DataView: The DataXu Blog
Category: algorithms
- The 5 Signs of Cultural Transformation in the Media Business
06/07/2010 | Categories: advertising, algorithms, analytics, data, machine learning algorithms, media, media agency, media trading deskI just finished reading “The Big Short: Inside the Doomsday Machine,” an excellent book about the financial crisis of 2007-2008, by Michael Lewis, one of the most entertaining and perceptive story tellers of our generation. I actually had the pleasure of meeting Michael Lewis the other day and hearing his views on a number of topics, the common thread being “cultural transformation” in industries where, as he put it, “aspects of what they are doing don’t make sense anymore.” Naturally, this led me to ask him about the business of advertising and media.
In his first book, “Liar’s Poker,” Lewis observed a cultural transformation changing the business of bond and securities trading at Salomon Brothers. Why? In his view, cheap computing power led a “rapid intellectualization” of the field. So Vinny, the high school graduate from New Jersey with hair coming up from under his shirt, now had to run his bond trades next to a PhD from MIT who could use mathematical models to understand the option value of the security. As a recent Ivy League graduate, Lewis was given the advice: don’t socialize with Vinny – he might hit you!
In “Moneyball,” Lewis chronicled this same “intellectualization” of the sports industry, as Billy Beane and the Oakland A’s relied on complex statistical techniques to scout and hire the most promising baseball players. Rather than rely on dated but widely accepted measures of performance, like runs batted in and batting average, people like Beane used the new less obvious measures, like on-base percentage, to identify under-valued talent.
With apologies to Lewis, I offer below my thoughts on how he might one day tell the story of a similar cultural transformation of the advertising and media business:
1. Something the industry has been doing doesn’t make sense? Check
- As the consumer migration to digital media roars on, the business of providing and monetizing the media is more manual and inefficient than ever; basic things like counting, tracking, and improving the performance of advertising are staffed by small armies of low paid workers.
2. Cheap computing power disrupting the traditional economic structure of the business? Check
- As advertisers look for “faces rather than places”, engaging audience is emerging as the focus rather than placing advertising is premium media contexts. Google has shown the world that with the right data and algorithms you can conquer the media world.
3. Math geeks are using advanced analytic techniques to better predict investment ROI? Check
- Predictive analytics, data models, and machine learning algorithms are starting to substantially improve the targeting of advertising and reduce waste.
4. Tensions exist between the old guard and the new? Check
- There’s a growing stream of angry sounding “algorithms don’t matter” posts from traditional advertising types, although unlike Vinny I don’t think they like to hit people.
5. The “new math” is creating new winners and losers? Check
- I recently spoke to a senior manager in a media agency who told me that his algorithm-powered “media trading desk” (which buys media through online ad exchanges) produced more profit in its first year of operation than the entire traditional media buying agency from which it was spun out.
The forces at work in the media business are indeed familiar, and most now agree that we are on the cusp of a transformation of the business itself. I’m looking forward to perhaps one day reading Lewis’ account of it. Meanwhile, let’s hope the leaders of this industry learn from the mistakes made by Wall Street.
- Mike Baker, CEO, DataXu
(**This piece was also posted on the MarketShare blog at Forbes.com – http://blogs.forbes.com/marketshare/2010/06/07/michael-lewis-and-the-cultural-transformation-of-the-media-business/ )
- If 6 Turned Out to Be 9
05/24/2010 | Categories: algorithms, audience buying, data, demand side platform, DSP, Google, Invite Media, media, media planning, online display advertisersIt’s no secret that Google has been shopping for a so-called Demand Side Platform (DSP) for some time. With the AdMob acquisition signed off by the Feds, the rumors are that Google will acquire Invite Media (http://mediamemo.allthingsd.com/20100523/with-admob-out-of-the-way-is-google-set-to-buy-invite-media/), a DSP known primarily for a user interface that enables audience buying across exchanges. This move seems sensible enough as a way for Google to shore up its exchange user interface, which even Google concedes needs more care and feeding. But would it undermine the core value proposition of the DSP?
A history lesson: before DSPs, online display advertisers have had to rely on the sellers of media to target, price, and optimize media. That’s why agencies have become so reliant on ad networks. And that’s why their clients increasingly are questioning the value added and the fees paid for media planning in the digital era. Enter the DSP. Using a DSP like DataXu, an agency and its client can for the first time effectively crunch their own proprietary data, develop their own campaign algorithms, and seek price/performance optimality across multiple sellers, effectively breaking the reliance on the seller to deliver all the value.
So the question is: would buyers rely on the largest digital media seller for their media investment allocation decisions? Would a Google system ever have a buyer take money out of Google media and allocate it to Microsoft, Yahoo, Apple or other supply-side aggregators? Consider the potential similarities to Google’s search engine marketing model – very efficient, but advertisers must accept the one-size-fits-all approach and seller-controlled black box algorithms. If you’re a managing director of a digital agency, consider how much value you are creating for yourself and your clients if you leave the work to the seller. Isn’t that how the agencies got into the current pickle in the first place? It will be interesting to see how this all plays out!
(**Another history lesson: For those who may not be Jimi Hendrix fans, the title of this post references one of his greatest tracks – and the potential role reversal of sellers becoming buyers.)
- Mike Baker, CEO, DataXu
- Shortening the Cycle Time for Data Driven Insights
03/11/2010 | Categories: algorithms, data, media, media efficiency, Yahoo!In a recent Q & A in AdAge, Scott Burke, a VP of Engineering at Yahoo, shared some great observations about the use and valuation of data by marketers today. Two points, in particular, resonated with me, as they reinforce key aspects of our mission here at DataXu:
1. Shortening the cycle time for data-driven insights will benefit everyone involved in online advertising.
2. “It’s not about how much data you have, it’s what you do with it.”
In the article, Scott mentions how Yahoo strives to accelerate their decisionmaking internally, and do so in an “iterative loop” of hypothesize, test, learn, repeat—a paradigm that describes our platform’s learning system quite well. Scott also echoes our philosophy that the ability to derive insights from data more quickly will make all parties involved in advertising more effective. We built our platform to analyze enormous amounts of data with unprecedented speed, in order to make highly informed decisions about the barrage of ad impression opportunities available at any given second on the Internet. We also recognized the value of learning from each of those decisions, to continually improve the system’s decisionmaking data models and algorithms going forward. This dynamic, closed loop learning model – especially when combined with real-time bidding – gives our clients a distinct advantage over more static approaches to buying media online. We extend the continual learning outward from our system, too. Our clients gain timely, granular insights from their campaigns that they can apply to better inform a range of marketing efforts.
The ability to leverage data in this dynamic way can also extract more value from the data itself, as the quote from Scott in my second point above illustrates. It can be challenging for marketers to anticipate what type of data will have the most impact on a campaign’s success, but when applied within an intelligent system, you can pick out the signals that matter for the specific campaign. More data isn’t always better. But more advanced decisioning can make limited data do big things or elicit the select, valuable gems among larger sets of data – both tasks that are nearly impossible for an individual to do without the benefit of some powerful technology. (For more on my thoughts about the role of data in display campaigns, please see a recent column I wrote for ClickZ.)
It’s great to hear validation in the marketplace that what we offer our clients is relevant and meaningful. The amount of data marketers have at their disposal is growing by leaps and bounds, so our goal is to continue to develop the best system for using that data to improve media efficiency and lower customer acquisition costs.
