Why are there so few great products?
Large companies can reliably turn money into solutions for most of their problems. But few can transform money into great products. Why is that?
There are objectively few great products in the world. The average piece of software is slow, buggy, and unpopular.
I've always wondered about this. Large companies seem to be quite good at reliably turning money into effective solutions when it comes to marketing, sales, recruiting, and many other problems. But somehow building great products is an exception — money and resources don’t seem to help that much. What makes it so difficult? And why are large companies particularly bad at it?
Products have objectively little surface area of significance. A product will only ever have one home page and one onboarding experience, for example. As a company grows, the ratio of important surface area per employee goes to 0. When you have something important that an increasing number of people care about, pressure accumulates. This battle to dictate how important real estate is used happens within product and design, but as a company grows demands increasingly come from other functions like sales and marketing. And those requests often aren't in the interests of users at all.1 Users don't really care about your startup's ability to grow, for instance.
The dynamic creates an environment where the typical path for company leaders is to meet in the middle, to compromise, and to delegate. These are also the traits that lead to a deteriorating product. The best product CEOs and product leaders — Steve Jobs, Elon Musk, Brian Chesky — are the opposite. They’re in the details, opinionated, and top-down exercise judgement. Great products are never designed by committee. 2
Product decisions, especially early ones, matter so much because some subset have high path dependence. If you run a disappointing marketing campaign with an agency, you can fire the agency and work with another. But some product decisions are like early branches of a tree. The latest iPhone is remarkably similar to the original one introduced seventeen years ago. The core Google search experience has barely changed in over two decades. It's often only apparent after the fact what decisions have high path dependence, and so you need to have a high-caliber thoughtful product culture from the start and not lose it, or you might not be able to course correct later.
Features have an innate tendency to get flushed out over time. Even if a feature isn't a hit, the excuse to "Make it a bit better and it might work" comes so naturally and is rarely questioned. The accumulated bloat that results and the reality that mediocre features are rarely deprecated is how great products lose the original simplicity that made them compelling in the first place. Everything you add to a product has a complexity cost and given the human incentives at play, subtraction is difficult. Products evolve to become easier to use only in cultures that encourage ruthless deprecation and re-imagining things that aren't working. In contrast to Apple, which quickly deprecated the Touch Bar on MacBooks and Force Touch on iPhones after going all-in, Google somehow believes users want to have all of Google Messages, Google Chat, Google Meet, Google Voice, Google Duo, and Google Allo.
As organizations mature more decisions tend to be data-driven. While analytics is a helpful tool, analytics should never be the thing that's most important. The problem with obsessive data-driven cultures is the inevitable end state of landing in a local maximum that blinds you from bigger opportunities. Any feature that's been A/B tested to death 100s of times will naturally outperform a reimagined approach because the latter hasn't undergone any product evolution. This is why Meta, being so obsessed with optimization on their big blue Facebook app, missed the novel UI paradigms that Snap and Instagram pioneered.
A data-obsessive approach is also how you end up with something that lacks any semblance of cohesion and character. The big blue Facebook app, shaped through A/B tests, performed better and better on metrics until suddenly it didn't. And at that point it became too rigid to substantially change things.
Getting the details right is what separates many of the best products. It's why the startup employee stack looks shockingly similar with Stripe, Notion and Linear almost always making an appearance. Are there no alternatives to these products? Of course there are — there are 100s of note-taking and issue tracking apps — but Notion and Linear won the hearts of users, not by having more features, but by being more enjoyable to use and by nailing more of the details.
In practice, it's difficult to sustain a product culture that is obsessed with details because arguments for new functionality always sound more compelling. Whereas a new feature means new stuff for marketing and sales to pitch, what's the argument for making something a tiny bit faster or a bit more fun to use? The right interpretation is that those improvements compound in ways that will eventually help a company set itself apart on a dimension that is indeed *more difficult* to replicate that a feature. For example, imagine how difficult it would be for Jira to become as fast as Linear, or for Adyen to become more developer friendly than Stripe.
So why aren't there more exceptional products? Product decisions are particularly difficult to delegate, business functions have emerging anti-user incentives, path dependence is high, analytics can be a trap, bloat is difficult to avoid, and prioritizing product details is unpopular. The default trajectory for a growing business is product decline. Startups have to actively fight against the many forces that will over time make a product mediocre.
I'm not saying these are necessarily bad decisions on the whole, just that they are not in the interest of current users.
There are numerous stories about this in Steve’s and Elon’s biographies. But I thought this interview with Brian Chesky and Lenny captures the contrast particularly well.