I’m not going to sugar-coat it, I’m just going to say it right out. The most useful lesson I have learned about pay is: there is no right way to do it. And I don’t mean “there is no one right way to do it”. I mean “every approach to pay is wrong in some important sense”. That might sound like an unduly negative way to look at things. It’s tempting to dismiss it as hyperbole — but I hope I can persuade you both that it is true and that there is something positive here.
I want to start off with a story from my own working life. In 2014 I joined an awesome company called GrantTree. The founders were deeply committed to running a different kind of company, and embraced a full-fat version of self-management that included a default to total transparency and very little hierarchy. Not longer after I joined, the CEO announced that compensation, one of the few powers that had been reserved to the founders, was going to be democratised. Partly this was motivated by a desire to reduce management power; partly the company had just reached the point where it was time to put in place a salary model (usual story: ad hoc setting of salaries, especially for early hires, had lead to a lot of inconsistency, and it was starting to cause friction).
I was one of the “lucky” group tasked with this problem, the so-called Compensation Committee. As part of our research we were lucky (genuinely lucky, no quote marks) to get speak to Joel Gascogine, CEO of Buffer, a company famous for publishing its pay model. One of the things Joel said to us was: “You cannot get this right. We’ve tried for years, and no matter how much you tinker, there are always problems.”
This, it turned out, was very good advice. We were being given a shortcut to wisdom via somebody else’s hard-won experience. But, as so often in life, we didn’t take it.
Over the next few years, various people tinkered with the pay model at GrantTree, not quite non-stop, but certainly regularly enough that it began to irritate even our highly-engaged workforce. We had several Compensation Committees and even a Pay Tsar. And did we end up with something significantly better than when we started? Probably not. (At least, not in the sense of leading to a better overall assignment of people and salaries. Our approach certainly become more interesting as we pushed decision making out to individuals, but that’s a story for another time).
Why did we ignore Joel’s advice? Keith Johnstone points out in Impro that your first idea is very often as good as any other idea you come up with, but we reject our initial impulses thinking we can come up with something better because we think that’s the clever thing to do:
I accepted the idea that my intelligence was the most important part of me. I tried to be clever in everything I did. In the end, I was reluctant to attempt anything for fear of failure. My first thoughts never seemed good enough. Everything had to be corrected and brought into line.
Or, as one of my old bosses used to say:
Everyone thinks they can improve someone else’s copy.
There was certainly some of that going on at GrantTree. We were tempted to tinker because we thought we could fix the problems that we saw. Sure, other companies might have flawed pay models, but ours would be better. With hindsight, I’d say we set out on a fool’s errand. I think that it’s impossible to come up with a compensation model that isn’t wrong because pay is too fill of paradoxes.
Let’s begin with an innocent-looking assumption: you would like to have a salary model that pays people in a way that fairly reflects the value of their work. This sounds immensely reasonably, and free of contradiction — but you’re already in the minefield.
What does ‘fair’ mean? Perhaps it means “fair in relation to the market”? In other words, if someone is a junior salesperson at your company, you’d like to pay them a wage that’s about the same as what they’d earn at another business of a similar size in the same sector. Let’s assume for a moment that equivalence is easy to define, and some comparable companies exist. In that case, your model should be straightforward to construct: you just need to do a good market benchmarking exercise. It might take a bit of time, but it’s a fairly standard research project, and once you’ve done it you can just base your answers on your data. Great!
But hang on, maybe that’s not what you mean by ‘fair’. Maybe ‘fairness’ is about paying people relative to their contribution within the company? In which case, you can’t follow the dictates of the wider market — because the wider market puts a premium on certain kinds of work that might not reflect how much value that work generates for your business. The market typically values sales work higher than delivery work, for instance. By ‘delivery work’ I mean making the product or delivering the service that your company sells. Why is this? It seems to be an example of “what gets measured gets managed”. The easier something is to measure, the lower the cognitive load in assigning it a value, and we human beings optimise for low cognitive load. Sales people appear to have very easily measurable work:
total worth of contracts signed = total value generated.
But this isn’t true. You don’t make money from sales alone. You make money when the product is in the customer’s hands, and you can send them their invoice. Value-generation involves both sales and delivery. So, should sales and delivery roles be paid more equally to reflect this?¹
Each of these approaches has its own difficulties (finding relevant benchmark data is not always easy; it’s not clear how to measure people’s contribution to value-generation), but they’re not insurmountable. With some effort you can make either approach internally consistent. What you cannot do is reconcile these two obvious and reasonable definitions of ‘valuing someone’s work fairly’. You have to chose to pay people unfairly relative to the marketplace, or unfairly relative to others inside your company. Either way, you will in some sense have made a bad decision.
This is just one of the paradoxes of pay. There are many others. Some of them come from the more touch-feely senses of ‘value’. We all have different relationships to money and it represents many different things to us. Consider a scenario where you bring in a new pay model and it will mean some people instantly get quite a big raise and some don’t. Someone who has several children to support will respond very differently to getting (or not) the big raise than someone who has no dependents. When a raise represents “more security for my family” it means something different to “more money for gadgets and subscriptions”. So even if you could come up with a system of compensation that’s rationally fair (which you can’t), it will still feel unfair to money of your team.
Here are some more quickfire pay contradictions:
- Does ‘experience’ matter in a VUCA world? And if it is, what does it consist of (other than age)?
- If you have a distributed work-force, does local cost of living matter?
- How should managers be paid relative to individual contributors?
- As you invest in someone’s learning, they develop new skills and their market worth goes up. How should you value this arrangement?
(If you can think of others to add to this least, please contact me and I’ll add them.)
So that’s why there are only wrong ways of doing pay. There are too many important things which contradict each other. What I haven’t told you yet, which I promised I would, is why I think this is not a bad thing.
I’m kind of a Buddhist. Well, I’m more Buddhish. I’m not a peace, and love, and universal harmony kind of Buddhist (although these are important things in their place). I am the kind of Buddhist that gets a kind of giggly joy from remembering that suffering is an inescapable part of life. This isn’t out of masochism. I don’t take an especial delight in being tortured. Rather, I find it liberating to know that I’m not unhappy because there’s something wrong with me². I’m unhappy because I’m alive, and experiencing desire, aversion and indifference. These are three emotions we can’t avoid without shutting ourselves away from the world. If, like me, you don’t to live as a hermit, then you have a lifetime of dissatisfaction ahead of you. No matter how many problems we solve, there will still be more problems. That doesn’t mean that individual problems aren’t worth working on — problem-solving is a big part of what human life is — but it does mean we can never fix the problem of ‘having problems’. There is always going to be something wrong. If it seems like there isn’t right now, and things are perfect, then… just wait a few moments. That old and familiar companion, trouble, will be along any moment. (Now I’ve got the giggles again. Sorry.)
“That’s great,” I imagine you saying. “But what has that got to do with pay?” Well, what I offer you from my own sometimes painful experience is: accept that there are only wrong ways of doing pay. If you come up with a pay model, and it looks like it needs work, this may not be because you have done a bad job. Every pay model has problems. Just like in life, some of these may be worth fixing, but even if you fix them, there will still be (big, important) problems left³. Maybe you can chose to live with the problems that you have, rather than working hard to replace them with some different problems? If you can accept that, you will make the business of working on compensation much less stressful — which would surely be a positive thing?
 This is an example of the same phenomenon that means bankers are paid more than nurses. Financial transactions is easy to measure. Patient care is not.
 Sam Harris goes even further than me, and takes pleasure in suffering because it shows you are alive. In his excellent meditation app, Waking Up, it’s almost a catchphrase of his to say “you could be dead!”. In other words, no matter how bad things are now, if you were dead you wouldn’t even be able to suffer them, and that would be worse. I’m nowhere near at that level of equanimity, but it’s encouraging to know some people are.
 If you’re left asking: OK, I accept this, but how do I get started on building an actual model? Allow me to recommend three excellent articles from other people that will point you in the right direction: