John Foreman, chief data scientist for MailChimp.com, learned an important lesson from the 1992 movie Sneakers, a lesson marketers might do well to learn, too. There’s a scene in which Robert Redford and Dan Aykroyd’s characters struggle to hack a keypad to open an office door. Eventually, Redford just kicks it in. “The takeaway there is you don’t have to hack the keypad if you can just break down the door,” Foreman says.
In his new book, Data Smart, Foreman argues that many businesses are metaphorically hacking keypads with their approach to analytics and data. We spoke with him to get some clues as to when that’s needed, and when it’s time to go all Robert Redford on the door.
What are your best pieces of advice for marketers struggling to turn all of their data into insights?
Businesses, especially businesses getting into the analytics game for the first time, should really scope out what they want to do before they try to do it. Rather than just reading a bunch of news articles about what other people are doing, actually look at your business and the problems you’re having and the places you can see the most gain from using analytics. It’s best to pick one problem and try to solve that. What I do in my book is go through a bunch of techniques and the types of problems they solve so people can know what’s possible. If you know what’s possible then it’s easier to pick out a problem to solve.
What would you say are some of the most common issues companies run into with data?
Some people run into issues with trying to build the perfect solution when often an 80% solution will do. You’ll see companies hire some Ph.D. who’s extremely proficient in some type of analytics and they’ll try to build some gold-plated solution that would’ve gotten them recognition in academia or something. That’s not terribly useful for a business. Rather, you should build an analytics solution that can live on even if the person who built it gets hit by a bus.
At what point is it is an analytics model too complex?
People should try to avoid complexity, specifically the kind of complexity that requires that they care for and feed their data-driven models all the time. It becomes such a real hassle that no one wants to use it and then you turn it off and you’ve wasted your time. You have to balance ease-of-use and complexity with whatever you’re getting out of it.
Do you think marketers put too much or too little emphasis on performance of their data?
I’ve seen marketers run into issues with performance a lot where they think, “I’ve got to get bigger and better performance out of whatever I’m building with this data.” It’s often important to step back and figure in what context are you going to be using this model or data tool. Is this the kind of thing you need to run in one second or is it the kind of thing that can take an hour? Can you change your business so that taking an hour to run this thing is acceptable? I think with analytics projects in particular people just tend to spin their wheels trying to make them perfect when often that’s not necessary. You get a better product in the end by releasing it into the wild faster and learning from how it does than you do from sitting there making it so good that you’ve gone bankrupt or something.
What other advice would you offer marketing people in terms of handling data and analytics?
Companies will often segregate the data or analytics people from the marketing folks. You’ll get people that are in leadership that will come up with a problem that they’ll want the analytics folks to solve, and they’ll just kind of throw it down to the basement. That can often be the wrong approach because the analytics people will solve the exact problem they’ve been told to solve, but was that the right problem to solve to begin with? You enter a situation where a problem that’s been given to analytics is one that’s been interpreted by management and leadership who might not understand analytics at all. They cast the problem in a light that might actually obscure the original goal. I encourage people to include the analytics team as early as possible.
