From Waterfall Mediation to RTB: The Rise of Real-Time Auctions

2023 may be remembered as the inflection point when the mobile apps ad tech sector finally started transitioning away from waterfall mediation to real-time bidding. App developers and publishers have relied on legacy waterfall mediation for years to manage their advertising demand — however, today, powered by machine learning and scalable infrastructure, real-time bidding (RTB) is quickly becoming the new normal.

At the forefront of that transformation is Alankar Agnihotri, a seasoned Product Manager with a FAANG company, with experience working closely with monetization platforms, top demand sources, and publishers around the world. According to him, “RTB unlocks tremendous value for both publishers and advertisers, and has substantial long term benefits for the ecosystem. Unlike the past incremental changes in mobile app ad monetization, the shift from waterfall to RTB is a paradigm shift for the industry”

The Waterfall disadvantage

In waterfall mediation, ad networks were hardcoded in order (based on historical eCPMs or priority set by a dev) and called one at a time until an eligible ad was returned. While this model was effective in the early days of mobile apps advertising, the downsides of this approach have become so stark that they can no longer be overlooked.

The fundamental flaw with waterfall, Agnihotri said, is that it "utilizes yesterday's averages to forecast today's impression value in a static way." What this often meant in practice was you missed higher eCPM for an impression if the ad source had a lower average eCPM and was not included higher up in the waterfall chain. Even worse, publishers were forced to keep adjusting and maintaining this configuration to constantly re-prioritize, tweak pricing thresholds and comb through logs to ensure they were not leaving any money on the table. This came with a high operational cost but there was no better alternative available.

The essence of Real-Time Bidding:

Real-time bidding turns this model on its head. Instead of bid list, RTB enables all eligible demand sources to compete for each ad impression in real time. And a single auction is run in real-time, meaning the impression goes to the highest bidder — assuring true impression value is paid, and received by sellers for their inventory.

“This change alters the process of maximising revenue at a fundamental level”, says Agnihotri. “That means instead of just trying to guess who should get the first look, you let them all compete — and the highest price wins. It increases direct competition for every impression, generating higher revenue for the publishers."

RTB doesn't just optimise dollar value — it also makes for a more transparent and fair marketplace. In older systems, often lower placed demand sources would not get full visibility into publisher’s inventory —regardless of their offer at a particular impression. Such smaller demand sources now also get full visibility into a publisher’s inventory, and can over time analyze and improve on their bidding strategies.

Machine Learning: The Silent Partner

The concept of real time auction isn't new, but it is machine learning that transforms buyers’ performance. The auctions — across billions of impressions daily — are far too sophisticated for raw computing muscle alone. They need intelligence.

“Machine learning brings unprecedented performance improvements opportunities for networks,” notes Agnihotri. Modern demand networks and mediation platforms leverage predictive models to optimize everything from which ad networks to call, to what floor price to set, to which creatives are likely to perform well for a given user.

Such real-time intelligence allows smarter segmentation, more effective placement of the ads, and better functioning of the campaign. Agnihotri says, "Every served impression becomes a learning moment — and the models become sharper with every auction."

Streamlining Monetization

RTB arguably offers one of the most readily available, albeit lesser touted, benefits through ease of operations. Real-time auctions, by contrast, are mostly self-optimizing, whereas waterfall setups required weekly — and occasionally daily — updates from monetization teams.

For Agnihotri, this change is as much about the attitude as it is about technology. RTB enables your product and ops teams to leave behind playing waterfall mechanics and to start focusing on strategic growth. The heavy lifting is done for you by the system.

It removes human error from the equation, and reduces maintenance. Forget about predicting the best performing network, having granularly segmented and tediously long waterfall chains, tracking eCPM across networks in spreadsheets, or tedious re-ordering of ad sources in waterfall.

A Fairer Ecosystem

Real-time bidding also contributes to a more robust adtech environment. In the waterfall model, several smaller or domain specific buyers would frequently not get access to publishers’ full inventory due to their overall lower eCPM. This created a vicious cycle, as lack of inventory visibility would hamper their ability to improve their mediation performance, causing further lower eCPMs. In several cases, demand sources would also have contracts to get first claim on the high value impressions. RTB changes that. As Agnihotri explains: “With RTB auction, access is not a privilege, it is something to be won in a competitive landscape.”

This turn encourages innovation. In particular, this unlocks competition and scalability for smaller DSPs and new networks that may have previously been relegated to the bottom of the waterfall. This translates to more demand sources, resulting in improved fill-rates for publishers. It means better inventory access and smarter targeting for advertisers.

Challenges still exists

While the benefits are obvious, the transition to RTB brings several hurdles. For one, the infrastructure of unified auctions is complicated. Real-time fraud detection, and privacy-safe targeting are not luxuries — they are essentials.

Not every buyer and app publisher is ready on day one, Agnihotri concedes. “There is an investment curve — in both technology and in enabling the teams that will use it — that cannot be ignored."

Then there is the change management issue. For teams with decades worth of waterfall structure ingrained in their DNA, they will need to learn a new set of metrics, how to interpret auction data and adjust to a new feedback loop. But as Agnihotri says: “the friction is short-lived, but the benefits are long-lasting and compounding.”

A Tipping Point for AdTech

And 2023 seems to be the tipping point to this change. Big publishers have since made the switch to RTB-first, seeing tremendous growth in revenue and efficiency. The biggest mediation platforms are going all in on supporting more and more bidding partners, and launching novel bidding features to further improve the publisher performance.

Agnihotri is of the opinion that we are heading into a new age. Moving to RTB is not really a choice anymore — it is rapidly becoming the norm. And the sooner you get used to it, the more beneficial it will be for you.

Looking Ahead

“While real-time bidding is the big change of the day today, that is only a starting point”, Agnihotri added. “Auction intelligence is going to evolve from here on out. Creative optimization will be driven by AI. Even managing user privacy will be an element baked into how ads are served in real time.”

Conclusion

Move over waterfall mediation: the industry transition from waterfall mediation to RTB is more than a tech evolution, it is a philosophical evolution. It brings incredible value to the overall ecosystem and its participants: buyers, publishers, and the end users.

As Agnihotri puts it: "RTB is not just about increasing revenues — it is about making better auction decisions, faster and in millions of ads at the same time. That’s the future of mobile apps adtech.”

Well, in 2024, it is already here.