John Fremlin's blog: Reducing tech productivity

Successful tech companies start out very high productivity places. A small group of people achieve great results. Then over time they grow, slow and some just stop.

The quintessential example is Twitter, whose main innovation after years of stagnation was increasing the length of tweets to 280 - and failed to grow revenue while sitting on a goldmine of engagement, cultural significance, and clearly massive potential.

Of course as a company becomes bigger and affects more people it's good to take input from stakeholders. The CIA’s 1944 Simple Sabotage Field Manual waxes lyrical on this pattern as way to undermine organisations: refer all matters to committees it enjoins. However, big companies like Apple and Amazon are able to set up these feedback loops, to gradually test and roll out changes in a controlled way, and keep on shipping really big things (like whole new hardware device categories).

Dan Luu offers a great example of 10000x productivity difference. I actually think he misses the point of his own example, which is that some groups are seem incapable of achieving an implementation given any time horizon. The factor is infinite. Not that the individuals involved could not, over time, learn and solve the problem - but that they are organised in such a way as to discourage this.

My experience at previous companies is that really a lot of these wounds are self-inflicted. Rapid changes to organisational structure, who is in charge, the official rules for how people get rewarded - all discombobulate. When this is expected to happen at a higher cadence than projects can be completed, rational productive workers may find complicated ways to sit on their hands, typically burnishing their CV with externally impressive open sourcing.

We need to admit there are self-interested people that want to reduce productivity and alienate the most effective employees by accepting bad behaviour. It's very threatening if you think 1000x more productive people would come in get promoted above you and then remove you very quickly (as they would presumably be productive at firing too). In the style of the sabotage manual: instead of rewarding outcomes, impact, results, etc. talk about "how" things were achieved. Particularly emphasize "sustainability" and "complexity"; i.e. reward things that affected a lot of other employees, grew teams and will be an ongoing effort for the company, instead of being solved once and for all. Productivity is naturally rewarded. It takes a concerted effort and mental contortions to punish it like this!

An example of a high productivity technical decision is WhatsApp not storing message history on its servers. Message history leads to a snowball of technical (storage) and feature requirements (search by users, search by law enforcement, ability to delete messages, etc.). Decisions like this avoid the need for great big teams. There is no way to reward them as much as the great big team would be rewarded.

We get low productivity because we don't commit to rewarding it. And on the other side, bloat and empire-building are very rewarded. People see teams get resources for doing a bad job (which is a totally natural first response). Trust that good work will be rewarded gets lost.

Like all management cultural issues the only way to fix it is to publicly, rapidly advance people - in this case, those who do unsustainable, simple things that have high impact. An uncomfortable trade-off with dangerous consequences of its own.

On the bright side, if you have the right people you don't need a lot of them. Fewer people means fewer incentive problems.