“About half of everything is below average”.
When you think about that, it’s pretty obvious isn’t it? Well keep that thought in mind as we explore a widely misapplied and misunderstood tool of performance management – The Bell Curve.
Without even attempting to get into the impenetrable mathematics behind probability theory and distribution curves (mainly because they go above my head), suffice to say that in simple terms, if you plot a range of values you often get a shape known as a bell curve. Here’s an example below. (No sniggering at the back!)
Essentially, the most frequently-occurring values tend to be grouped towards the middle. So, if we were plotting the IQ of 1,000 people in a room, there would be a few with very low IQs (left hand tail of the curve), a few with very high IQs (right hand tail of the curve), with the bulk of people being grouped towards the middle. This causes the distinctive hump that indicates normal distribution. As the name suggests, this is NORMAL…
Pay attention, anyone who might be about to use bell curves to rank their staff!
Now I’d be punching above my weight if I pretended to understand the algebraic formulae involved in constructing these things, but I think the bell curve conveys a pretty clear image that even the untrained eye can interpret, whether or not you have a Nobel Prize for mathematics.
Nevertheless, some managers think it is a good idea to use bell curves as a performance management tool, seemingly hoping to get everyone above average. Ha! Well guess what, if performance is assessed relative to peers then some poor so-and-so is always going to be at the bottom. This is regardless of how good their actual performance is. If someone was assessing my ability at making curries against my neighbours, I reckon I’d come out pretty well, whereas if you pitted me against Madhur Jaffrey or my good friend Faz, I’d be that sorry loser stuck in the left hand tail of the bell curve. This is because assessing performance relative to peers is the same as benchmarking against a moving object, like a bus.
Pitting people against each other has emotional and psychological consequences too. Being branded a failure, marginalised or made to feel inferior comes with the territory and makes Joy in Work (Deming) impossible. Recent research into performance management by Professor Phil Taylor of the University of Strathclyde noted the pointed inferences that go hand-in-hand with the use of bell curves, such as:
“You have not achieved. You are an underachiever”.
Brilliant for morale and self-esteem, isn’t it?
In addition to the sense of abject failure associated with being last in the sports day egg and spoon race (or having the least number of arrests etc), ranking of individuals always causes an even more toxic side effect – unhealthy internalised competition. This is where effort and ingenuity are turned inward to get ahead of colleagues and avoid being last amongst equals. The result is that focus towards purpose becomes distorted and team spirit is destroyed, as individuals concentrate on their own performance at the expense of the overall system.
As you know, an effective system relies on its components working together to achieve its overall aim for the customer or service user. Pitting individuals, teams or departments against each other defeats this object, leading to sub optimisation, waste and worse performance.
Forced distribution intensifies the dysfunctional behaviour associated with managers using bell curves as a performance management tool – this is where a predetermined proportion of people will be singled out as a result of being in the bottom 10% or whatever. This is seen in the example below. There’s always going to be a bottom 10% isn’t there?
Now, if sanctions for ‘poor performance’ come as part of the package, you can guarantee things will get worse. Who wants to be on an action plan or the subject of critical management scrutiny? Likewise, at the other end of the scale, if incentives are introduced into the mix (such as performance-related pay) this further intensifies the bad stuff. In the same way as with numerical targets, this particularly counterproductive device of performance management always causes gaming, cheating and other dysfunctional behaviour.
Another major weakness is that, in the same way as those useless tables of numbers often used to judge performance, a single bell curve is nothing more than a snapshot of that moment in time. It’s likely that people will have changed position next time the music stops, or whenever another bell curve is drawn. Roll up! Roll up! Spin the wheel and you get different results every time! Such comparison against moving variables always produces unstable results. Therefore, it is impossible to see any trends or glean anything useful about performance that aids managers understand and improve how their system is operating.
On top of that, consider the limitations of what gets measured. If a limited number of inputs or outputs are counted (such as number of arrests, detected offences and intelligence submissions) then people will be incentivised to produce whatever it is that their bosses are measuring. What gets measured gets done, after all. Furthermore, how do you incorporate ‘valuable-but-difficult-to-measure’ activity into such a performance regime, such as comforting Mrs Miggins when she got lost in the town centre, or standing on a crime scene for hours?
Answer: You can’t.
In summary, the use of bell curves in performance management is based on defective assumptions about behaviour, ignores relativity, attempts to outflank the law of averages, causes dysfunctional behaviour, creates waste, demoralises staff, is incapable of identifying trends or anything useful about actual system performance, generates unhealthy competition, obscures purpose, disregards important but-difficult-to-measure activities, breaks down the critical interdependencies that the system thrives upon, and makes performance worse. What’s to like?
The antidote? Dialogue, knowing your staff, doing the right thing and ‘Right Measures, Measured Right’.
Using bell curves as a performance management tool is a classic example of ‘Wrong Measures, Measured Wrong’. Don’t do it! Keep this monster locked in the tower.
InspGuilfoyle 
Excellent post once again my friend. What worries me is what our “leaders” are going to do when we get rid of TORA, CORA, League Tables, compliance,detections month on month and year on year etc etc etc. As the service gets smaller it will be difficult to feed the hunger of ambition, up to now the currency of promotion has been reductions and detections however they could be achieved. Not sure that many in the service are ready to give up on this just yet as they will need something too prove their worth. Its easy to measure nunbers harder to work on improving the process to make the work work, especially if you have spent most of your time not involved in it!
On the positive side, though, there is endless opportunity for fun in telling credulous managers that a whole half of their staff are achieving at a below-average level, and watching their reaction.
ah, but which half?
Excellent once again but I’ve noticed that you get a lot of slaps on the back from the people who do the exact opposite of what you’re advising.
Also, its a massive assumption by management that whatever they are measuring IS normally distributed. Normal distribution (ND) is common in nature, but work processes aren’t natural.
This link below is a good exploration of that and makes the observation that ND isn’t actually normal.
http://www.qualitydigest.com/dec99/html/nonnormal.html
Thing is, arguing numbers back at people has never worked for me. They didn’t think in numbers in the first place, it’s a different argument I’d be using if I argue from numbers to that that formed their opinion in the first place.
I work in an organisation where some previous mighty “leader”, luckily long since gone, wanted every part of the organisation to have plans that showed how they were ALL going to be in the top quartile for their National indicators within 3 years.
Lucky no other Council thought the same eh? Imagine a top 25% of local authorities crammed with 100% of them. That would separate the winners from the losers.
Good point re ND and fascinating article. Cheers!
Loved the article. Does your agency have metrics for measuring team work? Rewarding team behaviour instead of individual/single successes?
Yes. We also have team vs team, area against area, force against force. Bell curves at every level imaginable! 😉
Simple mathematics when applying mean, as the method to measure success or failure. There is some slight sophistication in attempts at mode, median and range, but ultimately your point, whilst correct (in my mind), fails to address the over whelming believe that ‘rank and shank’ as it was known in Enron (note not a success story), is seen as a fast and effective motivational tool when seeking to deliver short term objectives.
And this is the issue for me, not that you are wrong, but why should they do it the hard and long way when fast and demonstrable ‘delivers’.
In pure mathematical terms we cannot all be above average, although I do recall it has been set as an ambition! Interesting reading today in the papers that hospitals with above average mortality rates are being targeted for action – good luck trying to rid us of above average!
Despite the mathematical anomaly we all know what is meant, the same with ‘rank and shank’ – basically get rid or improve the worst and the outcome will be improved overall performance.
The reality as evidenced in such esteemed publications as the Munro Review of Children’s Social Care, is what such action causes. My technical skills prohibit my showing, but please review the causal loop diagrams in the review which demonstrate how trying to achieve an outcome through a misjudged method, no matter how well intentioned can cause the opposite outcome.
So whilst it is easy to undermine the bell curve approach, there are those who will evidence its success. The challenge for those of us who see it as wrong is to provide positive evidence that our way works, and not just by dismissing others.
For whilst we (and here I accept your key role) can build a body of support, there will always be ambitious individuals who see the evidence of success the bell curve gives, to get them the outcomes they seek to achieve.
For me this is the difficulty of delivery, in the private sector the segmentation of stakeholders allows you to target and deliver against the majority/influential group. This may reward shareholders, executives in some cases and occasionally the customer but rarely is their a sustainable model to satisfy the on-going needs of the all. And here is the crux, there does not need to be, private business would like to make infinite profit, but it does not need to, as long as it makes enough to satisfy shareholders.
Compare that with the stakeholder requirements of the public sector. Our shareholders, customers, employees, employers, are all the same group – the public. Yet depending upon which agenda they take there expectations are completely different. As employers and ‘shareholders’ they want value for money, as employees they want fairness and rights and as customers they want the right level of service for them. There is no adequate service, that we will just accept because the public, in whatever guise they take deserve a service that strives to try and deliver (alas I think we will never achieve for all).
This is why system thinking needs to evidence and demonstrate a better service delivery model, that not only sounds good and feels good, but can be seen to good.
For this reason I am pleased to see a growth in public agencies taking a purposeful systemic approach to doing what matters, more police forces, health and partnerships and hope that evidence of their successes can provide that positive evidence, alongside the evidence that the bell curve does not deliver sustainable results.
Thanks for your comments Nick. I suppose the reason they should do it the hard and long way is to avoid the consequences listed in the last paragraph of my post. I guess that virtually any ‘motivational’ tool, used aggressively enough can produce visible short term results. Look at the horrific examples of slave labour from history, such as the Burma Railway, for the extreme end of that argument, for example.
I agree that many will only buy into the systems approach when they see it producing results that outstrip the conventional methods – its up to us to do that.
Cheers 🙂
You talked about football in a previous comment, and it’s a good example of how the ‘rank and shank’ approach is used to drive performance. Players who underperform are dropped, and new ones found to replace them. If you can secure more good players, then you should get a better team. Keep changing out the bottom 10% and your ‘average’ should improve.
Of course, there are some big ‘ifs’ with all this. It takes time to train up the new replacements, during which time the performance of the whole may suffer. The new ones may not be any better than the ones you’ve dropped. If the manager is ‘underperforming’, then the same risks apply if you try to get a better replacement. Maybe it’s the strategy/tactics that are at fault, and although your players are individually all good, their styles may not work well together.
Then there are the longer-term side-effects. Minimal team loyalty if you think you’ll get sold off after a couple of bad games. The temptation to cheat or play dirty if your job depends on achieving certain results. Ever-increasing pay levels for ‘top’ players and managers.
But how do you apply this approach to service industry, and to public services in particular. How do you determine what is good performance? The police officer who makes the most arrests or issues the most speeding tickets? The division that reduces crime the most? I remember some performance ‘experts’ who once suggested that divisions with higher crime offered better value for money, because their officers were dealing with more incidents than officers in divisions with lower crime! They said we needed to cut the number of officers in the lower crime divisions so that officers there would have the same ‘productivity’ as those in the higher crime divisions. Fortunately I was able to point out that good productivity in policing was the absence of crime, not the efficiency of dealing with it.
If we sack the bottom 10% of teachers, we… start to run out of teachers. If we offer more pay to the top 10% of NHS managers, we… accuse them of getting fat-cat bonuses for letting patients down.
Ultimately, systems thinkers (particularly in the public sector) need to convince others that performance needs to be assessed in a wholly different way. Educating our children, reducing offending, solving problems, helping the sick feel better. Doing a good job. You can still draw a bell curve, but you’ll want it to help you identify what is normal, and what isn’t, where there may be a problem, and how you can help staff do a good job.
Hope you don’t mind the link to another site, but this blog post summarises the different focus well:
http://thinkpurpose.com/2013/02/09/4-pie-charts-you-wont-see-in-a-performance-report/
On a related note, from the 1st April 2013 the entirety of the Civil Service will become “less effective” than it was before, due to Civil Service Reform.
Click to access Civil-Service-Reform-Plan-acc-final.pdf
Guess Sir Bob Kerslake has’t read your blog!