How I Think About Growth

1. The prevailing logic

Behavioral economics, ads, and data form the prevailing logic of the consumer internet. Behavioral economics provides the most parsimonious model for understanding and influencing human behavior online — its principles are embedded in every popular consumer platform and many business platforms. In the smallest of nutshells (would that be a hazelnut?), if you want someone to do something, make it easy. And "free has a quality all its own." Zero is a special number so special that free is the price we expect to pay, even for high-quality, hyper-useful internet experiences.

How do companies make money if their products have to be free? Ads. The saying is old, but old sayings have lasted a long time and are therefore likely to last longer: if the product is free, your attention is the product. This business model was pioneered by London penny newspapers in the 1800s, when publishers realized they could make more money selling papers at a unit-economic loss than they could trying to profit on each paper sold. When they reduced the price to a penny, circulation increased significantly, which meant they could charge more for each advertisement, because each ad would be seen by more people, and the increased earnings from the advertisement side more than made up for the unit-economic losses on the paper side. Data sharpens this further. It increases conversion rates on those ads, which makes ads more valuable to the companies running them, which means the newspaper can charge yet more per ad. If you consider that the return on an ad is the number of people who see it multiplied by the number who take the advertised action, divided by the cost of the ad, then it becomes plain why a media company would be interested in data about their audience. This logic may be applied to each generation of technology, from newspapers, to radio, to television, to "social" media, to AI chatbots that show ads.

2. The great transaction cost

Lack of trust is the great transaction cost of the internet age. When each of us has means, motive, and opportunity to fake — to "cheat" in the game-theoretical sense — the platforms that facilitate mutual trust will meaningfully reduce transaction costs, and will be duly compensated for it. Is this a self-serving take? Absolutely.

3. Objective, strategy, tactic

Growth is a funnel, a journey, a flywheel, a symptom of value created and made manifest. It is acquisition, yes, but also activation, retention, reward, referral, loops that compound on themselves. The objective gives rise to the strategy; the strategy gives rise to the tactic, not the other way around.

Empathy is both strategic and tactical: you cannot design an experiment that moves a person if you don't understand what moves them, and you cannot prioritize which experiment to run if you don't understand which people to move. The method is experimentation. The prerequisite is instrumentation: you can't test what you can't measure. Growth is the objective, but objective-level growth is a lagging indicator.

Growth is the objective, but objective truth is too high a bar. Even if God does not play dice, we do not see the whole game board, and the decision to identify a result as "signal" from the noise will typically require an element of faith. Every measurement contains noise and bias; you can reduce both but eliminate neither. Growth, like any empirical endeavor, advances on bets that won't always win, and as in writing, one must be willing to sacrifice one's darlings.

4. Ancient wisdom, modern tactic

Many of the platitudes are true. Not all of them. But many of them. The trick is how to apply those platitudes in fluid situations, with incomplete data, and severe constraints. The magic is the combination of ancient wisdom with modern tactic. It may be a comfort that if you reject a true platitude for its commonality, it will still come for you.

5. The definition of slop

You must spend more time making something than the amount of time it will take to consume. This is why reading raw AI output purportedly written by a human feels so bad: it violates this principle. I believe this is the definition of slop: a product that took less time to make than to consume.

6. The worst form of information delivery

Lists are possibly the worst form of information delivery, at least if you have any interest in your audience remembering the information. They persist on the supply side because they are easy to make (no cognitive effort spent organizing the information in an empathetic way) and demand remains strong because lists are easy to consume. I've clicked on many a link for "50 things I've learned about love, life, and the pursuit," and I remember none of them.