Contempt for the glue people
surfingcomplexity.blog
surfingcomplexity.blog
I'm not wholly sure what Schmidt meant. Maybe vague hangers-on people who kind of dip in and out of things and cheerlead, and take a bit of credit for every project they have a finger in, but not doing any significant lift? I'm not sure at all though. I'm just trying to rationalize it.
On another note, I've seen in government that there is an exponential growth in not just glue people but glue teams. Here the problem is obvious, the bigger an organisation is, the harder it is to ensure adequate communication between teams working in similar areas because there's so many teams to keep track of. Whenever a communication failure occurs, someone puts their hand-up to make a glue-team, a manager thinks 'this will fix this problem', and thus a new team is born.
Of course, the issue is that the glue-team that might now have resolved the miscommunication problem between teams a and b has now added yet another team to the org chart. The consequent increase in organisational complexity begets more glue-teams elsewhere, you have glue-teams to link to other glue-teams, the solution makes the problem worse, which leads to even more of the solution.
To avoid this run-away explosion in glue-teams I'd suggest two things. Firstly, the fact that cross-team communication is hard means it should be minimised, organisations should be as modular as possible. This is what bureaucracy originally meant, e.g. the delegation of power to Persian satrapies which meant they could act autonomously. Management schools often teach the reverse, thinking that the way to solve communication failure is to make it a strength by maximising the amount of cross-team communication resulting in endless proliferation of glue people and teams. If your organisation requires good information flow from everywhere to everywhere, it's built to fail.
Secondly, glue should be the last resort. How many communication failures are actually because teams don't have well-defined roles and so things slip through the cracks etc.? How many could be fixed by simple process changes (e.g. monthly catch-ups between related teams)? I'm with Schmidt, the less glue the better.
IMO experience, the defining attribute of these people is that they give emotional comfort to insecure management. Having an org cheerleader around when times are difficult will help make a CEO feel warm and fuzzy inside and help to drown out the discontent.
The good ones are not entirely useless, sometimes they can facilitate org progress in ways others are unwilling or not capable of doing. But at worst, they can create more inefficiency then they solve.
The true mark of a person's value to an org is - in their absence, do things get better or worse? I have seen highly paid "glue people" leave and things actually get easier and better - those people are often forgotten quickly too.
I don’t understand this at all. “Someone who assists in the interstices” doesn’t sound like anyone I’ve seen or worked with.
Can anyone give examples of what kind of employee/role he’s talking about?
Obviously the ideal answer is to have teams do better work and documentation to reduce the need for cross team communication, so I can see where a "no glue people stance" could come from, but at most mid size companies I think its at odds with the business velocity demands.
Eric Schmidt
>Machine men, with machine minds and machine hearts! You are not machines, you are not cattle, you are men
Charlie Chaplin
It’s enough that it’s the last place I’d want to start a job at because I could get laid off just as soon as I start.
A Harvard/Stanford/MIT education used to be about the rigor, the education, the knowledge, the tutelage. Now it's about having that name attached to yours. Same goes for FAANG.
“No, it’s just, it’s survivorship bias. There were plenty of people like that there then too; you just haven’t heard of them because they’ve done nothing with their lives since.”
Another way out to avoid laying off or not giving raises to essential employees is better performance management with qualitative and quantitative recognition and characterization without creating specific KPIs falling into Goodhart's law. Peer 360 evals that feed into performance reviews for cross-functional areas to show their strengths when they're supposed to be strong in that area. Also, there should be some discretion given to managers to advocate for staff who improve performance of the team or of multiple teams.
The last person I respected there was Matt Cutts for his combination of public relations and “make the freaking training set at all costs” mentality that separates the people who make working ML models from the rest. Google doesn’t give that to the world, no it gives us Kubernetes and OKRs to kill startups which could be potential competitors. (Sorry when product-market-fit is job #1, #2, #3 and #4 you can’t waste 30 seconds on a system which will twist all your productive people into knots while giving narcissists free reign to determine how they and others are perceived by the organization.)
I worked in Apple retail way back when, and had a co-worker who was the absolute best "customer service" person I have ever worked with. They had an unnatural ability to make anyone both calm and if not happy, then at least open to talking. Our store was 45 feet from front to back and on multiple occasions I watched them intercept an angry customer who had practically torn the door from its hinges walking in and by the time they'd reached the back where the "Genius Bar" was, they'd gotten the customer cooled down and in many cases actually smiling and laughing. As the retail management became more and more metric focused, they were constantly in the lower rankings for metrics, and yet I argued to anyone I could that they were an essential part of what made our store work and we needed to keep them around regardless of their metrics. Eventually I left because of the ever increasing focus on metrics, and a few months later I heard from them that they'd also been let go for failing to meet metric targets. The few occasions I've been back to my old store, the lack of happy or at least content people waiting around for their turn for service has been noticeable (at least to me in comparison).
I've seen things similar to this play out multiple times over my career. Too strong a focus on numbers and "objective" measurements, and not enough focus on soft skills and overall cohesion and how people fit into their teams. A team of 3 10x developers that don't mesh well on their own with a 1x developer who keeps them all working together smoothly by being a "glue" person is worth more than a team of 4 10x developers that only mesh ok. Everyone I've ever talked to in life understands this concept, and can easily think of examples from their own life. Yet somehow the lesson flies out the window anytime it comes to making hiring / promotion / firing decisions. I think people are afraid to make decisions like that which they can't back up with cold hard numbers, and while there's good reasons to want objective numbers for these decisions and avoid appearances of bias and favoritism, we lose a lot if we only ever focus on the tiny slivers of things which we are currently objectively measuring.
Lukas Podolski played only 56 minutes but was important to the team at such a glue people position.
You need people you like at your workplace.
glue people connect facts and context and teams and plans. They make sure resources are effectively utilized. Actively trying to remove cohesion from a company is like running an engine without lubricant. Super dumb.
I've been the glue. I didn't want to be glue, I wasn't hired to be glue and I do not wake up every morning to be glue. But I glue because I care about optimizing for results and it's the fastest way from A->B. Eric Schmidt doesn't know what he's talking about.
But I think the point of the article is that Google seems to have pretty serious problems stemming from the fact that it apparently has lots of individual teams working on their own thing without enough coordination with each other.
And I say this, because the least generous reading I can make, and the one I think is true, is that they are all sociopaths.
Some people think they’ll learn something useful at Google but it’s a delusion because you can’t take the monopoly is useful.
If you’re not reflexively deferential it looks like they pay people 2-3x to be 1/3 as productive as they would be in a startup but they can get away with that because the cash flow productivity is magnified by their profitability and scale.
As for “sociopath”, see OKRs. I can’t think of a better weapon sociopaths could use against the rest of us.
OKRs comprise an objective (a significant, concrete, clearly defined goal) and 3–5 key results (measurable success criteria used to track the achievement of that goal).
I worked a startup that had exactly one goal: find product-market fit. It’s a little more complex than that because that kind of goal has numerous sub goals, and the exact route depends on what exactly our product is going to do for what sort of customer. It was also a technologically ambitious company as we were trying to develop foundation models just before the BERT paper came out so lots of things were up in the air.
We were zigging and zagging and going in circles, some of that was we were selling to big enterprise customers such as a large European aircraft manufacturers, Big 4 accounting firms, etc. We always had to change direction for a particular product which was fine, but challenging because we were also developing a core system that empowered these engagements. Lots of work in progress was being wasted because we were always pivoting for the next client.
To make matters worse, the founders didn’t completely trust the founders so they were releasing our investment in dribs and drabs so we did‘t spend it foolishly.
If you asked me what to do to right the ship it would have been “address the WIP problem”, but instead the funders got the founders to get us to read about an OKR process where each of us individually was supposed to decide on 20 or so goals, including work goals but also personal and professional development goals.
Now there are some sociopaths who really have no fear, despite being really dangerous women often feel “safe” around them because they they feel no fear and radiate that. Many of them have narcissistic personalities which give them superior abilities in “presentation of self”.
They do very well in organizations that have formal performance management systems because they can put 100% of their effort into gaming the system and probably get 800% of the results that I’d get if I tried the same. “Stack ranking” punishes and drives out conscientious people who put 100% of their effort into their work. Even if they put 100% of their effort into self-promotion they’d get maybe 15% of results out of that than I would because they lack the skills and would feel guilty about it.
Thus a danger of many corporate environments is that sociopaths ride a rocket ship to the top.
With OKRs, the inmates are running the asylum. People who already have the skills and attitudes to game any system can now decide the criteria by which they’ll be evaluated. If you say that squares the effectiveness of their core competence now they can outdo me by 16,000% at the evaluation game.
And what’s great about it from their viewpoint is that that vast majority of their co-workers have no understanding of what the game really is.
Even in cases where everybody believes you can attribute business value to individuals such as sales are problematic. Consider the star sociopath salesperson who is so good at persuasion that he beds the wives of his clients.
It may be beyond my pay grade to say what that company should have done to solve its problems, I think the trust problem between management and investors should have been dealt with more directly than it was. As it is I know OKRs burned up 2-4 weeks of every person’s time which was probably worth a quarter of a customer engagement or so.
You find this happening again and again; good scientists will fight the system rather than learn to work with the system and take advantage of all the system has to offer. It has a lot, if you learn how to use it. It takes patience, but you can learn how to use the system pretty well, and you can learn how to get around it.
Hamming, “You and your research”
Astronomers saw me that way. "Cliff? He's not much of an astronomer, but what a computer hacker!" (The computer folks, of course, had a different view: "Cliff's not much of a programmer, but what an astronomer!" At best, graduate school had taught me to keep both sides fooled.)
Cliff Stoll, “Cuccoo’s Egg”