Why Georgia-Pacific bet the future of programmatic on supply, not demand

It’s fair to say Georgia-Pacific arrived at that view rather than started with it. The first leg of this journey was spent almost entirely on the other side.

“We had really maximized all the changes we could make on the demand-side platform side,” Shah said.  

Those changes came down to sharper inventory control, tighter campaign pacing and better targeting — all of which were layered on top of each other since 2019, when Georgia-Pacific first brought its programmatic buying largely in-house. What didn’t get touched was the supply side, the part of the stack, Shah said, where “a lot of advertisers don’t talk about or think about” the value creation opportunity sitting in plain sight.

The opportunity boiled down to timing. Standard programmatic, as Shah describes it, is a reactive, linear chain: supply gets surfaced by SSPs, the DSP fields the bid request, the advertiser bids, the ad delivers, and only then does anyone check whether the impression was worth buying in the first place. By the time that check happens, the money’s already spent.

“Standard programmatic scores inventory after the bid,” Shah said. “We wanted to score the inventory before the bid.”

That’s the gap Georgia-Pacific went looking to close. But to do so it had to find something that could sit upstream of the auction and decide which bid requests were worth letting into the stream at all, using predictive signals instead of the after-the-fact reporting most optimization still runs on. The search led it to SWYM, a tool that sits between Georgia-Pacific’s programmatic team and the marketplaces it buys from, screening inventory before it’s bought rather than judging it after. It reads live signals to flag the good placements and adjust the advertiser’s bids on the fly, catching the waste before the money’s spent, not after.

“We wanted to make sure that the inventory that we procure is effective in the sense that it’s able to improve the viewability rates versus not having the tech there,” Shah said.

The payoff, on paper, is straightforward: Georgia-Pacific says it’s paying less for the same ad space, and getting better ad space when it does pay.

CPMs — what the advertiser pays per thousand impressions — have dropped anywhere from 17% to 44%, depending on the brand and how long SWYM’s been running on it. That’s the cost side. The quality side moved too: viewability is up 11%, meaning more of GP’s ads are actually being seen rather than loading off-screen or behind another tab. Video completion rates — how often someone watches an ad all the way through — are up 7%. Shah framed these umbers as the actual test of whether any of this worked. 

“We looked at it from the standpoint of efficiency and effectiveness,” he said — efficiency being the price, effectiveness being whether people actually saw and watched the thing Georgia-Pacific paid for. Cheaper alone wasn’t the goal, though that wasn’t an easy sell internally. As Shah put it: “You can get dirt cheap CPMs and lowest-cost inventory, but it will not always be a quality impression.”

There’s a waste number too. SWYM’s fraud-detection layer, built on a partnership with Fou Analytics, cut invalid traffic — bots, essentially — by about 8%, catching it before the auction rather than after. 

“We didn’t want to optimize the campaign after the fact,” Shah said. “We wanted to shape the inventory, or even figure out if the inventory was even eligible to compete for dollars.”

Armed with a clearer idea of what happens when that works, Shah and his team could be more discerning about which SSPs it bought ads from. Shah wouldn’t confirm a current number on the record — “I can’t say how many SSPs we work with today,” he said — but by Digiday’s math, working off the case study’s 80% figure against the 30-plus GP started with, that puts the roster at somewhere around six now.

The judgment call wasn’t about who had access to unique inventory — most SSPs are drawing from the same pool at this point, after all — it was about how they packaged and delivered it. That meant judging each one on how well it turned that shared inventory into higher viewability, stronger completion rates and more efficient CPMs for Georgia-Pacific specifically.

All of this put together, added up to six-figure annual savings — a number that came from a case study not Shah, who didn’t put a figure on the savings himself when asked. Those savings aren’t to be sniffed at, to be clear, but the advertiser has its eyes on more as the algorithms powering its curation of programmatic inventory get even better. 

However that shakes out, it won’t come at the expense of Yahoo, Georgia-Pacific’s DSP — an important caveat given how much of this conversation tends to get framed as supply-side disintermediating demand-side, or the other way round.

“DSPs and SSPs need to talk to each other,” Shah said. “That’s the most important thing… that’s how you can get a win-win out of this industry.”

For that to happen, advertisers will need SSPs to be a lot more transparent  about how they sell ads, who they use to do it, and what they take in fees along the way. Expect fees especially to become the next lightning rod, just like they were on the DSP side a few years back.

“Brands and agencies aren’t just applying intelligence upstream,  they’re applying scrutiny,” said Stacy Bohrer, svp of global buyer development at OpenX. “The SSPs that can’t demonstrate real transparency and tangible inventory quality controls are going to face increasing pressure. That’s not a forecast — buyers are already moving this direction.”

The holdcos are further along that path than most advertisers, for the simple reason that they have to be. They spend enough on programmatic ads that doing the kind of curation work Georgia-Pacific has done isn’t optional, it’s table stakes for keeping the business profitable — for better or worse. That’s unlikely to change anytime soon, since most advertisers simply don’t spend the money required to justify the investment.

“Intelligence is moving closer to supply — where signals are cleaner, identity is stronger, and quality can actually be measured rather than inferred after the fact,” said Bohrer. “Concretely, that’s IQ scoring, ad load and ad-adjacency filtering, brand suitability controls, and capping the supply chain at no more than two nodes. Buyers can see it, not just take our word for it.”