Back
NoteMay ’252 min read

Arc published its novelty tax

The Browser Company stopped building Arc and shared the numbers behind it: its signature features were used by about 5% of daily users. What dense tools owe everyone else.

Typeface

On May 26, Josh Miller wrote a letter to Arc’s members explaining why The Browser Company had stopped building Arc. Most of the letter is about the new browser, Dia. The useful part is three numbers about the old one.

5.52% of daily users regularly used more than one Space. 4.17% used Live Folders. 0.4% used Calendar Preview on Hover.

Those were the features people wrote about when they wrote about Arc. Miller’s name for the problem was a “novelty tax”: too many new things to learn, for too little reward. He was blunt about it: “We knew. We were just in denial.”

Last year I wrote that a new interface earns nothing for being new. Every new concept is a bill paid up front, and it only makes sense if the old job gets done better. Arc is the rare case where a team counted who paid the bill. About one daily user in 20 used the features that defined the product.

Miller also wrote that Arc’s early retention was strong and that its usage looked like a professional tool rather than a browser for everyone. I don’t read that as a failure of the features. I read it as a product that charged the tax to everyone and paid out to a few.

I’ve spent a lot of my career on tools for experts, and on Uber’s experimentation platform the first release cut every chart to get the core flow working. The charts came a few weeks later, for the people who wanted them, and nobody had to learn them to turn a flag on or off.

A power feature should cost nothing to the people who never go looking for it. That’s the pattern I’d take from Arc’s numbers:

  • The default path has to work for someone who learns nothing new.
  • Power features show up when someone reaches for them, not before.
  • Each new concept gets counted. If fewer than one in 20 use it a month after launch, it either moves out of the default path or goes.

The cost of that pattern is that your best ideas are harder to find. Some of Arc’s charm was that the strange new things were right in front of you. Hide them and fewer people discover them.

The same letter gives Dia’s numbers: chatting with your tabs is used by 40% of daily users. That’s one feature, measured early, and announced next to the 5% figures. Whatever happens to Dia, the team learned to count. Most product teams don’t publish their adoption numbers, and I’d like to see more of them do it.