Several weeks ago, I spoke with Claudia Perlich, *Media6Degrees' chief data scientist, and asked her to define "data scientist." Her definition spanned four paragraphs, but it boiled down to this: "Data science lives in the intersection of understanding not just the results of the algorithms, but also the subtle caveats of their applicability and the problem that should be solved."
With that definition in mind, consider this: Graham Cooke, an ex-Googler and current CEO of Qubit, believes CMOs should become effective data scientists. He believes that's the only way to keep pace with what he calls "real-time retail."
Cooke describes Qubit as a company founded to simply "make Websites work and sell better." He says that online retailers don't change their Websites very often, and do a "code freeze" on Websites during peak trading months. "They don't want to take any risks that the site will get knocked down," he says. "We'll look back at that as pretty archaic."
Cooke adds, "Consumers are demanding so much more online. They really expect to get relevant information, a good journey, and for the site to look great. When you have a Website that changes a design three times a year, and a retailer changing their site every three years, you get this massive disconnect versus what customers are experiencing every day on Facebook and what they get on their retailer's Website."
In 2010, Cooke and Qubit launched a platform to help marketers solve the "code freeze" problem, allowing them to more easily change their Websites on the fly.
The platform allows the marketers to rebuild specific parts of their Websites without needing to know code. The platform also taps into data wells such as IP address databases, weather databases and
consumer behavior databases.
"We're helping Web business move toward the real advantage of the Web: respond quickly, retail fast, and make changes to your site," Cooke argues.
And he's got the numbers to back that up. Well, predictions at least.
Global retail value in the U.S. in 2013 is projected to be $585 billion, with 21% of that value being created over the Christmas period (November - December). In other words, November and December are expected to draw in $11 billion more revenue per month than the average month throughout the year. This data comes from McKinsey & Company.
Using the Qubit platform, which is expected to be updated soon, the estimated amount of uplift on top of that extra $11 billion a month is 6% (Qubit estimates). So instead of generating $111 billion during the period leading up to Christmas, the online retailer would generate $117.7 billion.
To achieve that lift, do CMOs really have to become data scientists, like Cooke suggests? I'm not so sure.
Cooke gives two examples. His first example is something I've seen on Websites often, especially on Amazon. "If you were on the Website and looking at products then came back three days later, it would show you the most popular products that you were looking at recently. That's what we call real-time targeting in the Website," he says. That's not groundbreaking -- we've all been seeing that for awhile now -- but it's effective nonetheless.
His second example involves taking a region-specific event, like a snowstorm, and combining it with other data sets. "A very important part of Big Data is the ability to join data sets together," he says. "While it's not a large data set in that context (a snowstorm), the principle of joining sets together in real-time is a Big Data principle."
With the first example, I don't really see how it means a CMO is a data scientist. But the second example is one I like. The CMO would be able to make changes to the Website for consumers in the area, doing anything from offering discounts to simply putting out a reassuring message saying that delivery is still possible.
Would the CMO really have to be a data scientist to achieve that,
though? I guess…in some small way.
The CMO would use data about the snowstorm and data about the consumers in the area (Are they still shopping? Are they on the site but not buying?) to make a decision. To use Cooke's terminology, by mixing those data sets and taking action, the CMO would be applying Big Data principles. But simply applying Big Data principles does not equate to being a data scientist -- in my eyes, at least. Cooke's example just doesn't seem significant enough to warrant a full-on transformation from CMO to CMO/data scientist hybrid.
To be fair, I asked Cooke to dumb it down for me. I wanted to understand how Big Data fits into an online retail CMO's picture. I also wanted to know how they would have to act as a data scientist in certain situations, so he gave me a pretty basic example.
What do you think? Do CMOs -- regardless of what they are marketing -- need to double as data scientists? If they don't
now, will they soon?
*Editor's Note: Media6Degrees has been renamed Dstillery.