Beware of round numbers

A lot of times, people are just guessing

Not too long ago, Google co-founder Sergey Brin was in the news for claiming that 60 hours a week was the “sweet spot” for how much people on Google’s AI team should be working. Apparently more than that would lead to burnout, but less than that would leave some productivity on the table.

My immediate reaction was suspicion. If there is a sweet spot to how much tech workers should be clocking for peak productivity, how miraculous that it is a number that is exactly divisible by 5! That’s even crazier when you consider that very few people are as productive at 6 p.m. on Friday as they are at 9 a.m. on Monday. What a coincidence that that the “sweet spot” calculation somehow works that all out, automatically adjusting to create longer days on some days and shorter on the others, and yet still producing a round number.

Obviously this isn’t all a coincidence — that number is most likely not based on some longitudinal study of work logs and individual and team output. It just sounds good. In general, round numbers satisfy a human desire for something official sounding, but they tend not to be telling the whole story.

The overestimated workweek

I think I’m particularly sensitive to this because I write about time, and so I hear all the time about people’s “60-hour workweeks” or “80-hour workweeks” (weirdly, no one talks about 70- or 75-hour workweeks, even though those are round numbers too — there seems to be a total jump from 60 to 80). I also spent approximately 14 months as a fact-checker on the opinion page at USA Today, and my main take away from that experience is that a lot of the numbers people cite are wrong. Statistics often go through the equivalent of a giant game of telephone as they are repeated in various articles and then later cited.

Numbers sound official, which is why people like them. Recently, several tech company CEOs were in the news for talking about the percent of their organizations’ work (sometimes just coding, but also sometimes things like customer support, or work in general) that was being done by artificial intelligence. They could have just said “we’re already seeing some productivity gains,” but people like numbers, and CEOs also like to one-up other CEOs on numbers, and so some claimed 20 percent, 30 percent, 50 percent.

More exact numbers

These numbers sound like they’re based on something, but when we’re talking a percent of work, and a round number, again, it seems a bit suspicious. It would be far more believable if someone said “The XYZ department has been able to reduce the time for this particular sort of deliverable by 27 percent by using AI” or something like that. But I’m not sure people have those exact numbers yet. There’s no set way to calculate the percent of work for a lot of things. Hence the round numbers.

Of course, if you would like to lie convincingly, this suggests an obvious hack: Use a not-round number! If you told me your employees were most productive working 43.6 hours per week, I would definitely think you had tracked that for multiple weeks across multiple people looking at output. Since most people aren’t tracking, that number would likely be wrong, but it’s probably no more wrong than a lot of the round numbers people do use.