I like principle #3 of the principles from the SAFe framework: Assume variability, preserve options. It’s a very useful guiding principle for anyone creating products, especially in tech.  

Customers are very picky. They have very specific needs. Traditionally, companies used to pay a lot of money to research those customer needs at length. The idea was that if you’re going to invest a lot of time and money into making an awesome product, you have to be sure that you are addressing a customer need, and addressing it as precisely as you can.  

About 20 years ago, leaner and nimbler companies realized there was a cheaper and faster way to build products addressing customer needs even more precisely than cash-rich companies willing to dish out on research could ever do. Companies that tested multiple product design options in the market were able to more accurately understand what customers wanted.  

It’s useful to think of products as tests of hypotheses about the market. For example, a company building a cool new portable kayak is testing a hypothesis that there is a market need for kayaks that are more portable than the ones that existed before. You test that hypothesis by trying to sell the kayak. In the most basic terms, if you manage to sell it, your hypothesis was right.  

Assuming variability and preserving options means you should constantly maintain several different hypotheses about the ideal kayak design before locking in on one of them. For example, you should have a longer kayak, a shorter kayak, a heavier kayak and a lighter kayak, then see which one best matches your customer’s needs. Once you find out which kayak your customer preferred, lock in on that one, and try even more specific variations on design.  

This method gives you more solid grounding than research ever could: you have proved that your product matches customer needs using the most solid evidence possible: the customer actually used or even paid for your product.  

Source: SAFe 

In this drawing, the x-axis is time, the arrows pointing to the right are different designs being tested, and the yellow lines are learning points. When you take this approach, it’s important to consistently stop to learn which designs worked and which designs did not. Weed out the worst performers, then repeat.