Hacker News Reader

I think it's also important to look at scale - in 2007/2008, Google had ~25.000 employees. Now it has over 100.000. If I'm looking at the graph correctly, Google hired 20.000+ of those in last year or two. The size of whole company in 2007!

That's a whole different ballpark than what hiring was in 2008. A slowdown in this scale can easily mean "just" hiring more people than anyone else.

I think part of this it's easy to feel like you have multiple managers. Typically you only have one true manager but you probably have multiple ppl who tell you what to do and not all of them have direct reports.

On the topic of not having enough work to do, here's a couple questions to any Google employees:

1. As someone generally frustrated with the quality of the search results I get on Google, is improving search quality seen as any kind of priority within the company?

To be fair I can't think of any other search engine that returns better results. But I feel like Google's gotten very complacent with their search, which is not surprising given the fact that it's practically a monopoly.

2. Is there any initiative to rewrite gmail on the web? The damn thing takes like 30 seconds to load. For a company pioneering progressive web apps and web performance, it boggles my mind that they've put no effort into improving gmail's atrocious load times over the last decade.

I imagine the finishing in quality is due partly to choosing a more aggressive trade-off on monetization, and partly on typical "large scale enterprise" inefficiencies

gmail was just recently rewritten. it was the rewrite that made it so slow. and i say rewrite from the perspective of the user. i have no idea what they did. it was just that there was a version that, while quite old in design and lack of features, wasn't disfunctional. then it changed and is now so slow to be disfunctional.

You can use the "basic HTML" version that's much better

The cynical view about search quality is that the metric being optimized for is not how satisfied people are with the results, but the amount of ads and therefore revenue earned. Search results only need to be good enough to get you to come back and not worse than the competition.

I wish Google had a subscription option (with actual support) that eliminated all ads across their products. Imagine Amazon's customer obsession with Google tech... I'm sure it would do well for the 4 years they would support it.

Google Contributor was their ad-free subscription option.

It was always a curious product, never really seemed to have management buy-in, and was pretty much a one-man product, which a 'subscription to get rid of ads across the whole web' shouldn't have been.

Controversial for google.com, unquestionable for YouTube.

On 2: any initiative to rewrite Gmail will get shut down or reworked again without notice. See, Google Inbox.

Highly recommend giving up on Gmail as an interface, and treat it as an IMAP/SMTP server only.

2. They released Inbox, which was amazing and had a cult following.

Then they killed Inbox. Presumably because they could not figure out how to capitalize on an extraordinarily useful, fast paradigm-shifting email client.

Just the insight we were longing for. How about very few users define "cult following" for you ?

My list:

1. How about an option to Maps to stop suggesting unprotected left turns at busy intersections?

2. Also, a Maps option to stop routing me through residential neighborhoods to save ten seconds.

3. Find existing, dormant Google accounts that obviously belong to the same human as another active Google account, and offer to restore access.

I left Google a few months ago, but I can speak to (1). I never worked on search directly, and of course I'm speaking only for myself.

> 1. As someone generally frustrated with the quality of the search results I get on Google, is improving search quality seen as any kind of priority within the company?

Search quality is kind of Google's secret sauce, so it doesn't get talked about a lot. But here's one public example of a huge engineering task that ultimately led to improved search results: https://www.blog.google/products/search/search-language-unde...

Using BERT in search is about much more than plugging an algorithm into the query pipeline. Google is one of the main contributors to the field of deep learning, and developed BERT. And Google also designed and built an entirely new chip (TPUs) to run this model in production.

I believe that you're not seeing improvements. And changing things without breaking other people is always hard, but particularly hard on software that billions of people use many times a day like Google search. That slows things down for sure. Honestly I was amazed they managed to roll out BERT on TPUs for search despite those obstacles. That should show a little bit of how committed they are to improving things.

I'd like to see an org-chart depth analysis for different organizations. It'd be interesting to map bureaucratic processes to the structure of an org chart and verify Conway's law.

The U.S. Army uses the concept of a span of control that dictates each level of hierarchy should have between 2 and 5 reports. There's about 10 levels between a private and the secretary of the army. The Active Duty Army is about 470k which works out to 3.7 reports per level from 470000^(1/10). There's about 1M total personnel which works out to 3.9 reports per level.

Google has 114k employees. With your ideal of 7, that works out to 5.3 reports per level given by 114000^(1/7). I'd be pretty surprised if there's a depth of 12 at Google. That's 2.6 reports per level.

A major way that things are different now is that it's much harder to get new people up to speed. My team is doing ok remotely, but we're not that good at it and I think the only reason it's going as well as it has been is because we had a lot of face to face time earlier. Bringing someone new on where we didn't previously have that time is a lot harder.

If we were going to be fully remote forever this would be different, but since this is (very likely!) temporary that's a reason to prefer to hire more later.

(Disclosure: I work at Google, speaking only for myself) (Well, I guess I work at home, but Google is my employer)

In times of economic depression is best to build and acquire, if you have the money to do it.

Acquire skilled people (it would be much harder to do it after crisis), buy land, buy companies, build buildings, build tech.

> Eric said when asked about it that it was a mistake.

He had the benefit of hindsight. In 2008 noone could predict if the GFC would last for one year or three years and how deep it would be.

The problem with hiring freezes (or reduced hiring) is that it creates a bathtub curve in your company. You'll end up with a lot of senior people, a lot of junior people, and a gap in the middle. This is a problem (somewhat recovered by this last decade of hiring) in the US DoD, primarily on the civilian side but also the military side.

In the 90s through the 00s there was a massive reduction in hiring for civil servants. So you had a lot of people who were there from the 80s and early 90s, and then a massive drop. Around 2010 or so they started ramping up hiring again, because they finally acknowledged the need (they had a huge percentage of civil servants who could literally retire at any point). So you ended up with (in some orgs) 30-40% of the staff being able to retire, 30-40% under 28, and the rest spread randomly from the age of about 28-55.

As people retired (biggest loss of personnel before other kinds of attrition), teams were literally losing centuries of experience in a year. The mid-career people were moving into leadership roles, and leaving the junior people to learn the processes and technical work largely on their own.

I saw something similar in private industry as well, but that was due to a preference to hire junior people over experienced people (either the manager wanted people who could be molded or cost).

Sometimes hiring freezes/slowdowns are important. But if your business is healthy, it'll end up backfiring on you more often than not. Even a year or two of this can create problems, though not as dramatic as what the DoD experienced.

It also introduces cultural problems.

In a healthy organizational culture, the org grows steadily at a pace that's not too quickly, and more senior members acculturate new recruits into the organization. Acculturation can't be rushed; it's the process of transferring all the unconscious, tacit knowledge about how and why things are done the way they are, and because it's unconscious, it happens only through long exposure.

When you have a hiring freeze, the existing culture ossifies and becomes more insular without any new blood. And then when the hiring freeze lifts, you usually need to "catch up" on hiring, which means that you have a sudden rush of new recruits and relatively few senior people. The new people don't have enough old hands to acculturate them, which means that they end up forming their own culture that may be entirely divorced from the existing org culture. Because the new culture outnumbers the old, eventually the old hands leave (taking their institutional knowledge with them), and the dominant org culture reflects that which the newbies brought with them.

Google's culture changed dramatically from 2011-2013, driven by the combination of rapid hiring, a new CEO (this may've been deliberate: if you want to change the culture, one way to do it is bring on a lot of people who agree with you), and the departure of key early employees like Craig Silverstein, Amit Patel, Marissa Mayer, et al.

Come to think of it, the U.S's culture wars are largely a result of the "birth dearth" of Gen X. Instead of a continuous shift in culture, you have two "big" generations of Boomers and Millenials and the latter developed a distinctive culture amongst themselves without regards to existing American culture.

> Is the situation the same now?

There is some difference in the current situation, however I am not sure how much it factors into this decision.

Hiring and onboarding people is logistically very difficult when travel has been affected worldwide and everyone is working from home.

> Management overhead = Mean org depth / log(# of employees)

Your model is in the right direction but misses more important aspects. When consulting I measure overhead of the orgs I'm working for in great detail.

Cultural aspects like fear, autonomy, transparency, etc have a much higher impact on management overhead than the org chart structure itself.

I’ve seen tiny organizations with massive management overhead. Friend of mine got a vacation denied by a 5 person all remote company because they didn’t file the correct paperwork asking for a day off!

Management overhead is a self inflicted wound. But very common.

I often come in and ask people to stop doing meaningless tasks, just to be told that the previous, much more expensive, consultant had told them it's the bees knees.

There is also a common misunderstanding that if a bigger or cooler company is doing something it must be the right thing.

Cargo cult.

Management is also just a lot of hard work. A lot of people seemed to convinced that good management is all about issuing orders, which leads to disaster.

I once worked at a nine person company where a three-minute fix to a piece of internal software ended up taking about two months of screwing around with revising a project proposal and, in the end, it was decided to drop the fix because what amounted to the #2 person in the company decided the tool would "eventually" be thrown away anyway.

Sad stories.

Just get it done. Don't ask.

A potential customer had been debating a problem for a couple of months with feasibility studies, lots of meetings, etc.

They asked me about it as an aside over a coffee. I implemented a quick prototype while they were describing their intended project structure to me. I gave it to them for free.

I did not continue with this customer. But I heard later on that a VP refused to accept the solution because he insisted it must be wrong since it was completely inconceivable that it could be solved that quickly.

> Just get it done. Don't ask.

Yeah, I didn't think it would go the way it did. That was the first time I noticed a problem.

What I really should have done is start searching for another job a week into this "process".

I'd rather give employers what they want, and -- if what they want is stupid or insane like that -- leave before I have to give the employer very much of what it wants, because I don't want to be a part of that.

> I heard later on that a VP refused to accept the solution because he insisted it must be wrong since it was completely inconceivable that it could be solved that quickly.

I loathe crap like that.

Autonomy and Agency.

Trust your people. Make sure they feel safe (not afraid) to make decisions and take action.

It drastically reduces the need for management. It makes people feel happier about their work, raises quality and reduces employee turnover rates. It makes it simpler to keep high skill people, which in turn makes it easier to recruit high skill people.

The counter argument I run across is that the mistake rate will increase. But that is simply not true. Through all my data I have never seen that happen. Generally the mistake rate goes down, and the cost of mistakes goes down a lot. And the cost of all actions goes down since fewer people are needed.

Couple this with transparency, asynchronous communication and decision making, and you have drastically reduced your company's management overhead, increased speed, increased well being, increased innovation, etc.

“ So years ago this came up with Eric Schmidt, who was CEO of Google through the GFC. After 2007, Google had selective layoffs (probably not in engineering) and slowed hiring. They also paused construction projects in Mountain View and probably elsewhere. Eric said when asked about it that it was a mistake. He said that if Google was healthy there was no need to have a kneejerk reaction to the slowdown. And this caused significant growing pains later.”

Was this covered by the media? Having a hard time finding anything about it.

Probably. I can't be bothered looking for a reference.

I was an engineering program manager and people manager at Google previously and around 2010-2012, there were opportunities for the company to really accelerate revenue, product development but they couldn't hire fast enough to take advantage. This stemmed from many orgs freezing hiring for a significant period (maybe 12-18 months).

At large companies, you have queues setup for most roles because you often growing the n count of that role by 5-10% per annum.

Also very few orgs plan for attrition (I don't get why because it's always been in the range of 4-20% in my opinion). Freezing hiring means a company's ability to execute reduces. Sometimes that is needed but often you must backfill to keep the lights on. Guess the messaging is easy ("no hiring. No exceptions").

> The ideal is ...

Nope. That was perhaps the case in the 50-80s, but today we can do much better, and it's dependent on the people, the culture, and the market/product structure and strategy.

I don't know why you were downvoted -- you're exactly right. It's also the reason why there are plenty of directors at google with no team, and plenty of lower level managers with 20, 50 or more directs (often lots of TVCs).

An observation from my work experience is that, while things run efficiently with smaller teams overall, small teams also tend to make it easier for any given leader to have a reasonably good idea what's going on across their org (it's easier to grok a 50 person team's work when it's split into 10 mgr teams than if it were completely flat), but that also puts the lower level managers' jobs at risk because their director is more likely not to consider them essential -- ironically, because things are running smoothly.

Not just google. All large companies I've inspected have had similar issues. Appearance becomes important. People loose agency and autonomy. Communication stalls. Things take longer. More misunderstandings. More mistakes. No one takes responsibility. Redefinition of words like "success" and "failure". Endless meetings. Etc, etc.

Smaller companies tend to have stronger focus of just getting stuff done. It's clear who has responsibility. People can more easily make a decision on their own or in small groups and just get it done, without involving a clown car of middle managers.

When I build projects I only have two "levels". Management have final say on budgets, targets and priorities. Experts/devs/techs get the work done. They generally make decisions with full autonomy and self organise. If they need help with coordination, planning, followup, etc then management steps in. Management also monitor and track. Usually my management has very strong technical skills and tend to be part of the daily work in the trenches.

> The ideal is (IMHO) CEO -> SVP (PA level) -> VP -> Eng Director -> Manager (of managers) -> Manager (of ICs) -> IC.

That assumes a fanout of 7-8 at the current size of Google. I think that's too much at certain levels.

There can be no absolute correct number of layers. That is obviously too many layers for a 20 person company. And what if Google grew to one million employees. Would that still be the right number of layers?

I think it's possible that there is no correct number of layers for a 1 million person company, i.e. such a company should not exist and should split up into smaller, more focused companies.

It appears that the two US companies with > 1 million employees are Wal-Mart and McDonald's (counting franchisees).

Obviously a lot of people don't like them, but it's not obvious to me that they would naturally break up into more focused parts.

I think you're overestimating the mean org depth.

> In fact this would be a good metric:

On a slightly different note: this isn't a good metric. It's a metric of how "balanced" the company is, not the management overhead. If you imagine a company with 150000 engineers and one salesperson, with sales reporting directly to the CEO, hiring more people under the sales director will decrease this metric, while increasing the mean org depth. That probably doesn't measure what you want it to.

Advertising is taking a much bigger hit here than 2008.

your "ideal" of 7 seems invariant of size. surely the needs of the average ~x000 person company are completely different from Google.

Yeah, when I worked at thousand-person companies it was like this:

CEO -> SVP -> VP -> Eng manager -> IC

Or, alternatively,

CEO -> Director -> Manager of managers -> Manager -> IC

I didn't feel like either needed more middle management (each had about ~2k employees), but in the case of the latter some reorganization would have helped (some managers had 50+ reports).