advertisement
advertisement

Google’s Algorithm Guru: Design’s Fine, but Testing Rules

Google Fellow Amit Singhal explains the company’s “scientific approach to design.”

Google has always valued engineers over designers–data over design. Long before the infamous “41 Blues” incident–where Google tested 41 shades to decide on the “best” color–the company has practiced a design-by-testing approach, relying on voluminous analytics rather than an organically developed user experience.

advertisement
advertisement

Do Google engineers and designers ever work together?

We spoke with Amit Singhal, Google Fellow and long-time head of the search ranking and algorithm team, who gave Fast Company readers an inside look at Google’s development process. Singhal said Google takes a “scientific approach to design,” one that focuses on rapid experimentation and testing–and a synergy between designers and engineers.

“Our answer is to put these two sets of people in very close proximity,” he says. “Let them hear each other’s problems because bringing a list of problems to a meeting once a week is not going to help anyone. We put [engineers and designers] next to each other, and they start thinking aloud.”

advertisement
advertisement

Singhal discusses how the team designed auto-complete, the search feature that predicts queries as you type.

“One of the problems we faced early on in this design process was whether to have the entire query be one color or give a grayish version of the completion,” he explains. “This was a design problem. How about we use probability to pick the grayness of the text? If the probability is 100%, then you might as well make it black. If the probability is low, you want to make it more faded.”

This approach is very scientific (not to mention oddly familiar to 41 Blues), and Google wouldn’t have it any other way. To Singhal, the process must begin with a hypothesis: “Should the probability of completion govern the density of gray?” This hypothesis was tested internally and externally, which helped the team decide whether or not they were “hurting the user experience.”

advertisement

Ultimately, the gray auto-complete feature was never implemented. Although designers felt it would be a very poor user experience, Google seems to have based the decision off the results of its testing–not subjective opinions.

“Everyone can use their gut to say whether it would be great or not,” Singhal says. “At the end of the day, if it doesn’t benefit our users, it is not good.”

I asked Singhal about the drawbacks of such an approach. In the case of 41 Blues, the team did endless testing on dozens of shades, settling on the one that received the most clicks–unsurprisingly the average of all the colors, and the blandest of the blues. Wasn’t working with this scientific approach–experimenting with so many users–the equivalent of having too many chefs in the kitchen? Was Singhal worried his team would just end up making a bland dish?

advertisement

“That’s a very genuine concern, but it’s somewhat misplaced. There are not really too many chefs in the kitchen–it’s the same team of designers talking to the same team of engineers, hypothesizing multiple paths that one could take” he answers.

“Testing gives us confidence. We have systems where we can test one change or 41 changes on a large number of users, observe how they interact with our ideas, and refine our ideas to make them much better.”

[For an alternate take, check out Google Equates “Design” With Endless Testing. They’re Wrong]

advertisement

[Top image by fdecomite]

advertisement
advertisement

About the author

Austin Carr writes about design and technology for Fast Company magazine.

More

Call for Most Innovative Companies entries! Apply now.

500+ winners will be featured on fastcompany.com. Final deadline: 9/23.