Looking at publisher setups lately, I’ve got a distinct feeling of déjà vu – complex waterfall setups with indirect monetization partners are everywhere. What is this, 2009? Fortunately no, but the challenge of publisher waterfalls is still very much alive despite the advancements in RTB, and just as this issue put a cap on publisher revenues with networks all those years ago, a complex waterfall can limit publishers today from scaling their programmatic channel to its maximum potential.
The solution? Back in 2009 it was to amalgamate all those network relationships into an SSP, but given the fragmentation is now between SSPs, publishers in 2015 should look to migrate their implementations toward pre-bid integrations, also commonly referred to as header-bidding, advance bidding or tagless setups. This technology has actually been around for years, but hasn’t been widely adopted in the publisher community until the last 12 – 18 months. From my own personal perspective working with over 50 premium, global publishers at Yieldex, I’ve seen pre-bid as a transformative solution, creating major operational efficiencies as well as increased revenue.
The reasons for walking away from the waterfall approach are many and compelling, but it comes down to ad serving costs on passbacks, inventory loss due to platform-to-platform latency, and most importantly, low-fidelity yield management. Most ad operations people know what those first two are, but that last one probably needs a bit more explanation. What I mean there is that when publishers implement ad tags for a given programmatic partner today, they typically put in that platform’s average eCPM yield for the last week or month, or some recent period. This means from the ad server’s perspective (which is doing the yield management don’t forget), the value is only known in aggregate across millions of impressions and not on an impression-by-impression basis.
Since the average rate then controls when the ad server decides to serve that ad tag, though, what any platform will pay for a specific impression has no impact on if it actually wins that supply. Rather, the platform’s rank in the waterfall is what really decides who wins the impression, irrespective of value. Add in the fact that for publishers on DFP, AdX can bid impression by impression which may lend advantage to AdX that could be hard to beat by other programmatic partners.
How does a pre-bid setup solve all these awful publisher pain points you ask? Pre-bid provides a solution because it separates the valuation process from the ad serving process with programmatic platforms. Importantly, it moves the valuation part ahead of the call to the publisher’s ad server, so that a bid value can be injected as a key value parameter. With that key value in hand, publishers can target lots of line items to very specific bid values. Specific bid values mean high-fidelity yield monetization, because now the publisher can weigh the true value of an impression not based on an average, but at an impression level. As a side benefit, the publisher can also know if any given platform can monetize an impression to begin with – either there’s a bid or there isn’t – which eliminates all that fill risk, as well as the inventory loss to system latency, because the waterfall just isn’t necessary any longer. In other words, pre-bid enables any indirect partner to work just like AdX – yield managing at the impression level, and that’s a big deal.
The other key benefit here is that because getting a valuation doesn’t require redirecting the inventory, a publisher can ask as many platforms as they like to bid on the same impression simultaneously, which drives demand liquidity. Pre-bid is a total game changer for publisher monetization strategy.
Now pre-bid sounds pretty good, but like anything, not all solutions are created equal, and it’s worth it for ad operations to dig in and review specific partner nuances. One critical factor is whether or not a programmatic platform is actually running an auction as part of their pre-bid process, or is creating some kind of estimate instead. Estimates are lousy because they bring back some of the downsides of the waterfall process. An estimate won’t be 100% reliable, so it’s possible that those systems can’t fill every impression they bid on, and that they won’t fill at the rate they bid. This is because they only run the auction after they get the inventory, which is an inferior approach.
Another important factor is whether or not a platform can return a true numerical value as a bid, versus something more generic. I’ve seen platforms that can only pass “high / medium / low”, or even as limited as “yes / no” as pre-bid responses. That’s an improvement over just using an overall average as the waterfall does, but still lousy because it’s still low-fi yield management.
We also see a big difference in how well pre-bid works in the US vs. the rest of the world. Companies that don’t have internationally sourced demand partners or datacenters outside the US have major issues with latency (that is, even providing a bid in a reasonable amount of time), or for those that are providing estimates instead of a real bid, show an especially high rate of failure on fill and monetization. That is, their estimates are especially bad on international inventory.
There’s no question that more and more dollars are moving to programmatic channels, so if you are a publisher struggling under the weight of an unwieldy waterfall, it’s only going to get worse. Modernizing your connections through pre-bid not only allows you to scale your partners, it can increase revenue by eliminating wasted inventory, and most importantly, it provides hi-fi, holistic yield management between all sources of demand.