Why Publishers Should Be Wary of Optimizing Their Direct and RTB Demand Sources Together

esco-sell-siderThe Sell-Sider” is a column written by the sell-side of the digital media community.

Today’s column is written by Esco Strong, Director, Marketplace Strategy at Microsoft.

If you’re a publisher managing a sizable direct-sold, guaranteed-delivery display ads business, you’ve almost certainly been tempted by the prospect of squeezing a few more dollars out of your business by optimizing your guaranteed and non-guaranteed demand sources as a single, unified supply pool.

The pitch is relatively straightforward: rather than having reserved campaigns sit “first chair” and gobble up all of the impressions they want ahead of downstream demand, instead allow “discretionary demand” to pick off some of those impressions where they are willing to pay a premium above and beyond the reserved rates. One key assumption here is that the “cherry-picking” can be done in a controlled manner such that enough inventory will be left over to fulfill the volume guarantees of the reserved buys.

While I have little doubt that technology in our industry today can tidily solve for the problem of meeting the delivery guarantee in this situation, our focus really should be on the question of the long-term effects on the brand businesses and the quality of the inventory that they will receive in this scenario. One challenge is the lack of “expressivity” – or targeting parameters – of brand buys.

Here’s what I mean: since these buys are typically non-targeted (i.e. non-‘expressive’) or feature basic reach targets (DMA, profile, etc.), it is difficult to infer what the success metrics are that are being applied to the buys, and therefore, what the proxy metrics are for the purchase of these chunks of inventory. Add the fact that brand goals can sometimes be diverse and divergent, and it becomes a very difficult exercise to understand how to evaluate ‘quality’ in terms of a potentially shifting inventory mix post-optimization.

For example, some brand buyers purchase these campaigns as a proxy for a demographic distribution of users that is representative of the population at large. Others may buy these groups of reserved impressions as a proxy for clicks, and may find them to be more effective at generating clicks in general or from a more diverse set of users than they may otherwise achieve via CPC-model buys where targeting will be used.

Another class of brand buyers are those seeking an implied concentration of a particular type of user or demographic segment, particularly in brand buys that are contextually targeted to specific types of content. Context also plays an important role in defining state of mind, and may be useful in helping marketers reach users at a particular point in the purchase cycle, such as awareness or consideration.

Ultimately, two basic components – the inventory mix, and the reservation against it – provide the necessary materials that create the value proposition for these buys, and while the reservation is quantifiable to the seller, the expected inventory mix is oftentimes neither apparent nor something that can even be inferred from other sources of information.

By contrast, most “spot market” demand is incredibly expressive (targeted!) today, by virtue of the capabilities afforded by real-time bidding (RTB) and cookie syncing. Impressions are targeted against a host of different parameters that are often combined to specify a highly granular type of user, behavior, user state, or all of the above. These buys are telling their buy-side platforms exactly what they’re buying and what they’re willing to pay in order to skim the cream off of remnant supply pools and get a hold of just the stuff they’re looking for. Interestingly, while the bid price is made apparent to publishers, the targeting is not, which poses an additional challenge for someone trying to manage optimization across these demand channels and predict the effects that RTB buying would have on reserved inventory mixes.

My colleague Mark Hall developed what I consider the best analogy for this situation, which is to equate the inventory to a “fruit salad” example. Brand buyers are buying a healthy mix of inventory – a fruit salad of users and impressions – that sometimes varies by the specific buy (think “berry fruit salad”, “tropical fruit salad”, etc.), but always contains an implied mix of certain types of fruit.

Similarly, we discussed how brand buys aren’t highly expressive with their goals – some of which may vary between buyers – and the analog here is that the fruit salad appeals to different types of customers as well. Some buy for the mix of overall fruit, some for the blend of tart vs. sweet, some for an implied quantity of certain types of fruit they like (despite others that they are less interested in or may even outright dislike), etc.

Spot buyers, on the other hand, are the proverbial “cherry pickers” of this example – they are specifying only the fruits they want, and buying them up at higher prices than the average price-per-fruit of the fruit salad buyers. However, an interesting thing happens along the way: as more and more of the “cherries” are plucked out by the spot buyers, the fruits that are left over begin to change the composition of the remaining fruit salad. A balanced mix of fruits suddenly becomes very acidic when melons, for instance, are removed from the mix and the tartness of strawberries and pineapples remains.

It is important to recall at this point that it is very difficult to surmise the composition of remaining fruit salad that would be acceptable to brand buyers, since they have likely not expressed it as part of their buy. Given this challenge, it seems incredibly risky to expose the “fruit salad” of brand buys to the cherry-picking of spot demand in any voluminous way, as the long-term implications are possibly severe. The typical response here is to point out that there are protections in place whereby spot demand can only pick off impressions where it is able to outbid reserved demand, which typically have relatively high CPMs. However, if brand buyers were to eventually shy away from placing their reservations against some of these campaigns due to the shifting mix, it would undermine this safety net as more and more inventory would begin to go towards spot buyers, who would then be able to lower their bids and continue to win as the brand buys (and their CPM benchmarks) evaporate. Therefore, it seems central to publishers’ interests to protect the constitution of their brand buys as the cornerstone of their revenue pyramid, even if there are opportunities to squeeze a little bit more revenue out of their inventory.

One last question that begs asking is whether the adverse selection described earlier isn’t already occurring amongst just the reserved campaigns – irrespective of any spot market considerations – and therefore possibly isn’t as big a concern as we might think. While it is true that the potential exists for inventory mix shifts amongst reserved buys when some are more expressive than others – behaviorally targeted audience segments, for example – this is typically done within the controlled environment of direct sales and the publisher’s ad server, and is gated by considerations such as sales force activity, natural and self-imposed limits on the availability of this type of inventory for sale, and analysis behind the scenes that a publisher may be running against these buys. The primary difference here is that the publisher remains in control of the buys and is completely apprised of both the target and the pricing. This is in stark contrast to RTB buys where the publisher knows only the price the buyer is willing to pay – a moving target that changes impression-by-impression – and is unaware of the targeting behind that bid.

One thing that I love about this industry is that where there is opportunity, it seldom takes long for new products to fill the void. And in this case, a revenue opportunity exists that many publishers are in desperate need of and will find too irresistible to pass up. The plumbing is there in several quarters to turn the spigot and let the short-term dollars come in, but I am reminded of the words spoken by author Eric Ries of “The Lean Startup” fame at the recent AdExchanger Human Centered Automation conference, where he stated that our technology is now at a point where we can do and build just about anything we want, but the question really is whether we should.

In this case, the question of whether publishers should be comfortable making large changes to the composition of their brand campaign inventories without first truly understanding those customers’ expectations seems imperative to answering Mr. Ries’ question.

Follow Microsoft Advertising (@MSAdvertising) AdExchanger (@adexchanger) on Twitter.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Good piece Esco. I agree that publishers need to balance both short and long-term revenue needs. It may be possible to squeeze a few more bucks out of areas with decent demand, but to your point, it may changes the audience mix significantly enough that the direct buyer isn’t really receiving the audience they thought they were. It’s akin to two buyers buying the same section and one buying US-only, the other buyer is going to now get a higher percentage of non-US traffic which may have a real impact on how effective the buy was. In theory it’s all a question of capturing the right price for each customer segment but in reality, it’s as much about relationships and expectations as it is about performance, certainly in the direct world. My guess is that there will be few publishers who take a full all or nothing stance. Smart publishers will ID areas they want to protect zealously to nurture and build on strong buyer/audience affinity and other areas where the risks are low if there are impacts to direct campaigns. With revenue pressures growing for publishers and the sea-change in buying, it will be hard to sit out the programmatic dance entirely.

  2. Alejandro Correa

    Great article. I’ve used the same argument you make to try to convince the marketers that commission “non-expressive” campaigns to run their branding buys through DSPs, precisely so they can measure the composition of their “fruit salad” as you call it.

  3. Ecso, this is a well thought-out piece, and raises some fair points I think. Publishers have to move purposefully into this space, be informed, and think of the consequences of messing with the inventory mix they deliver to buyers. I have to say though that I’m unconvinced there’s anything particular about holistic ad serving which delivers a fundamentally lower quality experience for direct buyers.

    As you smartly mention in the bottom part of your article, the inventory mix any client might receive in a direct buy is already adversely affected by a number of variables today. Ad server pacing, competing demand in the channel, capacity that shows up to the page on any given day, the page quality distributed within a given targeting set, and etc. are all components publishers can influence that would impact the advertiser’s experience. Not only that, but I would say for most publishers, there is a small % of buyers that are paying rates on direct deals that would overlap with the rates paid in the exchange world. To be sure, this overlap is growing month by month, but the absolute amount of inventory impacted isn’t significant – at least not yet!

  4. Michael Lum

    If the same data is made available to both direct and RTB (for example, in an ad server that does direct selling and RTB selling in a single ad decision), then it’s already possible to meet both brand and performance goals. The ad server knows how much fruit is in the salad, the RTB buyers don’t.


  5. Jon Mansell

    Definitely a thoughtful piece and a solid analogy. The problem I’m having is the assumption that the RTB pipes aren’t being used for brand dollars and only a tool for “spot buys”. That may be the general case in the past, but that’s certainly not where we are headed. If we are only using DSPs to cherry pick the users we want, why then did PMP market start to pick up steam in the last couple quarters? The lines blur there. PMP controls were built so that a more controlled marketplace could be executed. One of the possibilities created by these controls is potentially removing the algorithm altogether (algorithm = “cherry picking”) from the equation. As a buy side guy, I have to say that publishers often over think what is going on over here on our side of the fence. Its not that complicated in many circumstances. Often, we’ve simply found a more efficient way to conduct a direct buy, no paper work, no tags, no discrepancies, etc. With all of that said, publishers do need to continue to ask these questions and continue to be diligent to understand the “why”. But I’d propose that RTB is an archaic acronym now-a-days as not everything therein needs to be bid on to win with the advent of sell side prioritization.

  6. Esco,
    This is indeed a thought-provoking piece. I wonder if any buyers have run into this problem, or it is viewed mostly as a potential problem?

  7. Fascinating…as an advertiser-side analytics person, I agree with Jon. The distinction between the spray-and-pray and data-driven individual cookie-level audience composition is lost on most brand advertiser’s and their agencies. Many are busily still trying to shoehorn GRPs into digital advertising and remarkably still obsessing about clickthrough rates (6 years after Natural Born Clickers).

  8. Good article, and a valid theory/ concern for publishers. As you point out in your intro, however, there’s also a good possibility that if done properly, the unbundling of the fruits improves yield, as some brands will pay a premium for the strawberries, while others will pay a premium for melon, etc — you might even be doing the downstream/ guaranteed brands a favor by removing melons and leaving them more cherries. By enabling better targeting, you’re unlocking value for advertisers (and users), so there should be upside for publishers as well.