Still, overcoming the cold start problem is deceptively difficult. Chen offers the example of Tiny Speck, the gaming company that would go on to become Slack. Tiny Speck had everything going for it: a star team, an exciting launch, and $17 million from respected investors (including Andreessen Horowitz, where Chen works). It also got plenty of people to try out the game, called Glitch. The problem was it couldn’t get people to stay.

What was the difference between Glitch and Slack? For one thing, Slack’s timing worked: It had anticipated distributed workforces and a need for text logs. But it also benefited from tiny, atomic networks. People joined in teams, and as those teams familiarized themselves with the product, they were likely to continue using it. (The magic number, according to Slack, is when a team has exchanged about 2,000 messages.) Later, the company grew by encouraging companies to adopt Slack across the entire workforce, knitting together many teams on one unified workplace tool.

Of course, network effects alone cannot explain a startup’s success or failure. Slack was just one of many workplace communication apps with a similar idea; not all of them had the same success. “For every successful launch like Slack, there are many more that are failures,” Chen acknowledges, “and they usually stumble right at the start.”

Both The Cold Start Problem and Anticipate Failure offer autopsies of several failed companies, but that can still leave a reader scratching their head. Chen points out that some startups achieve network effects because they provide services that are free, convenient, and easy to use