I've tested multiple startup ideas over the past year. So below is key learning I got from my failed attempts for anyone else who tries to do the same. I hope it saves you a bunch of time.<p>Remember, the only thing you can control when just starting up is the rate of iteration. But even if you iterate fast enough, you still have to judge your efforts based on the right criteria.<p>In fact, testing startup ideas can be tricky. You need to test your riskiest assumption first and choose the right metric to determine success/failure.<p>Contrary to a popular view, the number of customers received is not a reliable metric, as these can be casual customers.<p>From my experience of testing 10+ ideas, the most important criterion for building a scalable business is a conversion rate (CR).<p>You should be measuring CR not from a site visitor into a customer, but from a post/ad/message view into a paid user. This is a more reliable measure of how many customers you can potentially get in the future.<p>Back-of-the-envelope example<p>If we're to make a product offering internships to students, we shouldn't start from running Fb ads. It's better to choose a specific university and advertise on its forum/group chat first. Another approach could be going to the campus yourself and recruiting potential trainees manually, practicing the art of persuasion.<p>If we don't see any meaningful effect (conversions vs reach), it means that we'd see the same at scale. In other words, our product value prop is unconvincing at best.<p>What should I remember?<p>When testing your startup's riskiest hypothesis, you need to forcibly narrow your reach to measure and test CRs.<p>Because if you don't do it, it's likely that during the testing you'd cover almost the entire audience available in that channel. At this point, any further investment of money or time won't produce any proportional increase in customers, because you'll be hitting into the same people over and over again.