The working hours paradox

We love to boast about how much we work

It’s always been a weird status symbol of the middle-class to boast about how hard you work. A mother of a McKinsey consultant, lawyer or banker can proudly say “oh he/she is working so hard in a great job!” with little regard for whether that work is actually effectual (at their given task, or societally)…because that’s not the point, that’s not the thing to be proud of, it’s the self-flagellation that we are proud of.

This is because we are narrative creatures and the greatest narrative of all human history is the simple story that success and progress requires suffering, so where we see suffering we presume progress.

Why are we talking about this now?

It’s not a new cutural paradigm, but the discussion around it has been reignited by the recent trends in AI companies. Many of these companies have no moat and a thin product layer ontop of commodity LLM APIs. Even worse, they face huge distribution risks, because when you’re building a relatively simple product that naturally sits adjaecent to/on-top of (or even worse, within) someone else’s platform, you are only ever one product change away from being overwhelmed by that platform’s distribution advantage.

A similar example that comes to mind is Microsoft Teams vs Slack. It’s not a perfect analogy, as Slack wasn’t building on/within Microsoft’s enterprise platform (so if anything, it’s not quite as bad as it could have been), but it highlights the advantage of distribution nonetheless. In this example, Slack still is a great company, but the result of Teams’s distribution advantage (i.e. releasing a similar software to all enterprises who pay for your suite already) is the difference between Slack being a super solid company, and the opportunity that would have existed to be a generational business, defining how we do work.

Anyway, back to AI companies…why have they created a new obsession with working hours?

What’s the result of this existential risk they face? An imperative to move as fast as humanly possible, the true VC blitzscale playbook, to build enough of a business that something can survive even when the platform they’re building on releases the inevitable feature that would otherwise kill them.

Now we see companies burning millions monthly acquiring customers and reporting higher and higher “ARRs”, even when they know many of these customers won’t last a year and that the CAC and churn rate means that they are fully just torching a decent share of that cash. It doesn’t matter, because the play is not to build a business that makes sense (yet), the play is to use sheer scale of money and userbase to build something, anything defensible… a brand, enough unique data, enough product improvements/features over time, anything to have a hold of something defensible before the music stops, fully aware that the current product is not that.

If you’re moving as fast as possible, there are 2 resources you have available to throw at the problem: money, and time. Raise lots, burn it in marketing, hire more people, and then the final piece…get everybody to work more and more hours.

This is what has reignited the discussion of working hours, and particularly has people talking about this 996 paradigm (9 til 9, 6 days per week, meaning 72 hours a week), or in the case of the below WSJ article, has 20 year olds boasting about doing 90-120 hour weeks.

So is 996 a good idea?

In a word? No.

In several words, no because it more likely than not harms more than it helps.

Why?

Starting a company is like a marathon…

This is honestly the simplest and best analogy. I’ll give some specific reasons and evidence after, but lets start with this simple mental model for now. The best marathon runners maintain a steady pace that is designed to minimize their overall race time. The funny thing? Starting out, this pace feels slow.

Anyone who’s done even a 5k before knows this feeling well. If you’re not careful, you set off and you’re thinking “man, look at me go! I’m killing it, this feels great! My mins/km is great!” Then 3 kilometres in, you’re dying.

If you want to run at a pace you can maintain, you have to consciously run a little slower than you would like to. It takes real restraint. Even worse…the best marathon runners tend to actually speed up THROUGH the race, as they get comfortable with how their run is going! That means they’re not just starting at a pace that’s slow enough to maintain but even a little slower than that, leaving enough gas in the tank to accelerate through the race!

In a business context what does this mean?

It’s easy to think “I don’t need my friends and family! I’m not going to burn out, I’m just so damn passionate like no one else has ever been!”

Calm down. You’re human.

If you avoid social connections, don’t sleep well, grind every hour of the day on this, it will catch up with you, and ultimately that’s just gonna mean you end up going slower. Having burn out, losing motivation, alienating your team, they ultimately will make you slower, like the hare who loses to the tortoise.

This is basically why 996 doesn’t work, because it comes from a place of hubris; not accepting the frailty of our human bodies/minds, in a context where measuring the output is much harder so we can easily miss that we’re actually doing worse.

Is this just my opinion, or is there evidence?

There’s evidence.

I don’t want to do a literature review; you can easily do your own research if you’re curious (ChatGPT makes this trivial). Instead I’ll just say that there’s a wealth of evidence that supports the view that working very long hours (generally over 60ish) is both bad for your overall productivity and bad for your health, and highlight 2 key points with some relevant evidence.

1 - Beyond a certain point, not only is your marginal hour less productive but you are actually reducing your total output.

To understand why this is the case, you need only think of the hyperbolic example of literally working 24 hours a day. The quality of work would become so bad it’d clearly become counter-productive to work at all, and in fact you’re also making your next day’s work worse by ensuring you won’t be rested. So clearly, there’s a line…the question is where the line is drawn.

Here’s the relevant study backing this up (amongst many). As it says below, working over 70 hours probably net REDUCES your output. The numbers here shouldn’t be taken as sacrosanct, but it makes sense as intuitively 50-60 hours is where your working time would really start to limit your ability to meet human needs (general life maintenance, good sleep, exercise, maintaining social connections, etc).

Pencavel (Stanford, 2014) – The Productivity of Working Hours

• Productivity rises with hours only up to ~40–50h/week.

• Beyond ~55–60h, output plateaus; 70h workers produced the same as 56h workers!

• Extra hours after ~63h became counterproductive.

2 - Part of the reason this is true but we feel good about doing more work is because our cognitive abilites decline more than our perception of those abilities when we’re under-rested.

This is the hubris part. It’s an interesting finding that helps explain the mismatch between the obvious reality and how it feels like you’re doing something good by working more.

Here I’ll provide 2 studies. The results are kind of obvious if you think about it, of course if you’re doing knowledge work you’d prefer to be well rested, it just becomes a line-drawing exercise - of where the optimal balance between working more and sleeping enough is. Turns out, it’s probably way under 996.

Van Dongen, Maislin, Mullington, & Dinges (2003) – Sleep, Vol. 26(2)

• Participants restricted to 4–6h sleep/night for 14 days.

• Objective tests (Psychomotor Vigilance Task, reaction time, working memory) declined steadily.

• Subjective sleepiness plateaued after a few days.

Durmer & Dinges (2005) – Seminars in Neurology

• Review of neurocognitive consequences of sleep deprivation.

• Shows divergence between self-rated performance (little change) and objective measures (sharp decline).

• Notes “microsleeps” and cognitive lapses occur without awareness.

There are also many indirect effects such as greater stress from reduced focus on social connections and hobbies likely meaning you are sleeping worse and generally distracted, higher likelihood of anxiety/burn-out/insomnia, etc etc.

Long story short, working more than about 60 hours per week is probably more useful for bragging about than actual output of your work.

A picture is worth 1000 words, so a graph must be like 10,000 right?

Lets try and understand that Pencavel study. This is what they’re saying your marginal productivity looks like, given different hours of work. It peaks at about 45 - I’ve made some assumptions around the shape before and after, but this is roughly their point.

Excuse AI messing up the titles, it’s an index of productivity on the Y and hours worked on the X. I’ve also made the generous assumption that it never hits negative - but actually this obviously does become the case at some point (e.g. if you literally worked all hours every day).

There’s another interesting thing to consider here though. They specifically said people working 55 hours had the same output as 70 hours. This can be true without the marginal productivity per hour quickly dipping to negative if it’s not only the case that your marginal productivity declines, but that your productivity on all previous hours falls too - which, intuitively, is likely the case.

Again, you can use the hyperbolic example to understand this; if you worked 150 hours per week (you only have 168 total), you would expect pretty soon that ALL your hours become kinda of garbage, not just the marginal 80.

So what does that mean? You end up with a graph that looks a bit like the below (it’s not to scale, but it should make the point that by working fewer hours, you can end up in the same position if working too much is causing your typical hour to be less productive).

The Y axis is “productivity”, the x axis is hours worked - you can see that if our assumptions are true, the area of the left triangle and right triangle may be similar even though you’ve “worked more” in the higher hour case. This is exactly what their study is finding.

3 more assertions that are more just my opinion

These are just my view, not studied arguments, but…

1 - Most of the people hyping this stuff do not practice it themselves - some wholly, some because they’re not counting their hours properly

In many cases when people think they are working 996, in reality they are most likely generously misclassifying their hours of work. “I sent an email, I was at my desk, so that was an hour of work!”

2 - Even if someone IS doing this…that doesn’t mean they couldn’t be doing better by actually working less

Even if someone truly is working 996 and is successful, based on the evidence and intuitive benefits of being well rested and having a balanced life on our work, can we not say that just because you have managed to work 996 and be successful, might you not have been even more successful? Is your success inspite of your hours, rather than because of them?

Afterall, the person doing 70 hours matched the person doing 56 hours in output, in the study mentioned above…so yeah you’re output is still great at 70 hours, you’ve just wasted 14 of your time is all!

We see this a lot with startups where we assume it’s great because the founders as #$$holes, when actually it could well be DESPITE this fact…

3 - It is probably also a net loss in terms of restricting your hiring pool and worsening your retention.

It’s hard enough as a founder to put up with this lifestyle, when you stand to make millions/billions. As an employee? You’re probably cutting your retention by years by being asking too much of staff.

What even is a working hour?

I touched on this above…what even is a working hour? Are all hours of work equal? Obviously not. So when someone quotes “120 hours of work”…is it humanly possible to “work” for over 17 hours per day? Does it just mean being at your desk, even if you’re just flitting between chrome tabs and checking dashboards? At that point…is it even real work, in the deep sense, that you’re doing? I’ve had more valuable hours where I was showering and doing chores, then had an idea come to me, but I wouldn’t classify THAT as a working hour!

Ultimately, hours worked is a foolish measure and a difficult standard to start to compare…if I go to the bathroom in that hour, is it a full hour? If I flip between whatsapp and slack, is that a full hour? Is my hour of work the same as yours?

Output is what matters. We see this with sales people most plainly, knowledge work where output is most easily measurable. Would you prefer the sales person who works 1 hour per week but closes $1m a year, or the one who works 120 hours per week but closes 100k? So, if you boast about doing 120 hours of work a week…I probably am going to assume your output per hour is pretty bad and you are over-estimating your body’s ability to perform.

And you know what’s the ironic thing?

I specified middleclass because this is where we boast about working hard. For the very wealthy, instead it’s a boast to say how little you need to work in order to generate inordinate value, so obviously the takeaway for them is that they are more important and valuable (not part of a team and reliant on the team below them).

Slaving away on excel models for 12 hours a day is low-level work, the CEO needs only make a call and close a deal to return 100x the value.

Funny isn’t it?

What are your thoughts? Message me, comment here (if that’s a thing, I’m not even sure) or hit me up on Linkedin.