Comment by dfranke | Hacker News Reader

Comment by dfranke | Hacker News Reader

If I had to structure an organization of 1,000 people, here's how I'd do it: rather than tiers of management on the order of 1, 10, 100, and 1000, I'd just have two levels. At the top would be a small committee that controls allocation of money and from time to time publishes non-specific guidelines about where they'd like the company to go. One person on the committee is in charge of hiring salesmen, who once hired are basically unsupervised and work for commission. The rest of the company is a bunch of small engineering teams of up to 10. All functions other than sales and engineering are outsourced. The most senior member of each team has hiring/firing authority for that team. He gets to decide the size but 10 is the max. These generate their own ideas and speak to the controlling committee only for budget requests. Salary is not part of the budget request: everybody in the company gets the same salary. Bonuses are awarded to teams that implement profitable ideas. More money goes to bonuses than to salary. If a team shows a pattern of underperformance, the committee fires the entire team. If this happens, they get 30 days notice that they're going to be fired, and they can use that time to find jobs on other teams within the company.

If I then had to scale this to 10,000, I still wouldn't add another layer of management. Instead I'd just add more teams, and expand the controlling committee to a confederation of committees, which all draw money from the same pool but have specialized expertise. When teams request budget, they approach the committee with the best understanding of their proposal.

The 1,000 person model looks a lot like a bunch of founding teams and one VC firm. The 10,000 person model bears some resemblance to the federal government, which is perhaps why an organization the size of the government manages to function at all, however poorly.

One very serious problem with that idea. Team X isn't insulated from lawsuits that Team Y gets hit with (or contracts that have the potential for massive costs, for that matter).

Suppose Team X isn't doing well. Probability of getting fired is 50%. They decide (independently) to do Z, with a 90% chance of increasing their team's profits (and not getting fired) as well as a 10% chance of major lawsuit that will crush the company.

This nearly doubles Team X's survival probability, while exposing the company to a massive risk. This is a classical example of agency costs. And this is why big companies are usually pyramid shaped.

You also just described the major problem with the government artificially allowing such a thing as a limited liability (I'm talking about legal, not debt) corporation.

This has little to do with limited liability. In fact, the problem here is that groups within the company do NOT have limited liability (from each other), thus necessitating some oversight.

The only role limited liability plays here is that investors in this company to lose only their investments (but not their house and car) should the company die.

Could you structure the company (legally) such that each team was an independent company that the parent company holds a majority stake in? This way if an individual team got sued, the parent company's liability would be limited to it's investment in that team, and sibling teams would be isolated completely.

Of course, I have no idea if that is how liability works, and even if it was, the accounting situation for the overall company would be a nightmare.

If I had to structure an organization of 1,000 people

You might also want to read this: http://www.amazon.com/Maverick-Ricardo-Semler/dp/0712678867

Just based on the summary, it sounds like he has a lot of good ideas and one bad mistake: profit-sharing is socialized. So if the company does well then the employees do well, but individual contributors would have trouble getting rewarded.

It may not be as bad a mistake as you think. There's some experimental (if unintuitive) evidence that rewarding individual contributions in organizations is harmful, even to that individual's motivation. The extrinsic motivation extinguishes the individual's intrinsic motivation.

Many high achievers have an attitude of "achievement for the sake of achievement." Individual rewards crush that spirit.

Joel Spolsky wrote about it on more than one occasion. Here's one link: http://www.joelonsoftware.com/items/2006/08/09.html

See also:

A) Punished by Rewards (Alfie Kohn)

B) Herzberg's Motivation-Hygiene Theory (Two Factor Theory). The idea is that a very low salary can lead to dissatisfaction, but increasing the salary beyond the required minimum isn't one of the things that increases happiness. Link: http://www.netmba.com/mgmt/ob/motivation/herzberg/

There's a huge difference between the sense of ownership that comes automatically from small organizations and the 'incentives' that a paternalistic committee chooses to bestow on a few in its largesse, while bleeding motivation in all the other ways large organizations do that PG talks about. I don't interpret any of those studies (including Herzberg's in the sibling post) to mean 'socialism isn't so bad'. Yes it is so bad. But that doesn't mean you can't implement capitalism poorly.

Capitalism isn't about profits, it's about motivation.

That's not entirely true. Workers set their own salaries, subject to the review of their colleagues.

If you think you are awesome, you can give yourself a hefty raise. If your co-workers don't agree though, in a few months you'll be out on the street.

I think it makes sense to share profits equally, since the whole point is that the result was achieved collectively. Salaries can be all over the map.

This is why I love the HN community - not one but two people recommend the same book - that's a huge vote of confidence that forced me to take a better look at it.

Semco sounds just like Nucor and Gore (of Gore-tex fame). There's probably a few companies like this around.

I was thinking your 10k person model sounds like how research funding works... which seems to be a moderately functional model.

With your 1000 person model, I'd do away with the boss (in the sense of hiring and firing). With that few people, you can all just work together fairly easily, with natural leaders emerging for different projects (and hiring/firing decisions can be made by the team as a whole). There is a factory in NC that makes aircraft engines that works this way, and it is quite successful. The teams are self organizing internally, and send representatives to deal with other teams and plant management.

In practice that's how I'd expect/hope it would work. But while ideally the groups should act by consensus, there has to be some means of making a decision if no consensus can be reached. For groups that small, a benevolent dictatorship tends to work better than a democracy.

All in all, not a bad framework, though I expect much tinkering would be needed to optimize it. It’s worth noting that almost every large company strives to be “entrepreneurial”, although some are better at implementing this than others. The world’s largest private company (Koch Industries, I believe) prides itself on “market based management”, though I’m not sure what this means in practice. Koch is also a big funder of libertarian causes, which gives some hint at his mindset.

I know of one huge company that does exactly this. Except instead of outsourcing "all functions other than sales and engineering", they outsource nothing but sales and engineering. :)

Reply on news.ycombinator.com