Every appraisal cycle begins with a number from finance: the increment budget. Say it is 8% of the current payroll. That is a promise about the total, not about any individual. Turning one aggregate percentage into hundreds of individual raises, fairly and without overspending, is the job of a merit matrix.
Two questions, one grid
A simple approach would give each performance rating a fixed raise: 12% for top performers, 8% for solid ones, and so on. The problem is that this ignores where people already sit in their pay band. An excellent performer who is already paid 20% above market and an excellent performer paid 15% below would both get 12%, and the gap between them would grow every year.
A merit matrix asks two questions together: how well did this person perform, and how is their pay positioned against the midpoint of their band, measured by compa-ratio? The answer is read from a grid.
| Rating | Compa-ratio below 0.90 | 0.90 to 1.10 | Above 1.10 |
|---|---|---|---|
| Outstanding | 16% | 13% | 10% |
| Exceeds | 12% | 10% | 7% |
| Meets | 10% | 8% | 5% |
| Partially meets | 4% | 3% | 0% |
| Below expectations | 0% | 0% | 0% |
Reading across a row, the same performance earns less as pay rises relative to market. Reading down a column, better performance always earns more. The design goal is that over a few cycles, strong performers move toward and above the midpoint, average performers settle around it, and nobody drifts upward on tenure alone.
Testing the grid against the budget
A matrix is a hypothesis until it is costed. The cost depends on how payroll, not headcount, is distributed across the cells. Raises are percentages of salary, so a cell full of senior, expensive people costs more than the same number of juniors.
Suppose the ratings guidance expects 10% of payroll in Outstanding, 25% in Exceeds, 50% in Meets, 10% in Partially meets and 5% in Below. Within each rating, assume 30% of payroll sits below 0.90 compa-ratio, 45% in the middle and 25% above 1.10.
The average raise for each rating is then a weighted average across its row. For Meets: 0.30 × 10 + 0.45 × 8 + 0.25 × 5 = 7.85%. For Outstanding it works out to 13.15%, Exceeds 9.85%, Partially meets 2.55%, and Below zero. Weighting those by each rating’s share of payroll gives a total of about 7.96%. On a ₹50 crore payroll, that is roughly ₹3.98 crore against a budget of ₹4 crore. The grid fits.
The real distribution is only known after ratings are finalised, so the test is repeated then. In practice, this is a spreadsheet exercise: each employee’s rating and compa-ratio determine their cell, a lookup returns the percentage, and the sum of all raises is compared to the budget.
When ratings inflate
Managers often rate more generously than the guidance assumed. Suppose final ratings come in at 20% Outstanding, 35% Exceeds, 40% Meets, 5% Partially meets and nobody Below. With the same grid and the same compa-ratio mix, the total rises to about 9.35%, well over budget.
There are three ways to respond, each with consequences.
- Scale the whole grid down by roughly 8 ÷ 9.35, or about 0.86. Meets in the middle column becomes about 6.8%. Simple, but it quietly punishes the managers who rated honestly.
- Send ratings back for calibration so that the distribution is closer to the plan. This addresses the cause, though it is slower and can feel like forced ranking if done crudely.
- Give each department its own pot equal to 8% of its payroll and let department heads allocate within ranges. This contains overspend locally but allows inconsistency across teams.
Most organisations combine calibration with a modest scaling adjustment and publish ranges rather than single numbers in each cell, for example 7% to 9% for Meets in the middle column, so managers retain some discretion.
What the merit budget should not pay for
A merit matrix works only if the budget feeding it is truly for merit. Promotion increases, market corrections for roles where pay has fallen behind, and statutory changes such as minimum wage revisions should be funded from separate lines. When everything comes out of one 8% pot, a team with three promotions has less left for everyone else, and the matrix begins to look unfair even though the arithmetic is right.
Similarly, people above the band maximum usually receive any award as a one-time lump sum instead of a base increase, so their pay does not move further beyond the range. Those lump sums should be counted against the budget too.
Signs a matrix is not working
If the difference between the top and middle rating is only two percentage points, high performers may conclude the effort is not worth it; the matrix is too flat. If almost everyone lands in one column, bands are probably out of date and need re-benchmarking before the matrix can do anything useful. And if the same people sit in the “below 0.90” column year after year despite good ratings, the percentages are too small to move them, and a targeted correction is needed instead.
Written by the RP Strativa Team. The matrix, distributions and payroll figures are illustrative; each organisation sets its own.
Leave a Reply