benefits its business, with limited incentive conflicts
Obviously, technology is the common thread; DSPs will compete on the strength of their technology and networks with weak technology (essentially bucket shops, substituting people and excel for real technology) will find themselves increasingly squeezed between DSPs and exchanges.
One final point: “platform” implies broad capabilities. Many companies exist with valuable capabilities that meet the above criteria, but that couldn’t properly be called platforms. I would suggest that Demand Side “Tools” (DSTs?) is probably more appropriate for more narrow capabilities. These tools may be used directly by demand side entities and/or be packaged by DSPs.
This post is a continuation of my article in last Monday’s AdExchanger about some serious challenges with BT for Brand marketers. Interested readers should start there and then continue reading below, as I make some of my points here in the context of the example presented in that original article.
As I mentioned, BT does not outperform other approaches in driving offline sales. Specifically, Brand.net’s studies with Nielsen have proven that our campaigns deliver impressive offline sales impact. These results were achieved without BT; instead Brand.net uses high-quality media with contextual, demographic and geographic targeting managed to high composition, with controlled frequency and cost.
The average ROI of 141% on these Brand.net campaigns is roughly comparable to the average ROI generated by Nielsen’s largest offline measurement partners over hundreds of studies using the purchase-based / look-alike targeting approach I described in my original article, refined over nearly a decade. The Nielsen-powered BT those others use is state of the art; BT doesn’t get any better for branding. If it fails to deliver substantial ROI upside to other approaches in driving offline sales – we as brand marketers really need to question the utility of BT in general.
In addition to this fundamental problem, BT poses a variety of other important problems that brand marketers should consider carefully.
First, there are no standard definitions within the industry for behavioral categories so there’s a huge degree of subjectivity in defining which users are a close-enough match to the core users to qualify as “look-alikes.” This is a big deal because, as I outlined, 99.9% of the users in a typical BT campaign are based on look-alike modeling. In the context of the specific example I used, how similar does a user need to be to an actual CPB Baker to qualify for inclusion the behavioral category? What’s to keep the network doing the modeling from stretching that definition to create more inventory, particularly if there’s no direct measurement on the campaign?
Another related issue is lack of portability. Since there’s no consistent definition for any behavioral target, if an advertiser does find something that works with a particular vendor, the advertiser is stuck with that vendor. They can’t say, “CPB Bakers work great. Let’s figure out the best way to buy them.” because the CPB Bakers from one source could be completely different from the CPB Bakers from another source due to different look-alike definitions. Furthermore, if the vendor whose CPB Bakers “worked” changes look-alike definitions, loses access to data or goes out of business, the advertiser must start from scratch. BT can’t be used as a basis for a scalable, repeatable, progressively improved strategy driven by the advertiser/agency unless the advertiser is the one building the profiles from scratch – something that is far beyond what most advertisers today are willing to do.
Due to cookie churn and simple inventory volatility, impression delivery is extremely hard to predict for any reasonably focused BT target (and forget about reach or pricing). This makes forward delivery guarantees almost impossible – another barrier for scalable use by large brands that typically plan a significant portion of their spend in advance.
BT can also be used by networks or publishers as a way to mask inventory quality issues. Would an advertiser/agency want the media included in a BT buy if they actually knew what they were purchasing? Would they be willing to pay the same rate? I doubt it, but the glossy BT story effectively launders this sketchy inventory into a desirable commodity.
Finally, there are obviously high-profile privacy issues swirling around BT, and it’s anyone’s guess where those will settle out. I would hate to have a platform or media strategy built around BT if (when?) our friends in Washington decide that “opt-in” will become the law of the land.
Marketers considering significant or sustained investments in BT would be advised to think carefully about all of these issues and ask tough questions of their partners before proceeding.
A blockbuster report from the OPA late last week, at least if one were to judge by how it lit up the blogosphere (as AdExchanger humorously put it, “Is the OPA the greatest link baiting organization in advertising, or what?”). I reviewed some of the coverage and the report itself over the weekend and I have to say, with all due respect to the OPA and its members, this report doesn’t measure up to their previous efforts.
Here’s my take:
1) Most networks are focused on DR metrics and not the upper-funnel branding metrics that are the focus of the OPA study. So even if we stop right there, it’s not shocking that that the study shows weaker results for networks. This difference in focus is fundamental to Brand.net’s business by the way. Unlike other networks, the Brand.net platform offers a full suite of capabilities designed from the ground up to help brand marketers leverage the web to reach their audience efficiently and effectively drive these upper-funnel metrics.
2) The OPA report didn’t include or consider cost data. If you believe the >10:1 spread between publishers’ direct and network deals cited in last year’s IAB research, this is a critical omission. OPA pubs performing 50% better than networks doesn’t look so good in the context of a >10:1 price ratio. Obviously the devil’s in the details here – the IAB research isn’t perfect either for reasons I have discussed previously on this page – but it’s clearly perilous to draw the sweeping conclusions OPA is going for without considering costs.
3) I don’t wish to cast aspersions on the study or methodology overall, but a couple of the data points just seemed counterintuitive to me. For example, slide 19 of the OPA results deck states that ad networks deliver insignificant improvements in purchase intent for the financial services category. This particular point caught my eye, because I know that well over $1B has moved through ad networks from hundreds of financial services companies over the past 5 years, the vast majority of which has been measured on a CPA – as in actual purchases, not just purchase intent. It’s extremely hard for me to believe this money would have continued to flow in such volume over such a long time period if it wasn’t actually driving purchases. If you agree, then we’re left with only 2 possible explanations: a) the data referenced to make this point is somehow not representative or b) purchase intent as measured by DL was not correlated with actual purchases. Neither is particularly comforting.
4) In addition to the metrics OPA focuses on in this report, I would have liked to see an analysis of actual sales lift – i.e., the ultimate result that improvement in the attitudinal metrics discussed in the report is intended to drive over the long term. This certainly isn’t easy for every client on every campaign, but it’s a powerful capability that proves real business results for many. For the next study I would be interested in seeing similar data from OPA.
Some of these thoughts have already been expressed by others, including some who commented directly on WSJ’s coverage of the report, but I thought there was enough new here that it was worth joining the discussion.
Let me know what you think.