Gotta Stay on the Page: Index’s Alex Gardner Opens Up on Header and Server-to-Server Integrations

Does the Buy Side 'Get' Header Bidding?

It’s been a busy few months for Index Exchange. In April, Google announced the exchange as a launch partner as it opened up its Dynamic Allocation server-to-server connection within Doubleclick for Publishers to third-party demand. Then in May, Index and AppNexus announced seamless integrations with each other’s wrappers, pointing toward potential header standardization.

Before he leads a kick-off session with Time Inc.’s Kavata Mbondo at OPS on June 7 in NYC, we caught up with Index SVP of Partner Development Alex Gardner and assailed him with questions (well, some he knew were coming) about a wide array of topics. For starters, we dove into the ups and downs of integrating into Dynamic Allocation and the potential for server-to-server connections; why it’s essential for demand sources to have code on the page; and what the buy side currently gets and should understand about header implementations.

GAVIN DUNAWAY: So what’s the lowdown on this agreement you have with AppNexus?

ALEX GARDNER: We sat down with them and agreed that if we can ensure there are standards that will facilitate or simplify integration, create more efficiency and ensure that everything is working as the publisher intended, that will probably accelerate adoption and make everyone’s lives a little easier.

We’re still going to compete. We will still want our wrapper installed; they’re still going to want to have their wrapper installed. But let’s allow the publisher to decide, and based on that decision, let’s just support each other and ensure that we can deliver to the publisher what it is they need in terms of liquidity and demand without having to deal with all the headaches of integrations and ensuring that these two technologies play nice.

The other underlying current is we elected to be launch partners with Google to open DFP’s Dynamic Allocation to third-party demand.

The principled exchange for publishers.

GD: How did you know I was going to ask about that?

AG: It’s the hot topic. Our decision to go that route boils down to choice for the publisher. Why limit how they can access demand if approaching it via Dynamic Allocation within Google’s ad server is what’s going to allow them get there? And the reality is that they can leverage both: they can run the Index Exchange demand via the header and they can run via Dynamic Allocation. We want to eliminate the friction and barriers to providing choices for publishers.

Do I think what Google is doing marks the death of header bidding?

GD: It’s like you’re reading my mind!

AG: Ha! It’s absolutely not the end for header bidding. If we go back to where it all stemmed from, it was the buy side that really conceived the idea. Retargeters were the first to run tags on the header of the page. And now we’ve begun to enter interesting agreements with buyers that want that position, because looking for the true promise of programmatic, which was to eliminate the paper and inefficiencies, and reach the right person with the right ad at the right time. Essentially they’ll have on-demand access to really high-quality premium supply that they can purchase in the same way that they would a direct buy.

The true promise of efficiency that people envisioned in the early days of programmatic is what we’re now starting to properly leverage via the header. And that’s the kind of thing you just can’t really do via the server-to-server into Google’s Dynamic Allocation.

There are still a lot of unknowns with Google’s solution – it’s very alpha. There is no deal support, which is a big part of why we still want to be on the page. What direct header implementation provides in terms of sophisticated deal support and programmatic guaranteed versus server-to-server is pretty significant. Not being able to do those through the Google solution is a limitation.

GD: Why is being on the page considered such a necessity?

AG: Being on the page affords you opportunities that server-to-server can’t provide. When we have our tag on the page, we literally see everything. We’re not limited in any way to having visibility across all supply, which is incredibly powerful.

It also creates a conduit to do other interesting things. When we have that position on the page, it’s conceivable that we could start to incorporate intelligence or functionality similar to ad serving that is currently relied upon today via the likes of DFP. It also helps us manage things like viewability. Say if a particular slot is consistently running below the fold and is negatively impacting the viewabilty score, we can easily turn it off on an individual ad unit basis because of the mapping we do.

Server-to-server introduces uncertainties. If another provider was running a header-based server-to-server solution on a given site, they may own that position. For other demand sources to gain access to the site’s inventory, that one provider could mandate a server-to-server connection into their platform, as opposed to wrapping other third-party bidders, which is the strategy we’ve elected. There’s a potential for black box. It brings us back to a place that we were working really hard to get out of with Google’s Dynamic Allocation, where the strength of our bids weren’t always being accurately represented to the seller.

GD: Is the server-to-server path intriguing at all to Index?

AG: We fully acknowledge that server-to-server has a ton of advantages. We can support it today. Any major platform that’s interested in doing a server-to-server into us we would support. Hooking into another platform’s server is where we get a little bit apprehensive.

GD: That is pretty much the biggest hangup with people who are trying to do those kind of server-to-server connections.

AG: It’s a challenge. We know server-to-server makes a ton of sense, because it’s low or no latency, incredibly fast and a very clean approach. But, there are issues. Unless everyone goes in that direction and there’s some kind of standardization, I think it’ll be hard to go that path. But if people want to integrate into us, we’ll happily support it.

GD: Is header bidding changing buyer approaches to the market? I’ve heard talk about buying more on an upfront basis. 

AG: We’re advocating for such practices and I think the buyers are definitely keen. Buyers like Omnicom Media Group are essentially using the programmatic pipes as a utility and they’re transacting or negotiating directly with the seller. More important, they’re paying the seller directly, which is the truest form of transparency out there.

MediaMath was arranging this for some special clients with the “guaranteed plus” model. There are no hidden fees, no dynamic floors or soft floors or anything of that nature. It’s a conversation between buyer and seller, the terms and details of which are fully open and shared, and the seller is keeping as much of that dollar from the marketer’s wallet as they possibly can.

However, there are potentially interests in the marketplace that want to avoid or stall the trend, because it eats into margins for intermediaries like ourselves. We’re in this to make money, obviously, but we’re organized in such a way that we can withstand narrower margins. If you look at our publicly owned competitors their take rates have been moving up and to the right. Supporting this model becomes more of a challenge, because they have responsibilities that would require they continue to make more money. And while they’ve scaled well, we’re scaling with more efficient operating costs.

GD: This is something we’ve been asking a lot at various conferences: What does the buy side need to know about header bidding? 

AG: It’s a good question. I think the first thing the buy-side needs to know is that header-based selling isn’t going anywhere. The second is that it serves to help them meet their marketing goals, not hurt them. In February, for the first time, spend on our platform was greater via our header implementation points than spend from traditional tags, and the gap has only widened since. We contemplated signaling header versus non-header in the requests to provide more transparency to the buy-side, but we realized it was kind of pointless because eventually it’s all going to be header – in which case it’s just inventory, and thus kind of eliminating or diluting the need.

When I run a single tag in an ad server waterfall, what is first-look Tier-1 inventory for me may actually be sixth in the chain. It was the first look that I was seeing of that particular user, but where I stood in the hierarchy or in that waterfall could’ve been much, much further down. The notion of premium has a certain baggage that comes with it, depending on where you actually reside. So header inherently levels all of that because all the exchanges are seeing things in parallel to one another. This means the buy-side will have more opportunities to see the users they want to reach and more opportunities to buy them.

GD: It really brings true the promise of programmatic – it was supposed to level the playing field.

AG: And there were so many weird inefficiencies with the waterfall, not the least of which was just the quality of the supply that was being represented to buyers. Having that level playing field just provides more clarity and a more accurate representation of the underlying supply corresponding to whichever publisher.

GD: So with the buy side, do the buyers themselves understand what’s going on with header bidding, or is it kind of a mystery to them?

AG: It’s hard to avoid what’s going on. The trades, yourself included, do a good job of identifying what’s hot and talking about it. There will always be some buyers that are more sophisticated than others and those who lead the charge. I put MediaMath with their guaranteed plus program and OMG with what they’re doing amongst those leaders. Everything that we’re doing ultimately stems from a place of trying to establish more credibility and confidence and trust in the buyers’ minds.

GD: What about DSPs? You mentioned MediaMath, but you’ve got a lot of agencies using multiple DSPs, so there’s worry about people bidding against themselves now in the header. In the end, the buyers are fine as long as they get what they went looking for, but I can imagine DSPs roaring in frustration.

AG: There’s certainly concern about bidding against yourself that gets raised all the time, but it’s been a concern since the very beginning, when you had multiple exchanges running via tags. The scenario of course is a little bit different this time.

Truthfully we don’t hear about it a lot; it doesn’t sound like it’s a major concern. The DSPs are asking us, “What is your strength and who are the publishers that you are accessing via the header, and can we create packages around those?” That suggests that those are of particular interest.

It’s possible that they’ve bid against themselves. Does it happen at high frequency? We don’t observe it to an extent that would be of concern to the buyers we’re working with, to the extent that they’ve mentioned it as a problem, and we only hear good things from their perspective as to what we’re doing and the effort we’re putting in the marketplace to gain access to header. So we have to acknowledge it, and we address it, but we don’t think it’s hindering that kind of progress.