Supply-side platforms and demand-side platforms often make tweaks to auction logic to try to get better results for their clients, with many SSPs moving to first price auctions, and demand-side platforms implementing supply path optimisation. However Index Exchange’s recent bid caching scandal highlighted how clients aren’t necessarily briefed on these tweaks, and the effects they have on programmatic trading. In this Q&A Shane Shevlin, IPONWEB’s SVP of strategic development, argues that it’s vital that publishers and advertisers understand these auction dynamics, and that SSPs and DSPs must be transparent about tweaks they make to auction logic.
Shevlin will be on stage at New Video Frontiers next Wednesday at 2.10pm on a panel titled ‘Sell-Side Strategies: What’s Working?’ Tickets are available via the website.
What’s your view on the current state of in-housing among brands?
While there is clear appetite amongst certain brands who spend big on programmatic to understand better where their spend is going, the number of brands who are in-housing the totality of their programmatic operations is still lower than some trade press might suggest. Yet there is a discernible trend of more brands having serious conversations with programmatic technology vendors.
I witnessed a similar trend roughly a decade ago, when the likes of Sky and Expedia “in-housed” their ad serving. For Sky that meant taking control of their vendor contracts and more scrutiny on the work their agency was doing on their behalf. For Expedia in-housing meant much more because digital marketing had a direct effect on the vast majority of transactional revenue on their web properties. Expedia were/are a digital-first business with the vast majority of their revenue coming via their web properties; fully integrated, in-housed digital marketing operations makes complete sense, whereas Sky see better to deploy their in-house best and brightest on other areas of the business, like Adsmart; the service their agency delivers against their own marketing goals still measures up.
I think a similar split exists today but as more and more business see their revenue shift into digitally-connected areas, I think we’ll see more brands beginning to experiment with programmatic. Where they will continue to need help, however, is understanding the dearth of data that in-housing throws up – how do I normalise, what does it actually mean, how do I benchmark? For this I think many will come back to their agency – for uncovering true insight and really helping to understand data in the right context. For direct to consumer brands like Netflix, this is less the case, since they have lived and breathed this data from day zero.
DSPs and SSPs are making lots of tweaks to auction logic to try to get better results for their clients. How important is it that the publishers and advertisers understand these tweaks themselves? What kind of uplift do they typically see?
I guess you’re referring to the move from 2nd price to 1st price auction model within SSPs? And the practice of SPO on the demand side? In the main, I see these as natural adjustments to a market that was to begin with massively over-supplied but also dominated by one player. I think we can also throw into this discussion things like viewability and fraud detection that now exist under the umbrella of quality control – all necessary innovations to ensure the long-term viability of automated online display advertising.
As regards how important it is for publishers and advertisers that they understand these tweaks, why would it be less important for them than for anybody else in the supply chain? If anything it’s more important for them. Auction tweaking is absolutely fine, as long as what you’re doing is transparent. We operate in a free market and those who understand the mechanics of that market better will and should thrive (just like any other market).
Where the risk emerges is when we have actors who are not telling the market what they are doing. This behaviour scares advertiser spend away from the programmatic channel and encourages dominant walled gardens to take advantage of this potentially crippling weakness of markets that are supposed to be based on open trading and measurement standards.
The devastating thing for many independent ad tech business owners in our space is that their best customers will run straight into the arms of closed ecosystems simply because they do a better job of PR than we do around both their own and our own mistakes.
On your question of uplift, the answer varies massively depending on the set up. In the early days of the mass migration to 1st price auction, there was relative carnage for a while. Big budgets happily bidding under what they thought was the safety of ‘2nd price wins’ were however quickly adapted – but in some cases using crude SPO methods. The reality is that there is so much duplication and division of spoils across the header, tag and RTB, that you really need an ML solution that optimises at the atomic level – and one that can do a very good job of pacing & delivery simulation so that your SPO efforts don’t knock campaign delivery out of kilter. In some cases we have seen up to 30% swings possible – but it really depends on budget size, campaign parameters etc.
IPONWEB frequently talks about the need for transparency in programmatic trading. Some are sceptical about the extent to which many of the key players — agencies and ad tech — really do want transparency. Is that scepticism justified?
I think that one reasons there is scepticism is because there is a disconnect between how each part defines transparency in the digital ecosystem. Is it an end-to-end view of all cost transactions? I think you’ll find that some brands actually don’t rate that very high on their list of priorities if the outcomes they ask for are achieved – in the direct response segment, for example.
In general, we as an industry do need to take a common sense and ethical approach to transparency; we should provide it when asked but sometimes total transparency simply serves to create unnecessary tension between vendors/agency/client because of the complexities of the programmatic supply chain; for me it boils down to better contract scrutiny for principals and a mutual awareness between vendors and their clients of the value that is supposed to be delivered. Historically, many actors who leverage the programmatic ecosystem have neglected to focus on what the final value they are delivering their customer is; this is where things begin to break down.
A recent IPONWEB/ExchangeWire survey found that almost half of media agencies use proprietary ad tech, and 34 percent use a mix of custom and third-party software. Is it conceivable that agencies will own their own full-stack or will they always require third party solutions?
To guarantee the continuity of existing client relationships, 3rd party solutions will continue to be used by many agencies. Vendor contract ownership by brands suggests this will continue to be the case. However, when asked a question like this, I can’t help think back to that wonderful interview that Gabe Lydon, CEO Machinezone, gave at the Recode conference in New York a few years back. I don’t believe anything has changed. His world view if anything is more pertinent today, with the relentless march of Amazon into advertising, for example. I think the smarter agencies will look at this and think that it’s now or never for them to take control of their own and their clients’ success. That means control of data, which ultimately means control of your own technology.
I’m not talking about contractual data ownership, which is now pretty standard among many ad tech vendors (except SSPs, who need to catch up here), but rather the ability to test and burn through ideas with customer data at the atomic level – not just swapping out segments in line item targeting! I’m talking about powerful flexibility in automation that can unlock the sweet spot of success for a specific set of desired business outcomes – and it’s only with proprietary tech that the most sophisticated strategies can be executed (custom predictive modelling, truly custom algo logic that affects bidding and ad selection, custom ‘short circuit’ logic for better user id management etc.). These techniques are becoming more essential for agencies as they grapple with the increasingly complex task of trying to marry up customer journeys and leverage their data from marketing stack into relevant browser, connected TV, phone or any other personal device experience.