Swimming in the red ocean of ad networks in the mid-2000’s, BlueKai was fish of a different colour: a company that wasn’t trying to sell you ads or buy them for you, but rather offered information about people. Were they in the market for a car? What kind of car? This sort of labeled intent info attached to a browser or mobile device became a way of turning the internet from a “stateless protocol” into an organised machine for advertisers.
Behind the bubbling of BlueKai was the ad tech pioneer-turned-AI entrepreneur Omar Tawakol. Born in Cairo, he re-pointed his compass from mechanical engineering to computer science as a Stanford grad student in the mid-1990s, a time when he says it was possible to “map the internet and fit it on a wall.”
- WHICH-50 Reader Survey: Help us improve Which-50 by taking our annual reader survey. You can also win prizes including Google Home, Google Home Mini, Kindle Paperwhite, Beats by Dre Headphones, or Philips Hue Smart Lights.
Aware that the marketplace for AI wasn’t ready – it was “nowhere near its current popularity,” he says, – Tawakol was drawn to the idea of imputing a person’s interests from observed behaviours on this wall-mounted internet. But he took a job at Navio (a spin-off from Netscape) before sailing forth on his own.
Martin Kihn: What was your first entrepreneurial venture?
Omar Tawakol: I started a recommendation engine company called CoRelation. We used a suite of different methods that were akin to collaborative filtering, but we factored in the product context like category and content. So we didn’t just recommend the most popular product, like Harry Potter. Barnes & Noble used us for almost a decade. Nordstrom used us. We were acquired by the company that became AudienceScience in 2002. We were like ten people and AudienceScience was 100.
After that, we approached the board and said why don’t we take real-time analytics and apply it to ad serving. Put people into groups, find segments, and recommend content [i.e., ads] back into the ad ecosystem. It was the beginning of behavioural targeting at AudienceScience. It started out as just a project with a few customers and within a year it became the whole company.
Tawakol is quick to acknowledge he wasn’t the only person with the idea of targeting ads on behavioral data. His mentor was Dave Morgan, founder of Tacoda, a similar company ingested by AOL in 2007. Tawakol spent a year at mobile analytics company Medio, acquired by Nokia, and it was during this time he “got the itch” that would become BlueKai.
MK: What did you think was different about your idea for BlueKai?
OT: Back then, everybody believed that to make money in the ad business you had to run some variant of an ad [network]. You had to get paid from the campaign budget. But I asked myself a really simple question: I said, you have companies like Google and Yahoo, and do I think that this new company can become the #1 or #2 ad company in the world? No. But I think we could become the #1 data exchange that supplies the entire ecosystem.
And we had the belief that over time data would become an increasingly important part of the ad ecosystem. One day it would become more important than the actual ad unit itself.
Of course, Tawakol wasn’t alone in these musings, but he got inspiration from an unlikely source: a high-profile disrupter in another industry.
OT: I remember talking to the guys who kicked off Kayak. They looked at businesses like Expedia who sell tickets and they said, you know, a huge portion of their employees are doing customer service, because they’re actually taking the transaction. And this small company Kayak appeared that realised the real value [to the consumer] is the search. I looked at the ad ecosystem and realised targeting data is more important [than media], and the ecosystem can’t get enough of it.
So I founded a company that was almost religious about: We’re not going to sell ads! And we never did.
After its initial funding round and a pause to assemble the crew – an essential step in any Hero’s Journey – BlueKai had its first all-hands meeting in the first week of January, 2008. It included Tawakol’s co-founder Grant Ries, Mike Bigby and Alexander “Hoosh” Hooshmand, a veteran of our old friend Right Media.
They started fast, beta testing by July. Tawakol and Ries were able to convince five major players to opt into their data exchange: Kayak, Expedia, Cars.com, eBay and Datalogix.
MK: What was the product you were building at BlueKai, exactly?
OT: We basically built a real-time data exchange. We wanted a real-time bidded auction system for data, and we did that. There were some really interesting problems about what it meant to set real-time prices for data – which is an invisible thing that can be infinitely replicated. So why would someone bid for it?
Our most valuable segments in the early days were auto intenders. Also retail. We were focused on intent. But buyers started to come to us and said, we want demographic data – which we thought was lower value then intent in terms of what they could do. But it was the common and accepted language for targeting from the offline world. I was surprised how fast we had to add a very large volume of demographic data.
We knew the utility of the demo data that was coming through the major sources wasn’t as good as the intent data we had assembled. But that’s what the market wanted. An entrepreneur has to listen to the customers even when it doesn’t fit with your theoretical model of how the world should work.
MK: Who were your buyers at first? Was it the media agencies?
OT: No, the agencies weren’t ready. They liked the concept and they wanted to test it. But the ad networks were our first buyers because they weren’t testing – they were buying. We’d give them a pixel and the next day they were buying, they were very fast. They realised they could use this [data] as a differentiation. We had intent data literally from eBay and Expedia and Kayak which nobody else had access to.
I give Hoosh credit for making us focus on the right sources of revenue. He’d draw these charts and say … Google and the DSPs are taking revenue away from ad networks, and the [buyers] won’t spend unless the agency loves us. So every six months or so we had a rethink of our strategic focus.
And since BlueKai happened to be launched around the time of that rectangular harpoon into the heart of all ad tech voyagers, i.e., the iPhone, I have to ask …
MK: How did you handle the rapid shift from desktop to mobile behaviour?
OT: It wasn’t as much of a risk to the business as I thought. There was one thing if I could go back in history and do it differently, I would. And that is the approach we took was to build mobile identification capabilities were probabilistic. They were high volume but not high accuracy. That’s pretty much the way everyone was going at the time. We should have taken the time to build a much higher accuracy but lower volume mobile match based on known identity because over time, that approach would eventually scale.
The problem was moving from online to mobile, the targeting accuracy wasn’t as high, and that wasn’t good for the long-term adoption of that category. What ended up working well in the long run… was when people started doing direct match to hashed email identities.
Famously, BlueKai morphed within a few years from a pure-play data exchange that derived its revenue as a per cent of media spend into one of the first data management platforms (DMPs). The DMP is a subscription SaaS product that has a UI, storage, connectors in and out – many things other than data.
MK: Why did you decide to launch the BlueKai DMP product?
OT: From the very beginning, customers were pushing us toward it. In addition to buying our data, they wanted to use our tech [for management]. Expedia was doing this. eBay eventually did. This was as early as 2008, but we just didn’t have a separate business model for it. We had contract addendums. We didn’t call it a DMP and didn’t focus on it as a business.
For the DMP, we had to separate out an entire stack for a different [purpose], different salespeople, different business model, different revenue recognition — everything changed.
MK: But why did you go through that ordeal?
OT: Picture yourself on a boat, and you’re on the river, and there’s people on one side of the bank with their weapons and people on the other side with their weapons, and you’re floating through. You’re vulnerable – getting hit by people who have ad systems, the DSPs or the ad servers, and they have a position of respect in the end-buyer, the customer. And then you have people like Omniture who are on the site. You weren’t the system of record for the advertisers.
In the early days, we’d talk to [the C-suite] personally. They thought data was kind of sexy. As the category matured, they would relegate us to speak to the agency, and we didn’t have that warm relationship.
Also we realised that people would always value their first-party data before looking to third-party data [like BlueKai’s]. They would see the value of third-party data through the lens of how it intersected with first-party data. That was strategic. They could invest in multiple ad networks and DSPs and portals, but they needed one system of record. That was the DMP.
We took more than half the company and put it on the DMP. It was a risk from the investors’ perspective, a speculative thing. But by the end of the year, they were popping champagne.
The late 2000’s were the heyday of the independent DMP, with players like Demdex and Krux, and DMP-DSP combos like MediaMath and DataXu riding the high tide of programmatic buying. Then in 2011, Adobe acquired two year-old Demdex, which TechCrunch described as a “behavioural data bank,” and the big clouds started to storm.
MK: Who were you most worried about as competitors?
OT: We weren’t worried about Demdex, but we were worried about Adobe owning it. Krux was just getting started. A bigger worry were the DSP-DMP bundles [like MediaMath]. Now in the end it turned out that our thesis was right – that in the long run you could prove to advertisers that you want to keep your data and your media distinct … I would say that the category is slightly in favour of our approach. That’s why Adobe scaled and Krux scaled. All three of us stayed independent of media.
And then – in a bold strategic swoop – Oracle acquired BlueKai in 2014 for $420 million, a 10X net revenue multiple. So began a dizzying round of Oracular decisions and revisions. Tawakol was named GM of the Oracle Data Cloud (ODC), not to be confused with the Oracle Marketing Cloud (OMC) – both of which held parts of BlueKai.
OT: When Oracle acquired us they split BlueKai in half. They wanted to build a business bigger than BlueKai by monetising data, growing beyond cookies. So we acquired several other companies: AddThis, Crosswise, Datalogix – and they were planning to acquire Moat when I left [in late 2016]. That became a very large business. The pure data business was even more scalable than we originally thought.
The DMP is strategic, but the data business just kept on scaling. What I’ll never know is if it was because we had Oracle behind us. Datalogix and AddThis owned their data [and] didn’t have to pay for their margin…. The data business on its own is tough, but once you get to a critical mass – and you have enough gross margin coming along with revenue – it became even more [successful] than I originally imagined.
MK: What happened to the BlueKai DMP after it was separated from the data cloud?
OT: I don’t think it helped the DMP business in the short run. Many of the people with the expertise in advertising went to the [Oracle] Data Cloud. Many of the new people who ended up servicing the DMP came from a different type of mar-tech, and that mismatch had a temporary cost to the business. A few years later, the DMP ended up getting tightly aligned with the Oracle Data Cloud and I believe that was a very good move.
Having said that, the advantage of the original split was that the Oracle Data Cloud was extremely focused on selling data to the ad tech ecosystem.
The day after leaving Oracle, Tawakol says, he talked with Ahmad Abdulkader, a renowned AI expert and lead architect for many of Facebook’s applied AI projects including Deep Text. Realising “we’ll never have more data than Google or Microsoft if we try to optimise productivity using email or calendar,” they talked about applying AI to voice in the workplace.
MK: How did the idea for your new start-up come to you?
OT: I wanted to build something that impacted people everywhere, not just advertising … People spend a lot of time in meetings. How could I make that time more productive? So we applied AI to the problem of [voice capture] and productivity. Because I had Ahmad as a partner, I was able to pull talented engineers from the majors including Apple, Google, Microsoft, and LinkedIn. That is very hard to do as a startup.
In October, we started prototyping. We were funded in January 2017. The full founding team showed up to work in Menlo Park in February.
Called Voicera, Tawakol’s new venture features an A.I. avatar/assistant named Eva who you invite to your meetings and calls; she provides transcriptions, highlights, to-do lists, reminders and so on. And indeed, Eva herself attended my video conversation with Tawakol. Afterwards, she sent me a recap and this summary word cloud, pictured below.
And so Tawakol closes the loop, from AI to AI. In between, he changed ad tech. “Blue kai” is English and Hawaiian for “blue ocean,” an allusion to a strategy for finding uncontested markets. For now, Tawakol and Eva are charting their course for the open seas.
*This article is reprinted from the Gartner Blog Network with permission.