For the past 20 years, NASA’s been using VR to train astronauts for spacewalks — you know, those things where they have to dangle off the end of a robotic arm and move something the size of a bus. NASA’s journey with VR has been intense, but over the years they’ve figured out a lot that we can learn from.

Here are three big takeaways for the future of virtual reality from NASA’s VR Lab.

First, success is going to have a lot to do with finding those spaces where virtual reality is the best solution for a problem. The thing about space is that it’s only in space. You can’t easily recreate certain circumstances on the ground, like becoming detached from the International Space Station. Unless you use something like virtual reality. And that’s exactly what the lab does — they have a simulation that lets astronauts practice flying back to safety in the unlikely event they start floating away.

Second, expect iterations. The VR lab has been through about 6 different headmounted displays and 7 versions of their graphics engine, but have also iterated countless times in between. So keep in mind that even though everyone’s excited about the first wave of commercial headsets about to hit the market, just remember the key word there is first.

Third, consider the customer. That’s a good rule in general. For the VR lab, the relationship between them and the astronauts, who really, are their customers, has helped them create something that’s actually useful and, in a lot of ways, it’s helped them survive things like budget cuts. Early on, latency was a huge problem. The lab had to pick between low latency or detailed graphics. They sought input from astronauts who said they’d rather have details, and the latency didn’t matter much because in space they moved their heads slowly anyway. That basically means that while the rest of the VR industry was struggling with lag, they could focus on other innovation and wait for the tech to catch up.