A few years ago, I was lucky enough to be a student and an employee of Milton Glaser. My mind was a sponge to his sage advice and philosophies, which, like many other legendary faculty in the MFA Design program at the School of Visual Arts, centered on an ontology that a designer should experiment (mostly in private) and present a single version of their work. Data does not play a major role in how he designs and he smirks with pride to admit he has never used a computer.
When Milton Glaser was born, the Empire State Building wasn’t finished yet. Sliced bread was literally only a year old. His career developed at a time when big data, analytics, and design research were luxuries too expensive for most to afford and were therefore impractical. Even the best design agencies could only perform market research at a barely-significant sample size. His methodologies were born out of an era where, in a vacuum of proof about the efficacy of a designer’s work, one needed instead to speak confidently and persuasively. Stefan Sagmeister and designers from Pentagram have echoed the “one logo” approach. But these figures are confident, sharp-witted designers with a career to back up their bold take-it-or-leave-it view.
But my experience has been a bit different. Data science has crept into the most fast-paced fringes of the design world and is quickly spreading inward. I wonder if iconic design studios will soon be required to prove the success of their ideas with metrics or ever be flexible enough to change their final designs in realtime. And as much as I love the idea of testing, I can’t help but wonder if some of my favorite designs from Glaser’s studio would have been forced to alter in an age of aggressive user testing.
The design leaders of tomorrow, the next Milton Glasers, will be expected to design within highly measurable, highly volatile constraints. When I design, the work’s performance is tracked down to the millisecond and pixel. I know how long someone looked at it, whether they were drawn in enough to click it, and whether it had successfully convinced them to to take action. I’m encouraged to release versions of my designs that aren’t a sure win. In fact, the idea of a “sure win” with any design seems like hubris to me now. The client (and the public) don’t get one version. They get dozens. Designers now have access to A/B testing platforms and machine-learning systems that are always comparing performance and serving designs that are most effective.
In some cases, this performance can even be tailored to an individual level. Person A will see version 4 and 5 of my design, but never 3 because their behavior pattern suggests it wouldn’t be effective.
This process seems antithetical to the advice I’d received from mentors. It suggests that the “one logo” approach no longer holds any merit in a world of big data. And perhaps this is ok. Isn’t it best that designers validate their ideas with actual information from their audience? Will the next Milton Glasers be known for their perceptibility and taste, or their ability to detect the signal in the data noise?
There is, of course, such a thing as taking this data-minded approach too far. Marissa Mayer, former steward of the Google search page, gave several talks over the years on Google’s obsessive data-driven design process, including a story about testing more than 40 shades of blue to get the search button right. Mayer went on to become CEO of Yahoo! where she thumbed her nose at the design world boasting about her rebranding of the company over a weekend without consulting any design agencies. The result was an off-tone Revlon-inspired new logo that shook confidence in her decision-making as a leader.
A few points will bring us to a middle ground:
♦ Design should drive, not data. In my experience, designers craft questions and use testing to validate hypotheses and better understand the context of their work. It doesn’t mean that a poor performing design always gets killed.There can be reasons to disregard the data.
♦ Data isn’t for amateurs. Data can answer questions for designers, but designers must be asking the _right_ questions first. Nate Silver’s book, “The Signal and the Noise,” talks about the pitfalls of overconfidence in data. Biases, heuristics, and general naivete are all real threats to basing your design on testing. Anyone using big data must have at least a cursory knowledge of statistics, forecasting theory, and different methods of testing.
♦ Stay open. This is more of a general philosophy, but if you’re the type that prefers cold hard facts, consider the impact such systems would have on some of our favorite designs. If you’re wary of analytics, try using new research methods to learn about the environment you’re trying to affect, and be ready to admit when it proves you wrong.
Design and big data are just getting acquainted. One thing seems clear: data is here and it is forcing designers to resolve artistry with loud, immediate feedback.