Performance Management: Revision Notes

Performance management is the continuous process of aligning work with organisational goals, setting expectations, supporting and evaluating performance, and using the results for decisions. A performance appraisal is only the periodic evaluation event within it.

PM serves four purposes: strategic (translating strategy into goals), administrative (evidence for pay, promotion and exit), developmental (feedback and growth) and documentation. The administrative and developmental purposes conflict, because people avoid discussing weaknesses in a conversation that decides their pay; modern systems separate frequent development conversations from periodic evaluation.

Three findings explain why PM works when it does. Goal-setting theory (Locke and Latham): specific, challenging goals with commitment and feedback raise performance. Feedback research (Kluger and DeNisi, 1996): feedback helps on average, but in over a third of cases it reduced performance, mainly when it targeted the person rather than the task. Expectancy theory (Vroom): ratings must visibly affect rewards for PM to motivate.

In most Indian companies the cycle follows the financial year: goals are set in April–May, progress is monitored through check-ins with a mid-year review around October, ratings are reviewed and calibrated in March–April, and increments, bonuses, promotions and development plans follow in May–July. The employee owns goals and development, the manager sets expectations, coaches and evaluates, and HR designs the system, trains managers, runs calibration and checks fairness. A system needs clear goals, actions and consequences; without consequences it becomes paperwork.

A KRA is a broad area where results are expected; a KPI measures performance in it against a target. An OKR combines a qualitative objective with three to five measurable key results; OKRs are stretch goals (about 70% achievement is often a success), often public and often kept separate from pay. They originated with Andy Grove at Intel and reached Google through John Doerr in 1999. OKRs do not replace KPIs: KPIs track ongoing performance, OKRs drive a few changes.

SMART (Specific, Measurable, Achievable, Relevant, Time-bound) is the quality test for goals. Weak: “improve hiring”. SMART: “reduce time-to-fill for engineering roles from 65 to 45 days by 31 December without lowering 90-day retention”. Goals are cascaded from company to individual, ideally combined with employee-proposed goals, weighted by priority, and limited to four to six per person.

Results (the what) and competencies or values (the how) are rated separately and combined through a matrix. A person who exceeds results but falls short on behaviour should not receive the top rating or promotion until the behaviour changes. Rewarding results regardless of behaviour teaches everyone what the organisation really values.

The BSC exists because financial measures show past results but not future drivers; the airline example links ground-crew training to turnaround time, on-time departures, loyalty and revenue. 360s work best for development, with anonymity (at least three raters per group) and coaching. The 9-box was adapted from the GE–McKinsey portfolio matrix, not invented by McKinsey for people.

Forced distribution was popularised by Jack Welch’s vitality curve at GE (top 20%, vital 70%, bottom 10%). It prevents rating inflation but assumes teams are normally distributed, penalises strong teams and encourages competition with colleagues; Microsoft (2013) and GE, Accenture and Infosys (2015–16) moved away from it. Guided distribution applies a recommended spread across a large population and relies on calibration.

Calibration: managers submit ratings with evidence, HR prepares distribution data, managers discuss top, bottom and borderline cases, ratings are adjusted with recorded reasons, HR checks for bias, and managers explain final ratings. Risks: dominant voices and unfamiliar employees being disadvantaged.

About 62% of rating variance has been attributed to the rater’s own tendencies rather than the ratee’s performance (Scullen, Mount and Goff, 2000). The main errors are halo and horn (one trait colours the whole rating), recency, leniency, strictness and central tendency, similar-to-me bias, contrast effects and proximity bias against remote workers. Controls: specific goals and behavioural anchors, evidence recorded through the year, calibration, rater training and analysis of ratings by group and manager.

Effective feedback is specific, timely, task-focused, balanced and forward-looking. The SBI model (Center for Creative Leadership) states the Situation, the observable Behaviour and its Impact. Feedforward (Goldsmith) offers suggestions for the future instead of criticism of the past. Check-ins are short, frequent conversations about priorities, progress and obstacles. Coaching often follows GROW: Goal, Reality, Options, Will. Every move to check-ins depended on manager capability; without training, check-ins don’t happen.

Diagnose before acting: lack of ability needs training and support, low motivation needs a conversation about causes and consequences, and problems in the environment need the work system fixed. An IDP develops anyone towards future roles; a PIP (usually 60–90 days) brings an under-performer to the required standard through stated gaps, measurable targets, real support, documented reviews and clear consequences.

In India, termination of workers for misconduct must follow the standing orders and a domestic enquiry respecting natural justice; non-disciplinary discharge of a worker for poor performance may be treated as retrenchment, requiring notice and compensation. Managerial staff are governed mainly by their contracts, but a documented PIP remains important evidence of fairness.

PM outcomes feed increments (merit matrix of rating and compa-ratio), bonuses (pool × individual multiplier), development (IDP, training needs), promotions and succession (talent review with the 9-box). Promotion should depend on readiness for the next role, not on performance in the current one: promoting the best salesperson to manager without assessing leadership potential is the classic mistake.

Adobe (2012) replaced annual reviews costing about 80,000 manager hours a year with quarterly check-ins and no ratings. Microsoft dropped stack ranking in 2013. Deloitte (2015) replaced annual ratings, which took about 2 million hours a year, with performance snapshots in which team leaders answer four questions about their own intended actions towards each person. Accenture and GE dropped rankings and annual reviews in 2015–16, and Infosys removed its bell curve in 2016. Google moved in 2022 to GRAD, with one annual rating plus development check-ins.

The lasting change is more frequent conversation, not the absence of ratings: several companies reintroduced simple ratings to make pay decisions explainable. AI can summarise feedback and flag bias, but should not decide ratings.

Unionised workers are paid through long-term settlements; use team or line productivity incentives (output, quality, safety, attendance) negotiated with the union and skill-based progression. Sales uses objective metrics balanced against margin and customer quality. Project work aggregates feedback from several project managers. Hybrid teams need outcome-based goals and protection against proximity bias. Gig workers are managed by algorithmic ratings, which raises issues of transparency and deactivation appeals.

Organisation-level measures include revenue and profit per employee, human capital ROI = (revenue − (operating expenses − employee costs)) ÷ employee costs, early attrition = hires leaving within 90 or 365 days ÷ hires in that cohort (the compendium’s denominator is wrong), quality of hire, and rating distributions by group.