Among lean startups, a common mantra is “launch and learn.” Braden Kowitz of Google Ventures Design Studio says that’s precisely the wrong approach and suggests seven strategies that he says will save time and lead to a better product.
If you’re running a lean startup, “launch and learn” is undoubtedly a familiar mantra. But launching a new feature can take weeks or even months, and for a scrappy startup that’s a potentially make-or-break issue. Our design studio works with dozens of startups each year to help teams define their products and features. Through the process of doing this over and over again, we’ve collected a time-tested toolkit of methods for learning that are cheap, fast, and perfect for startups to find those crucial mistakes earlier and then adapt their plans more nimbly. The result is almost always that they ship better products and do so even faster.
Most teams think they need to build an interface that functions and looks real before showing it to customers to get feedback. Nope. It turns out that if you string together a few simple mockups with clickable hot-spots, you can still get great feedback in a fraction of the time. We’ve done this with companies like HomeAway, AVOS, and Duo Security by designing a few screens in a flow and then building a clickable version, using basic consumer software tools like InVision or Apple’s Keynote.
At first I thought these prototypes would be too rough to be useful. But time after time I’ve seen customers engage with click-throughs like they’re real products, and that helps you learn if the designs are working. It’s a great method to use before engineering starts to build a design.
Instead of working in a vacuum, gather data to use as fuel for designing your product. Specifically, go out and find the people you think will use your product and talk with them about the problem(s) you’re aiming to solve. I know you’ve heard this a hundred times. Customer interviews are like flossing — everyone agrees it’s good for you, but it’s hard to build the habit.
It’s easy to get hung up on the details: How do you find people who will talk with you? What do you talk about? Relax. User researchers have been doing this stuff for decades, and there’s a wealth of knowledge about how to do it quickly and accurately. For starters, you can write a short survey called a screener to help you recruit the right people to talk to. Then, create an interview script to help guide the conversation.
If you want to know more, we created a research guide with plenty of tactical tips for finding and interviewing customers. Now you have no excuse. Get out of the building! (Then come back — there’s more good stuff below.)
You can quickly see whether customers will engage with a new feature by launching just the first part of it. We did this with CustomMade, a startup that lets people order custom-built products. Our idea was to let visitors save others’ projects for inspiration. But instead of laboriously building the whole feature, we just launched the first button. When we observed a huge number of visitors clicking the button to access that function, we knew we were onto something and built the rest of the feature. After a few changes like that, we saw a 3x increase in engagement. For more on fake doors, see Jess Lee’s excellent talk.
When teams design a new product, they come to the table with all sorts of assumptions about the competition. It’s easy to look at another product and have an opinion about which parts are valuable and which parts are broken. But if you guess wrong, you might just copy a bunch of functionality that your customers don’t actually need.
So we like to think of competing products as free prototypes. We watch customers use these products and learn very quickly which features are loved, unusable, ignored, or hated. With this knowledge, we can make better decisions in product design, marketing, and sales.
Surveys are a tempting way to learn from the comfort and safety of your office chair. But designing a good survey is surprisingly tough. Whenever I talk with survey scientists, I’m overwhelmed by all the ways you can screw up a survey design and unknowingly get bad (read: useless) data. So when we run surveys, we stick to a pattern we know works well.
We put the survey as close as possible to the behavior we’re trying to study. For instance, if we’re interested in why a customer picked one of our pricing plans, we’ll ask them with a small pop-up survey in the moment, not an email that might get read days later.
And we rely on open-response questions that let us hear directly from customers. You’ll learn more from reading 100 short responses than knowing that 32 percent of users chose option B in your survey. Here’s more on micro-surveys.
Prototype with Real Data
Clickable mockups are a good first step, but you can learn even more when you build a prototype that integrates real data. You might be tempted to start building the actual product at this point. You might even call that work-in-progress a prototype. But it’s not. Building a real product always takes longer than you think. If you really want to learn fast, build a true prototype – one that you’re not afraid to throw away.
When we were designing coupon pages with RetailMeNot, we needed real coupon data in order to evaluate our designs. So we built a prototype in two days. It was buggy and didn’t have many features, but it was just enough to get useful feedback from customers. And it was good we did, because it turned out that half our ideas weren’t working.
We iterated three more times, building prototypes and showing to customers, and were able to get to a design that improved both usability and click-through rates. Few startups build true prototypes, but it’s an immensely useful way to learn fast.
Go to wherever your customers are, and watch them actually use your product. I know that sounds like common sense (or it should). But it’s too easy to think we know our customers from all the meetings, phone calls, and reports we’ve read about them. To deeply understand how people actually use our products we need to go to where they work, where they play, and where they live.
Recently, we were working with Foundation Medicine to improve their clinical cancer genomics reports. So we decided to visit oncology centers, watch how doctors used the reports, and see what we could learn. We were surprised to discover that the reports we’d worked so hard to design were often received by fax. Tiny text was hard to read and all color information was lost. It was an easy problem to fix, but we only noticed it through a site visit.
Being a lean startup means that we should first consider all these ways to learn, and then pick the fastest, cheapest method. I’ve listed seven methods that we’ve found work well at startups, but there are plenty more out there. Once you start looking, you’ll be surprised at the variety of ways you can learn incredibly fast, saving you and your team precious time and money (and heartache).