Playing To Win
It’s Time to Accept that Pay for Performance Doesn’t Work
I am frequently asked about strategy and pay for performance. By and large, the questions aren’t at all neutral. The tend to be phrased as: Isn’t it really important to have pay for performance to get your strategy executed? It isn’t, as I will discuss in my 37th Playing to Win/Practitioner Insights (PTW/PI) piece: It’s Time to Accept that Pay for Performance Doesn’t Work. (Links for the rest of the PTW/PI series can be found here.)
An Article of Faith
We have long taken it as an article of faith that if you want performance against your strategy, you have to pay for it by way of incentive compensation. Hundreds of millions of person-hours get spent every year designing Pay for Performance (PfP) systems, then arguing about the design, then measuring performance against PfP systems, then arguing about the measurement, then paying out against the PfP system, and then arguing about the payouts. And then more hours the next year on redesigning PfP because it never seems to work out the way we thought it would.
The logical assumption is that PfP just has to work. Employees will give greater effort if the results of that effort are tied directly to greater levels of compensation, right? That is the way humans work, isn’t it? And it would be so embarrassing if all those person-hours spent designing and operating PfP systems were in vain, we don’t even want to contemplate that. No: if you want to make sure that the things in your strategy actually get done, you have to provide strong incentives through PfP!
My Crumbling Faith
My faith in PfP began to crumble during my work with Harvard Business School (HBS) professor Michael Jensen in the mid-1990s. As the father of agency theory, Mike had important views on how managers responded to incentives, and I wanted to integrate those thoughts into my approach to strategy. As part of that work, he introduced me to one of the most interesting articles I have ever read, Quota Restriction and Goldbricking in a Machine Shop, published in 1952. It was written by a young scholar named Donald Roy who went to work for 11 months as a drill operator in a shop that machined steel parts as a hands-on way to collect data for his PhD dissertation — if that isn’t commitment, I don’t know what is!
Michael liked the article because it demonstrated clearly that if you put kinks in the curve that related performance to compensation, you will drive performance to the kinks as is illustrated in the chart above, which shows a vast majority of the outcomes to be clustered around two kinks. One is around the least work you can get away with to earn minimum daily day without being fired (24% of observations). The other (47%) is around the highest pay you can earn without risking being the worker that caused the ‘standards department’ to rerate the task to make it less lucrative, thus becoming the least popular worker in the shop.
My central takeaway was somewhat different than Michael’s. In the article, Roy estimated that workers were putting in four or few hours of real work per eight-hour shift to optimally game the system for their own benefit by alternatively not working at all and earning the floor pay when the machining jobs were particularly tough (“stinkers”) and working slowly to earn the maximum safe pay when jobswere particularly easy (“gravy”). The productivity cost to the shop was astronomical. Plus, it created an environment of cheating and covering up, an environment into which Roy was indoctrinated the minute he showed up.
Then in 2001, I reconnected with one of my HBS professors, Malcolm Salter, to work on a project. One day as we were working together, he casually mentioned that there had never been a rigorous study which demonstrated a beneficial impact of monetary incentive compensation on organizational performance. I pushed back but he was adamant. I decided to check with him 20 years later and in an email response to me last week he said: “Amazingly, I think my earlier comment on the lack of rigorous research on a positive relationship between financial incentives and organizational performance still stands.”
A dozen years ago, an article in the 2009 Review of Economic Studies, Large Stakes and Big Mistakes, provided the coup de grace to my crumbling faith. Dan Ariely, Uri Gneezy, George Loewenstein, and Nina Mazar demonstrated that for tasks involving the brain (e.g., problem-solving), performance declined as incentive compensation increased. It turns out that incentive compensation only produces better performance in tasks that are primarily physical in nature (e.g., repetitive button-pushing) — not the staple of the modern manager’s work agenda.
Pay for Performance Won’t Help Bring Your Strategy to Fruition
PfP simply is not a useful tool. The problems with it overwhelm the good that it hypothetically produces. There are five big problems that in combination are just too big for PfP to overcome.
First, PfP is an extremely blunt instrument. Roy’s machine shop illustrates it well. The incentive doesn’t skew behavior a bit: it skews behavior immensely. Almost half of the total observations are in a narrow band around the rerate line. And it isn’t even a management-defined line. It is the guesstimate of workers as to at what point management might take deleterious action. Management isn’t even in control of the impacts of its own system. Wells Fargo provides yet another object lesson in bluntness. Workers did exactly what management wanted and gave them incentive compensation for: they opened accounts. Management just didn’t know the extent to which workers were willing to go to do exactly for what they were paying. And that cost Wells Fargo billions and worse yet, its reputation.
Second, there is evidence that when it comes to tasks of the mind, monetary incentive compensation creates choking behavior of the sort described in Large Stakes and Big Mistakes, rather than the desired superior performance
Third, it is really hard to coordinate the effects of incentives across entire organizations. At an individual worker level, incentive compensation might produce more of what the designers want, and not too much, but that is difficult to accomplish without creating problems elsewhere. Sales may well have an incentive to do something that makes life miserable for operations. Finance to do something that makes life miserable for customer service. And the conflicts that inevitably arise drown out any positive impact.
Fourth, the gaming is without limit. Roy’s machine shop is a poignant example. For every worker, half their day was spent doing nothing productive. Tanking a whole year to reset the budget at an easy level. Stuffing the distribution channel to make the quarter. Chopping advertising spending to jack up this year’s profitability. Diluting the ingredients in the termite spray at the end of the year to make budget — yup, freely admitted to me by field staff at a former client. Opening accounts that customers never approved. Recognize that there is no limit!
Fifth, you can’t fool customers. They figure out that they have a big target on their backs. Your incentive compensation ranks far above their satisfaction. They figure out that are merely a means to your end. But of course, they aren’t powerless. They know that you are trying to make budget to get your bonus and they can wrap you around their finger as year-end approaches. Monetary incentive compensation is toxic for customers. And that is even if it is about customer satisfaction scores. With two of my last three car purchases, the salesperson pleaded with me to give him a perfect satisfaction score to help with his compensation — resulting in me never intending to purchase a car from their companies (Lexus and Range Rover) again — great cars, but terrible customer experience thanks to monetary incentive compensation.
Reward in Other Ways
If you really want to encourage the behaviors that your strategy requires, reward it in other ways. Four Seasons Hotels & Resorts treats its employees the way it wants its guests to be treated. Employees get great eating and changing facilities, are treated with respect, are given high compensation, and are provided the best training and career development in the industry, and guess what, they treat guests the way they are treated, which results in Four Seasons earning the highest ratings for guest service in the industry.
Almost all the leading New York corporate law firms pay their partners explicitly for performance measured by a combination of hours billed and clients acquired. But one of the very best and most respected, Cravath, Swaine & Moore, has zero PfP. Every year, partners get paid the same as every other partner of equal partnership tenure, what is called lock-step compensation or most commonly the Cravath scale. It is because the Cravath strategy holds that cooperation among partners is the key to the best client service, which in turn is the key to keeping existing clients and earning new ones. And firm success combined with a pleasant and cooperative work environment is what will attract and retain the best partners. Law firm PfP works entirely against such a strategy.
Both the Four Seasons and Cravath approaches represent Enabling Management Systems designed to build and maintain the Must-Have Capabilities that enable the How to Win in the chosen Where to Play. They illustrate the productive way to tie compensation to strategy.
Don’t accept PfP as an article of faith. If you are on the receiving end, recognize the dangers and the downsides. You will have to determine your own stopping mechanism; the point at which doing more would earn you more incentive compensation, but you know it is bad for the business long-term. In addition, you will have to show self-control on gaming. The moment you start gaming the system, you create a schism between you and management. While they deserve it for being delusional, the price is that it will rot your soul and that isn’t worth it.
If you are in a position to put in place a PfP system, just say no! I am not going to give you advice on how to. Just don’t. I know that it is counter cultural. But you have a chance to avoid the traps that PfP creates. If you default to the blunt instrument of PfP, you will get blunt results — and you won’t like them. Instead, work on building a culture in which your people care about the customer and care about each other.