November 6, 2019

Why SPO Isn’t as Mysterious as It Sounds

You’ve likely been hearing a lot about SPO lately.  In this article I’d like to lay out some clear and simple steps marketers, agencies and trading desks can take on the road to SPO.  While you can do a lot of sophisticated SPO work, as with most things in life, you can get 80% of the results with 20% of the effort and focus on that last 20% later.  These simple steps are pulled from my experience in SPO tech work as a primary author on numerous relevant specs like app-ads.txt and secure bid requests (ads.cert) as well as a member of the IAB Tech Lab’s Commit Group overseeing all specs.  

If you’re confused about why SPO is the topic du jour lately, these two articles on what led us to SPO and on the latest developments enabling SPO will give you some background.

First, what is the goal of SPO?  In a word: de-averaging.  Today as a buyer you are purchasing a mix of inventory from various sources and you see the average performance.  You are likely already accustomed to de-averaging on things like domain, device, or geo to find and optimize to top performance.  You’re likely not de-averaging on things like exchange or reseller path.  SPO is simply about including the supply path in the list of vectors on which you optimize.

So, what is actually happening on the ground today?  Mostly, I’ve seen that buyers have begun exchange audits and whitelisting of exchanges.  Whitelisting exchanges that implement business policies you agree with does make sense, but I wouldn’t suggest taking it too far - for example deciding on one exchange per format is likely missing significant inventory and denying you the opportunity to experiment and learn across exchanges.  If you haven’t had conversations with your exchanges yet about transparency, take rates, what they do to create value, the directness of their inventory, etc I’d suggest you start with that and whittle your list of exchanges down to a handful for each format.  That gives you a better handle on things and you’re not likely missing out on much if any inventory.

Moving Past Exchange Short-Lists

The following are some recommended practices as you begin to think more about SPO

  • Pull an exchange-level report - most buyers have never looked at performance by exchange - pull up that data in your DSP and see what is happening in terms of relative performance on a cost-per-action and whether your budget is following the performance.
  • Exchange-level path optimization - once you have a handle on relative exchange performance, blacklist bad performers and/or consider allocating budget specifically to higher-performing exchanges.  Pulling this data varies by DSP - here’s how to do it in The Trade Desk, DV360, and Xandr.  This is easy and can be VERY effective - this is most likely to be that 80% solution you can implement NOW.
  • Look beyond the exchange - most exchanges buy from a variety of inventory sources, and this data is now available via sellers.json and the supply chain object.  The tools aren’t necessarily easy for leveraging this data yet, but innovative buyers are looking at it, and some companies are building tools to help, chiefly Jounce Media, which also provides some really interesting public data on their site.  For now, this data isn’t easily actionable, but you can build out custom bidders to utilize or use this data to inform which exchanges you buy on or ask your exchanges to create deals for you.
  • Don’t blindly cut all resellers - While it’s tempting to eliminate buying resold inventory, reselling comes in many flavors.  In some cases, particularly for midsize publishers, they may work with publisher networks who handle their monetization.  In this case the only way to buy that inventory is via reselling.  More sophisticated analysis, again such as that done by Jounce Media and others, looks at whether a path is “the most direct possible”.  Even this may not lead a buyer to the best outcome as a particular reseller may enhance the transaction in such a way as to make it worth the extra hop.  It’s best to look at performance by path and not worry so much about reselling or not.
  • Consider unique ad formats - Don’t forget about the creative and the user - the actual creative shown on the page at the end of the day can vary dramatically by path.  In the sea of data it’s easy to forget about this - although it should ultimately show up in the performance data.
  • Run small tests - You may find you don’t necessarily have the data on performance by path.  To gather that, consider broad always-on campaigns to gather data - you could run a small budget always-on campaign with open targeting and use that to learn about the performance of various paths.  This can be a useful input to your strategy.  Your DSP partner may also be willing to share some level of this data with you.

While those are a lot of ideas, I want to again emphasize that you’ll likely get some big benefits from simply looking at cost-per performance by exchange and allocating spend accordingly.  Ultimately being perfect on SPO isn’t necessary to get some benefits from it.

So What's Next?

First, DSPs are getting in the game.  The initiatives helping with transparency - ads.txt, sellers.json, and supply chain object were all pioneered by DSPs.  As I write, The Trade Desk is stepping up their already aggressive push for transparency via the supply chain object.  The DSPs are taking two strategies: 1) take unilateral action to completely block buyers from purchasing certain bad or highly unknown paths, and 2) giving buyers information and control, for example through The Trade Desk’s Gold Standard and MediaMath’s SOURCE initiative.  So far, Google seems to be playing catchup and has yet to even publish a sellers.json file, but this will undoubtedly change, particularly given they were a pioneer in forcing ads.txt adoption two years ago.  

The more prominence these initiatives gain, the more they will benefit the entire ecosystem, even those slow in adoption.  The most complex, inefficient, or fraudulent supply paths are likely to refine their operations or go out of business, inadvertently cleaning up the paths for everyone.  

So, your DSP may help you, but ultimately buyers need to take the reins.  The ideas in this article are a good place to start, and several consulting companies such as Jounce mentioned above and other entities can help buyers get more sophisticated.  Perhaps new types of companies will spring up as layers on top of buying platforms to help provide guidance or custom bidding algorithms.  In an example from a completely different industry, a website and browser plugin called AlgoTransparency attempts to clarify to the user what is really happening when they use common sites with recommendation algorithms - for example helping users understand they are being recommended increasingly extreme YouTube videos because YouTube’s algorithm has optimized for watch time.  Perhaps these sorts of tools will emerge for ad buyers.  Imagine for example based on the boxes you check in your DSP the software alerting you “Did you know the options you’ve selected will mostly target annoying popups on a reseller ad network that usually create negative brand perception?”  

While this is a lot of ad tech inside baseball, I want to conclude by re-grounding in the fundamental “Why?”  To me, there are two core drivers.  One, as an advertiser you want the best possible performance at the lowest cost from your ads.  Wasting money on inefficient or fraudulent supply paths won’t accomplish that, and won’t let you offer the best product at the best price to the most customers.  Without SPO, as a buyer you are likely purchasing a mix of great and terrible performance, and just seeing the averaged result.  Second, we all have a vested interest in the ad-funded internet continuing to thrive.  By simplifying supply chains, more money will tend to flow to real publishers and allow the open and free internet to continue to thrive.