On performance management

In the last post, we outlined how to attract, select, and onboard talent. In this post, we will examine aspects of performance management.

Why do we need performance management (when it's actually expensive to implement)?

• Dunning-Kreuger effect: Novices are more confident than they should be, while experts are less confident than they ought to be. People have inflated perception of themselves. Performance evaluation brings this to light. 
• Helps identify who deserves to be promoted, given bonuses, needs development etc. 

How does performance management help?

• Increases motivation and engagement (there is a sense of equity)
• Results in learning and development
• Insight into current talent's skills and competencies
• Better clarity on the state of company's goals
• Can be used to identify discrepancies between strategy and talent (happens often enough with disruptive companies: Blockbuster vs Netflix, Borders vs Amazon)
• Bad performance management can result in outcomes contrary to above, as well as risk of litigation. 

What are good practices for measuring performance?

• Identify several dimensions of performance (based on goals and behaviors)
• Assign scores to each (5-point scale is good: most of the variability of 7-point is captured, while a 3-point scale causes people to feel that it's harder to progress)
• Get a weighted score of the dimensions, to obtain a single overall number
• Do this often, and not just once a year.

What is the scale used for rating? Two fundamental approaches: 

• Absolute rating scale: all employees are compared to a standard
• Comparative rating scale: employees are compared to each other 
• Stack rank
• Order by the single overall number listed above
• Pair-wise comparisons is one way to stack rank: order by number of wins
• Forced distribution approach (10% low, 70% middle, 20% high)
• Jack Welch thought this was kind :) For employees in the low bracket:
• Gives them opportunity to improve
• Gives them a head start for job search
• So, which is better?
• 8% better performance with comparative ratings (research here)
• We work harder when comparisons are done
• Partly because incentives are clear cut between different levels
• Top performers better rewarded, and more motivated 
• Companies do not fare well switching from absolute to forced distribution
• Initial performance improvement, then drops precipitously 
• People get demotivated after the initial rounds of appraisal
• Forced distribution could result in:
• Lack of collaboration (drag others down to get ahead). Seen as a zero-sum game, and highly political process. 
• Unethical practices like cheating, cooking numbers, sabotage etc. 
• Abets extreme risks, especially if rewards are skewed to top 
• Perception of inequity (especially when the curve causes you to push down people), and accompanied low morale
• Another popular system is the 360 degree feedback; you can see a detailed post here. 

What are some rater errors in performance management? 

• Availability error: We overemphasize very available information (typically tends to be recent, or highly marketed). 
• Make sure you track performance over the entire year 
• More frequent feedback helps
• Leniency error: Tendency to inflate errors (especially with absolute rating scale). 
• We want to be liked; reluctant to deliver bad news, avoid escalation
• Central tendency: avoid extremes (no bad news for low performers, no promises for high performers)
• Top performers do not get separated easily 
• Attribution errors: Needlessly stringent ratings, and development plan might not work because of underlying reasons like lack of business unit support etc. 
• Sample size errors: Not having many data points results in extreme judgements. 
• Halo error: Having a good perception in one dimension causes us to rate a candidate good on several other dimensions. 

How to handle rater errors?

• Rater training
• Educate about rater errors, so errors are minimized
• Have a facilitator walk raters through examples, and point discrepancies
• Frame of reference training 
• Establish a common standard of what is good/bad/ugly
• Research here shows what errors are reduced by:
• Rater training: Halo, leniency, better rating accuracy
• Frame of reference training: better rating and observational (how we measure performance) accuracy

The effect of stereotypes

• Gender 
• Women are penalized, especially when they are below 20% of population (research here). 
• Women are perceived more negatively than men when they are autocratic (because they are "supposed" to be nurturing). Gives lesser options to women in terms of styles of leadership. 
• Women receive lesser salary than males. 
• Racial 
• African-Americans receive lesser salary increase than whites (research here)
• Beautiful
• People with "attractive" faces get 3% more salary (research here)
• Players with symmetric faces receive higher salaries
• Order of magnitude higher bias than gender
• Halo factor plays a role here
• Framing the conversation in terms of "Group A has an unfair advantage over group B" helps move discussion forward (as opposed to saying "Group B has a disadvantage over group A"; see research here).