508 words | ~3 min
Reading this piece by Simon Law, on the creative challenge of programmatic and adaptive media, this thought stuck out:
We’re all trained to find the best answer – both at agencies and in marketing departments. But the best answer is inherently singular. It doesn't include a set of 'maybes' – so we need to change our attitudes, too.
I worked with and for Simon for three and a half years at Fabric, and it's a principle we put in practice as an agency, testing and measuring creative ideas and scaling up the ones that worked.
But 'designing for "maybe"' strikes me now as a good foundational principle of building analytics functions, as well as creative departments. There are only two components of an analytics team: the people, and the technology. We need to design for 'maybe' when assembling both.
People first. Everybody recognises the need for expertise - you can't really bluff your way in advanced statistics or data integration. But we should be hiring people who approach their jobs as collaborators and inventors, not just as experts. Being an expert is a defensive posture (I'm an expert only insofar as you're not, and I'm an expert in a particular thing...); being a collaborator and an inventor makes you the kind of person who can be approached with a new problem and look for a way to say 'maybe we can...'.
And then technology. More and more of the challenges we face in analytics are problems of technology - its capability and its scale, but also its ability to help us respond to uncertainty. There are a lot of good, powerful marketing technology software products around these days, with serious amounts of data behind them; but most of them exist to make doing certain things more intuitive, user-friendly, foolproof. They offer a set of definite 'yeses', lots of 'nos', but very few 'maybes'. We need technology and software that we can tinker with, recombine and plumb together in new ways to answer original questions - Lego bricks, not works of art.
At Maxus our analytics technology stack is designed for maximum flexibility - scalable on-demand big data warehousing, a lot of SQL, a lot of R and a bit of Python for analytical programming, and build-your-own visualisation layers using Tableau, Shiny and PowerBI among others. It's not always pretty, but it lets us say 'maybe' a lot more, and 'no' a lot less, when we're asked to help solve a problem we haven't encountered before, and get solutions working in days and weeks rather than months. And, of course, we put them in the hands of people who see 'maybe' as a challenge.
If you're buying analytics products or services, look beyond the elegant user interface. Most analytics tools, behind the scenes, involve a lot of people prodding scripts. Ask how open they make the underlying data; ask how locked the development roadmap is; and ask whether they will let you answer 'maybe' to an interesting question you haven't thought of yet.
# Alex Steer (05/08/2015)