The Greatest Power

At Squarespace, every employee can comp two websites, for life. They can comp sites for their own use, or for friends, or they can even end the comped status on one site and move it over to another.

What this means is that you have a company full of dogfooders: most employees are actively using Squarespace for personal sites or projects. As a result, the platform grows, rough edges get smoothed over, and needed features get implemented faster.

But that's not crazy. It's just good sense. When your team is invested in your product, you get a better product.

What would be crazy would be not to give away the product to employees, but to give away customers to employees. Imagine it: every engineer had a thousand users they could roll out changes to at a moments notice, no approval or overhead necessary. Imagine they could dream up a feature or an improvement or a style and see how people react to it within the hour. Wouldn't that be neat?

Of course, that's great power—and, like Ben Parker said, with great power comes great responsibility. But given that at many companies, any engineer already has the agency to deploy breaking code to master, we should mitigate irresponsible behavior by building good tests, stressing proper employee education, and working hard to build cultures that highlight responsibility, vigilance, and good choices—not through shackles and red tape. And while perfect performance is an enviable goal, it's probably better to see experiments fail on a smaller group of customers than on the entire userbase.

So: what does a company need to enable such a radical idea? It comes down to a few things:

  • Partial deployment/code splitting. Relatively speaking, stylistic changes and frontend-only experiments are easy. But for more advanced experiments, or anything requiring server changes, you’re really looking for the ability to deploy a branch on an individual node. Even frontend-only changes benefit from code-splitting: keeping multiple variations of markup and styles and conditionally rendering them increases page loads, adds bloat, and inflates load times. And slower pages are bad for business.
  • An easy-to-use and flexible reporting framework (and education on how to use it).There’s little reason to push a variant of your product to a subsection of your user base unless there is 1, some kind of metric you’re hoping to affect, and 2, a plan for measuring that change. Oftentimes, those metrics may be things that aren’t already being measured—having a flexible reporting framework in place that empowers engineers to add event or behavior tracking on a whim is key to making their experimentation effective.
  • Empowerment and education. Let’s face it: an anyone-can-test-a-variant plan is going to result in some big bets that don’t go over well. For a consumer product, that’s probably fine: with good reporting in place, failing experiments should be easy to identify and terminate. But, reducing the frequency and severity of these misses is possible with a good, well-rounded education and open channels of communication. If an engineer wants to learn about user experience, is there some way she can do that? If an engineer would like to bounce around some ideas about design, is there an obvious way to do so? What about statistical significance? Building cross-functional teams is a good start; formalized cross-functional training is a great way to take things to the next level.
  • A bottom-up culture. Who defines the direction of the product? The upper management, the ICs, some combination of both? This approach will work best at a company that inspires its employees to creatively and proactively find ways to improve the product, and gives risk takers and exceptional performance the recognition and appreciation they deserve.

Clearly, there are some ground rules that need to be in place as well: user data isn’t to be tampered with, experiments need goals, mission-critical aspects of the application may merit more oversight. But let’s dream big, and imagine for a second that empowering your engineers in this way was completely possible. What kind of things would they come up with?

I’ll give you a hint: there’s no way to know. If all of your employees and co-workers think the same way you do, you’re doing something terribly wrong: as former CEO of McDonald’s Ray Kroc once said, “I believe that if two executives think the same, one of them is superfluous.”

The beauty of a system like this is that it allows for the innovations too radical to make it out of boardrooms to skip that step and make it to users, where their impact can be measured directly. Sure, maybe some of those innovations are miserable ideas. But maybe some will become core features of your business. Sound ludicrous? There’s only one way to find out.

Timothy MillerffComment