Monday 13 June 2016 5:14 am

The next internet: AppNexus CEO Brian O'Kelley lays out his vision for the programmable age

At six foot five, there isn’t much which can dwarf Brian O’Kelley. His adtech company AppNexus, which allows marketers and publishers to buy and sell advertising space in real-time, has grown exponentially since its founding in 2007. Today, the firm’s ad-serving infrastructure processes 5m requests per second at peak periods, facilitating an estimated $2bn of transactions through its platform, and taking 10 per cent.

Nonetheless, AppNexus sits between two bigger beasts. Google and Facebook have giant ecosystems in which users spend hours every day, and both control the advertising sold within. In contrast, AppNexus is not a consumer platform. It remains independent. While the other two giants are its rivals, it is not in direct competition with them. AppNexus is able to sell into Google and Facebook (before FBX closes in November), and as Martin Sorrell said in 2014, there is virtually no other “agnostic” game in town.

Rumours are swirling around where O’Kelley might take his adtech titan next. There might be an IPO. The firm might be bought. The pace of change is so quick in his industry that O’Kelley can’t say where AppNexus will be in a decade.

“Facebook was tiny 10 years ago. It was a college kids’ app. If you can tell me what consumers will be doing in 10 years, I’ll tell you where AppNexus will be.

“Advertising sits within a circle of consumers, publishers and marketers. AppNexus’s job is to find any standardised, digital market of scale within it – to be the most efficient way for buyers and sellers to connect. So if everyone’s wearing virtual reality headsets in a decade’s time, then we’ll be the best real-time decisioning virtual reality advertising company.

“Those which can adapt will thrive. Just by virtue of their size and scale, I bet those giants won’t be able to make those changes as fast as we will.” 

Father of the ad exchange

A computer scientist, O’Kelley earned fame in the adtech world for founding the first ad exchange while he was working at Right Media.

“It became clear to me that advertising’s technology was incredibly inefficient. People were making very bad decisions. A publisher would get a report from a buyer which said that their ads had been sold at a $10 cost per thousand impressions. In reality, only 10 per cent had been sold at that price; the rest were thrown away. But the publisher thought that getting those $10 was amazing.”

Read more: Why publishers need to keep pace with programmatic

He built the technology which allowed publishers to choose the best buyer for every impression, before turning his attention to how buyers could be helped. The ad exchange launched in 2005, and has developed into the real-time auctioning infrastructure which today allows millions of buyers to bid for an ad impression before the publisher’s site loads it.

At the time, it was hard to sell this efficiency to either marketers or publishers, he explains. “For any kind of exchange, you need liquidity. When we launched, around 80 ad networks joined, transacting billions of impressions every month because of the efficiencies they gained. That started to create liquidity, then publishers and agencies started to join. We needed critical mass to get past the chicken and the egg problem.”

It was a year and a half before the first “major” – Yahoo – joined. Nine months later, Yahoo wanted to buy Right Media.

After the sale, O’Kelley and another Right Media colleague founded AppNexus, which has received $300m of financial backing to date, from the likes of Sorrell’s WPP. A number of acquisitions have allowed O’Kelley to plug technological gaps and increase its market share. Last year, he bought Yieldex for $100m, allowing AppNexus to beef up its offering to publishers to include forecasting, pricing, and direct-sales analytics.

Last week, AppNexus announced that it will add video to its deal with Microsoft. And O’Kelley has been open about approaching Goldman Sachs and other investment banks about a flotation.

The programmable age

If AppNexus is having a big moment, O’Kelley thinks the internet is entering a whole new era. “In the age of programmatic, buyers and sellers of advertising have been connected. Publishers are simply able to choose the marketer who will pay them the most money. But it’s still a struggle for marketers.”

Once again, it’s a human problem. There is currently 30 times more traffic than in 2007, making it hard for marketers to bid on the inventory best suited to them.

“Buyers have to bid into 100bn auctions a day if they want to get a full view of all traffic. Most of the trading desks we work with have a team of people creating campaigns which fit certain inventory sets, and they have a hard time managing more than a few thousand campaigns.”

Automation and machine-learning would provide scaleable ways for buyers to find the right inventory, he says. “If you know a customer has multiple devices, and browses certain sites, when is the right time to serve them your ad? If you miss your chance, when is the right time to try again? Data science can find incredible patterns in human behaviour. It’s a very Matrix-esque way of having all inventory cascade down your screen, enabling you to choose that one, and that one.”

Read more: As social media gets better at monetising, what can marketers expect?

AppNexus, he says, would provide a data science platform to build marketers their own algorithms, allowing them to bid on the very best inventory across the internet.

Technical challenges

The kind of proprietary and portable algorithms O’Kelley envisages are far from today’s reality.

“There are incredible technical challenges to allowing an arbitrary algorithm to run in real-time and bid on 100bn ads a day. We’ve just crossed the threshold where we can see that it’s possible though, and we can now play with that.”

He does think that agencies and their holding companies could be investing more into data science than they currently are.

“Only a few have made substantive investments into data science. The worst thing for agencies and their holding companies would be for their clients to be better at this kind of programmable marketing than they are.”