What's Fueling Rocket Fuel: Smarter Machines, People Too
by Joe Mandese, Jan 17, 2013, 2:36 PM
Real-time programmatic buying optimizer Rocket Fuel this morning reported record revenue of $107 million for 2012, a 238% increase over 2011, and outpacing most programmatic marketplace growth estimates by a wide margin. The company attributed the rapid expansion to improvements in its core artificial intelligence, machine learning and robotic media trading systems, and the fact that existing customers have been expanding the amount of buying they place through Rocket Fuel’s systems, because they work so well. In fact, while Rocket Fuel did add an impressive number of new advertisers in 2012, its 93% growth in new customers trails its overall growth rate by a margin of more than two-to-one.
But don’t worry about machines taking over the business anytime soon. If Rocket Fuel’s own growth is any indication, the type of technology they have been developing actually increases the need for human talent -- albeit not in the manual media trading part of the business. During 2012, Rocket Fuel increased its human resources 126% to 289 employees, including even more rocket scientists, but also sales, marketing and support teams. Much of that expansion, and a good chunk of Rocket Fuel’s revenue growth, came from its expansion into Europe. Rocket Fuel added six offices in 2012, including Amsterdam, Atlanta, Hamburg, Raleigh-Durham, Toronto, Washington DC, and now has offices in 15 cities worldwide.
Rocket Fuel also entered the emerging Japanese programmatic buying market, aligning with that nation’s largest digital marketing company, cyber communications inc. (cci), a wholly-owned subsidiary of Dentsu. Under the terms of the strategic alliance, cci runs its digital ad management service, “PerformanceX Management,” on top of Rocket Fuel’s platform, offering full-service brand and direct response campaigns to clients.
On the eve of this morning’s announcement, RTM Daily spoke with Founder and CEO George John to find out exactly what’s fueling Rocket Fuel’s growth.
RTM Daily: What were the main drivers behind your 2012 success?
George John: Well, the product just works well. Part of what we love, and what we love to hate about the advertising business, is the different phases of partnering with an advertiser or an agency, that are rational or irrational. We always know that it will be hard to get in their door, because there’s so many other people trying to get in there too, but once we’re in, it’s kind of beautiful and rationale things start to take over, because we perform. And if you’re always at the top of what performs for them, they’ll keep allocating more budget toward that.
Most of our revenues -- about 84% of revenue in 2012 -- were from advertisers that were at least in their second quarter of spending with us. What we have seen is that between the first and fifth quarters of working with us, advertisers are spending 2.6 times what they spent in the first quarter.
Some of that revenue came from some new things were were doing, including our work with Facebook and their FBX exchange, but for the most part, it was organic. Even without the Facebook campaign work, 90% of the revenue growth was from existing customers.
RTM: So your product is so good it sells itself? But what did you do to improve your product in the past year that helped it sell itself?
John: If you were to look at our core product, it was about a year ago that we pulled away the last bit of daily human oversight and attention that used ot be required. A little over a year ago, we would run a performance campaign by a target rate or a cost-per-action, and we would instruct our models to buy over a certain threshold of quality, and we would have to check in on it daily to see if we were meeting or exceeding that threshold. Well, it’s been a little more than about a ear since we got away from that too, and said, “Can’t we just have the computer do it?”
What we discovered is, that when you map it into a machine environment, you’re no longer constrained by a human viewpoint, and you can get much more granular than a person can. Now it’s flying on autopilot and we get to relax and look out the window while the machines do all the work and the humans are focusing on superior analytics.
RTM: Explain what you mean by “superior analytics.” What are people doing now that the machines are doing their own quality control on themselves?
John: We think of those superior analytics as experiments -- people making adjustments and trying different scenarios until they prove themselves. And that's when we let it autopilot. There’s not a requirement for a human to be around in the beginning. All the machine needs to know is the budget, the goal and the performance target we need to hit. But the humans are sort of the upside, really.
RTM: Do different humans, you know, clients, have different upsides?
John: What we see, is direct response advertisers are pretty rationale. We we love the fact that they just look at their DART reports, or whatever reports they are using, and we’re the best ones delivering for them, so they just keep upping their their budgets every quarter.
Brand advertisers are different. They may be doing a new product launch campaign with us, and that’s where we see brand advertisers come and go with their spend, depending on what they are focusing on at that time. Sometimes it is difficult for them to translate their brand objectives into specific thing that we can make happen online. There’s the rating point style campaign that can be measured online. Or there’s some kind of brand consideration metric that can be measured via surveys or engagement actions on their site. Once we have that, we just let the machines run.
RTM: If the machines are so good, why don’t you just let your machies sell your product for you?
John: I keep threatening to create a robot that will sell the pants off our sales team. But selling is still a very human process. When we launched in Europe in 2011, it took time to sell it, and it wasn’t really until 2012 that we got critical mass there.
I’m an engineer, but I was also a sales rep, and I understand the value of both sides of the business. When I was an engineer, I didn’t think we should pay our sales people that much, because if the product is so good, it should sell itself. When I was in sales, I felt our product is so good, we just need more people to go out and sell it.
RTM: With all your focus on machines, robots and AI, what’s the biggest challenge you have selling your product?
John: Well, as I said, last year we removed the last requirement for any human intervention. AI was the most significant part of that, in my view. But we still need to humanize that for our customers. Otherwise, it’s just a black box. Some of them come back to us and say, “We’re happy with our performance, but we don’t really know how it’s doing it. So last year, in the first quarter, we created an early version of our Insights Booster, which basically creates a nice infographic on the campaign explaining the audience segments that are working well. It has a persona tool that helps describe what audience is working well by selecting images of different personas from a library of thousands of pictures coded by variables representing those personas. It’s a way of putting a face behind the performance aspects of the campaign. That’s one way we are humanizing what we do. When you’re dealing with AI and machine learning, it’s not always that obvious what is the best way to explain to customers what you are doing that is working well.