[jdmichal]

The biggest power of a CEO is culture. Look at Uber and all the negativity their culture is getting, fed by the specific instances where that culture allowed terrible things to happen.

And yes, a CEO does not change the culture on their own. That's impossible by definition -- the culture is the sum result of all the actors within the company. But a CEO is expected to lead the culture of the entire company, typically by stating values and setting goals and incentives that actually match those values. If they cannot do this, then they are unable to do their primary internal-facing job. This is one of the reasons that Steve Jobs was so well regarded. For whatever flaws he had, there was no mistaking the culture he was leading, and the results matched.

A great example, from a development perspective, is unit tests. Everyone loves to talk about how much we need unit tests. How great unit tests are. How they will make our software better. This is stating values, but you also have to back that up, and that's where things fall apart. The moment any kind of crunch happens, unit tests are the first things thrown out the window. Despite the fact that when you're working quickly, you probably need those tests even more. And all this is because, even though everyone states the value of tests, it's not actually part of the culture. The goals and incentives aren't there; there's no penalty to not following the line.

[speedplane]

Also, developing big projects practically require tests. I haven't seen a large project without them. Not sure where you work, but if they are the first thing that is thrown out, there is something definitely wrong not just with the culture, but with the engineering.

[jdmichal]

Completely agree with you. Part of the big fight is getting people to understand that tests are an asset and not a liability to working quickly and safely. There's a huge visability problem, in that management doesn't have insight into all the times when a bug wasn't committed because of a unit test. Functional and integration testing doesn't have this problem as much, because those typically have much better visability.

[speedplane]

> The biggest power of a CEO is culture.

I'm not exactly sure what you mean by this, but it seems a bit simplistic. CEO's are also in the position to be thought leaders, public advocates, build hype, and internally, guide the company to specific results. Workplace culture is extremely important, but not everything.

Also, I'm not sure that enforcing unit tests is exactly a "cultural" issue. How many unit tests you write is a decision that weighs current development speed against future development speed and reliability. That is of course influenced by culture, but not exactly culture itself.

[jdmichal]

> I'm not exactly sure what you mean by this, but it seems a bit simplistic. CEO's are also in the position to be thought leaders, public advocates, build hype, and internally, guide the company to specific results.

At least the first three of those are different ways of setting culture. The last one is so loosely defined that it could mean anything.

The sum of all actions of all actors within a company is its culture. If people act in an obsessive way about UX, then you have a culture of obsessive UX.

The CEO has the power to shift these cultures. There are many tools, as you point out, but the biggest one is setting appropriate goals, with the proper aligning (dis)incentive structure. These are set on his subordinates, who then filter it down and specify it for their subordinates, and so on down the entire chain.

> Also, I'm not sure that enforcing unit tests is exactly a "cultural" issue. How many unit tests you write is a decision that weighs current development speed against future development speed and reliability. That is of course influenced by culture, but not exactly culture itself.

If you supposedly have a culture that values writing unit tests, then that should be reflected in action. The idea of tossing unit tests should be met with disdain, the same way we value not murdering people and murder is meet with disdain. It's not about tradeoffs, it's about stated cultural values (talk) versus the actual culture (walk). If there's nothing holding you to the stated cultural values, that's a management failure.

[speedplane]

Normally, I would consider culture to be work ethic and how employees treat each other beyond the purely transactional business sense. You're defining it to be something much broader, that's fine, but I think that broader undertanding is better captured by the word "leadership".

And as far as my last factor ("internally, guide the company to specific results"), I mean that CEOs make a lot of day-to-day tactical decisions. E.g., how many engineer/sales/marketing people do we add to that team, do we need a bigger office, do we sue someone or not, do we take VC money on the offered terms or hold out for something better.

These tactical decisions are less visible, but probably make up a good deal of a CEOs day.

[jdmichal]

Culture is both a tool and side-effect of leadership. I talk about culture being the sum of all the actions of all the participants, but the funny thing is that culture also informs all those actions. Basically, there's a self-reinforcing feedback loop here, and that's one of the reasons why changing negative cultural aspects is both hard and important. It's hard because you'll be working against the currently ingrained culture. And it's important because you must break that self-reinforcing loop in order to affect real change.

And yes, CEOs can make tactical decisions, or at the very least give final approval on them. I think in smaller companies they tend to be more involved in such decisions, and as the company grows it becomes impossible to keep up and it gets delegated out. But in my opinion, those tactical decisions like the ones you list don't exhibit leadership as much as they need a leader to make them. It's a reversal of the reasoning: People aren't leaders because they make those decisions; they make those decisions because they're leaders.