Hacker News Reader

Comment by dsr_ | Hacker News Reader

In Peopleware, DeMarco and Lister write:

    Count on the best people outperforming the worst by about 10:1.
    Count on the best performer being about 2.5 times better than the median performer.
    Count on the half that are better-than-median performers outdoing the other half by more than 2:1.

What nearly everyone who has read that remembers is the following:

    The best people outperform the average by 10x.

Spot a difference?

View on HN

>Spot a difference?

The difference you point out doesn't matter because not everyone who uses "10x" in conversation is rigorously tracing it back to the 1987 DeMarco and Lister "Peopleware".

And arguably, the "10x" had earlier origins in the 1968s studies by Sackman, Erikson, and Grant : https://www.google.com/search?q=10x+Sackman%2C+Erikson%2C+an...

But again, that 1968 origin doesn't matter because many people are just using "10x" as a rhetorical synonym for "really good programmer that's much better than average".

The problem with dealing with these concepts quantitatively rather than qualitatively is that the deeper you dig, the less sense this sort of statement makes. There's two ways to read your quote. If "the best people" are the best in an absolute sense -- say, the top ten in the profession -- then the chance that /your team/ will sample the long tail of these high performers depends on the size of the team, but is nearly zero for any reasonable team size. And the distribution described is non-normal, with long tails skewing towards high performance. So if /the/ best people outperform /the/ worst by 10:1, /your/ best will outperform /your/ worst by less than 10:1. Alternately, if the statement is true for a typical team, then the expectation varies significantly with team size -- which has changed (in my impression, for smaller overall teams in many cases) since Peopleware was written.

I think we should assume the distribution is pretty non-normal. Feels to me like performance in many careers, especially so knowledge work, follows a power law. There's pretty good evidence for it: http://www.hermanaguinis.com/PPsych2012.pdf

Tangent: apparently contra DeMarco and Lister, I'd argue that "the worst" are often outright detrimental to a team or project. Which in turn makes me question the "math".

It is an interesting observation when put side by side with the sibling post [0].

I think it really depends.

A senior that is 10x more productive than an intern or junior? That's perfectly fine, they are here to learn.

A mid-level dev that is 10x more productive than another mid-level? A bit of a red flag, and probably detrimental to the morale of the team.

[0] https://news.ycombinator.com/item?id=31074653

I'll never forget having a negatively productive programmer on my team. If he spent an hour on a PR it would take 1.5+ hours of work to get the PR into shape to merge by other people because it took so much explaining for this guy. It was usually better to just scrap his code and start over. If code of his did slip through reviews often times there would be production defects that would then take more time to solve (tracking down the issue, figuring out where the code was, overhead of service tickets etc) than the original implementation. Just totally wild, he had somehow survived years in the org being shipped from team to team and being a nice pleasant guy to be around. How does one calculate the productiveness of someone on a team compared to that?

Being outwardly pleasant can take precedence over just about anything. I find it ironic as this can lead to a lot of other unpleasantness, perhaps even resentment, people leaving, etc.

Being technically competent while also being perceived as merely neutral personality wise can lead to a lower overall qualitative evaluation, potentially affecting raises, promotions, etc.

I guess the lesson is that you (almost) always have to play the game or at least that playing the game well can yield many personal benefits.

Just be thankful there is a process to get the PR's into shape.

My worst experience is someone who was highly prolific at creating enormous amounts of tech debt without the process in place to block it. We are still only about 50% through working off this giant stack of tech debt 2 years later.

I heard of a creature like that. I think they are called human being. Someone who was assigned to a team of highly efficient robots that never make mistakes and produces bug-free code. Production defects? Yeah, it's obviously that guy's fault alone. /s

If I didn't know any better, it sounds like you are just using the guy as someone to conveniently blame.

I disagree with that, it really isn't "normal" to have someone like that, but it does definitely happen. Doing a crap PR once in a while is absolutely normal for everyone, but doing that regularly isn't.

This is clearly someone who is working beyond their abilities, and needs help and mentoring.

This is not the person's fault, though, this is a management issue, and maybe also a cultural one, where people just assume this person "should know" just because they've been at the company for a long time, and mentoring would be "beneath" them.

The original study was timing how fast developers could solve a given task. The slowest took 10x the time of the fastest. With this setup you can't really get negative results. And participants who didn't finish the task were excluded from the results.

A summary of the research: https://www.construx.com/blog/productivity-variations-among-...

I don't think there is anything wrong with the math, it is just that real-world productivity is not as easy to measure. The study didn't measure the "maintainability" of the resulting solutions, only how fast they were completed.

Also note how sensitive the 10x number will mathematically be to your assessment of "the worst".

Since I believe in the existence of negative-value programmers, being 10x "more" than the worst would actually be a very bad thing.

McConnell from code complete mention this, because in one of the studies the multiplier only comes from people who completed the task, meaning in actuality it was infinite.

this aligns with the hypothesis that some people have that a handful of people do most of the work (and the others are the "worst" performers").

What this misses, is that in any highly functioning teams everyone is important and in order for the "rockstar" to be the "rockstar" they need people to help keep the machines running and to support them. This is usually why a team that has a mix of both veterans and newbies works so well. Everyone can operate at their own level while learning and growing. Is the veteran a 10x when compared to a newbie? If you make up some metrics, maybe, but the veteran cannot do meaningful work without the newbs (think rock concert without all the supporting crew. could any band do their own concerts without all the supporting crew?)

If a "veteran" cannot do meaningful work without the newbs, either that "veteran" is not very good or your processes are too manual.

The best programmers I worked with were not "rockstars", but "icebreakers" -- they would solve the hardest problems and leave the clear path for others. For example, they might solve a sporadic memory leak in a few hours, and thus save a less experienced team member at least a week of effort.

They were "10x" in the sense that any team they were on would get 10x better because everyone does more.

yeah. talking about "rockstars" and "newbs" is probably not helpful. but... the veteran not only can do what the newbies are doing, they can do it better and faster. The point is that in having the newbs do that, the veteran's time is better used on harder / more open ended problems.

to use an analogy, if we are building a sewer line, the veteran knows how to dig as he's done it before and it represents no challenge to them. Not only that but they have better technique and stamina. Now, someone has to also decide where the line goes and organize the workers. So the veteran's time is best used figuring out where the line should go and coordinating, not digging, but his work is enabled by the people that do all the "actual" work

If the worst performers on the team weren't creating negative value (the team would be more productive without them) then that was an incredibly good team to start with.

Based on this a median programmer is a 4x engineer

Reply on news.ycombinator.com