Building in Public, and Why We Shut Down the Privacy Policy Decoder

We don’t want to pretend every idea is a winner. Sometimes you spend time with something and realize it’s weaker than you first thought. The market isn’t there, the economics don’t work, the technology is too difficult, or the product starts to make less sense once you get into the details. Those are the usual reasons people give for pivoting or shutting something down.

That wasn’t really what happened with the privacy policy decoder. The idea was pretty simple: take a privacy policy, turn it into plain English, and help people understand what they were actually agreeing to.

There was nothing obviously wrong with it. It sounded useful, the technology was interesting, and we probably could have built it. I can imagine people using something like it. But the more we thought about what the product would actually be doing, the less comfortable we became with it.

Privacy policies are legal documents. As soon as you start translating legal language into plain English, you’re making judgment calls about what matters, what a clause means, what can safely be simplified, and how certain you are about the explanation you’re giving. That made the idea feel a lot less straightforward than it had at the beginning.

If the tool says a company does or doesn’t collect certain information, or summarizes a policy in a way that gives someone the wrong impression, that isn’t a minor error. “Mostly right” starts to feel like a pretty bad standard when someone may be making a decision based on your interpretation of a legal document. That was the point where the decision stopped looking like a normal startup question.

We weren’t saying there was no market. We weren’t worried that it would be too expensive to build, and we didn’t think the technology was impossible. We just didn’t like the amount of trust the product would ask from the user unless we were extremely confident we could justify that trust.

It made me realize there was another question we hadn’t been asking clearly enough. Entrepreneurs spend a lot of time asking whether an idea is viable. Can we build it? Will people use it? Can it make money? Is the market large enough? Those are all reasonable questions, but something can be viable and still not be something you want to build.

There are products you could probably make work, and maybe even sell, that still don’t fit the kind of company you want to build or the kind of responsibility you want to take on. That was where we landed with this one.

There was another reason the idea had been appealing in the first place. The privacy policy decoder was a very good demonstration of what the system could do. You could give it something dense and difficult, have it pull out the important parts, and explain them in a way that was much easier to understand. That makes for a good demo, but it doesn’t necessarily make for a good product.

So we shut it down. I think this is one of the parts of building in public that is easy to leave out. People naturally share the launches, the progress, the new ideas, and the things that worked. But some of the more important decisions are much less exciting. Sometimes you look at something that probably could work and decide that you still don’t want to keep going.

That’s what happened here. The privacy policy decoder wasn’t killed because the market disappeared or because the technology failed. We just became less convinced that it was something we should put into people’s hands, and that felt like a good enough reason to stop.

What hasn’t changed is our view of the problem itself. Most people can’t realistically be expected to read and decipher every privacy policy they agree to, and as a result, people routinely agree to things like having their data collected, shared, or sold without really understanding that they’ve done so. We still think that’s a problem, and one the tech industry needs to do a much better job of fixing. We just decided that our privacy policy decoder wasn’t the right way for us to fix it.