Merit matrix: how to build a merit increase matrix from compa ratio
A merit matrix is a grid that sets each person's raise from two inputs: their performance rating and their compa ratio. Performance goes down one axis, pay position against the range midpoint goes across the other, and the cell where they meet is the increase percentage. High performers paid below midpoint get the most; people already above midpoint get the least.
That is the whole mechanism, and it solves the two problems every merit cycle has. Managers stop negotiating raises one by one, and the total cost lands on your budget instead of 300 basis points over it. This guide covers the two axes, a worked grid that costs exactly 3.2 percent, what the 2026 benchmarks actually say, and the mistakes that make a matrix useless in year two.
What a merit matrix is, and what it needs first
A matrix only works if both axes are real. The performance axis needs a rating scale that managers actually apply differently, usually three to five levels. The pay axis needs a midpoint for every job, which means you need salary ranges before you need a matrix. If you do not have them yet, build the salary structure first; a merit matrix laid over invented midpoints just distributes money with extra steps.
Compa ratio is salary divided by midpoint. Someone earning $88,000 in a grade with a $100,000 midpoint has a compa ratio of 0.88. Someone earning $112,000 in the same grade is at 1.12. The whole point of putting that number on an axis is that it turns "she has been here five years and is underpaid" from an argument into a coordinate.
What is a good merit increase percentage for 2026?
For US employers, 3.2 percent is the number to build around. The three sources HR teams cite most are close enough to each other that the planning answer is unambiguous, and they all point at a flat year rather than the accelerated raises of 2022 and 2023.
| Source | 2026 US figure | What it measures |
|---|---|---|
| Mercer QuickPulse, October 2025 | 3.2% merit, 3.5% total | 1,013 US organizations across 15 industries, surveyed October 20 to 31, 2025. |
| WTW, January 2026 | 3.4% salary budget | Overall US salary budgets, described as stable rather than rising. |
| WorldatWork | 3.6% mean budget | Mean projected US salary increase budgets, a slight contraction on the prior year. |
The gap between the merit number (3.2 percent) and the total number (3.5 percent) is the part people miss. Total includes promotions, market adjustments and equity fixes. If you budget 3.5 percent and hand all of it to the matrix, you have nothing left to correct the two people your last pay equity audit flagged. Split the pool before you distribute it.
A worked merit matrix that costs 3.2 percent
Below is a four by four grid built for a 3.2 percent merit budget. Compa ratio runs across the top, performance rating down the side, and every cell is the base salary increase for that combination.
| Performance | Compa under 0.90 | 0.90 to 0.99 | 1.00 to 1.09 | 1.10 and up |
|---|---|---|---|---|
| Outstanding | 6.5% | 5.5% | 4.5% | 3.0% |
| Exceeds | 5.0% | 4.3% | 3.5% | 2.2% |
| Meets | 3.7% | 3.2% | 2.6% | 1.6% |
| Needs improvement | 0% | 0% | 0% | 0% |
Read one row and one column and the logic is obvious. An Outstanding performer at 0.85 compa gets 6.5 percent, because they are cheap and you would like to keep them. An Outstanding performer at 1.15 gets 3.0 percent, because they are already paid above midpoint and more base salary pushes them toward the range ceiling. Same rating, different money, and the reason is written down.
How to check the matrix costs what you budgeted
This is the step that separates a working matrix from a wish list, and it is arithmetic you can do in a spreadsheet in ten minutes. Multiply each cell by the share of your workforce that lands in it, add up the products, and compare the result to your budget.
Take a company where ratings come out at 10 percent Outstanding, 25 percent Exceeds, 60 percent Meets and 5 percent Needs improvement, and where compa ratios distribute as 20 percent below 0.90, 35 percent between 0.90 and 0.99, 30 percent between 1.00 and 1.09, and 15 percent at 1.10 or above. Run the grid above against that population and it costs 3.20 percent of payroll. On a $3.2 million salary base that is $102,456 against a $102,400 pool. Close enough to sign.
Change either distribution and the cost moves immediately. If managers rate 20 percent of people Outstanding instead of 10, the same grid costs noticeably more, which is why rating inflation is a budget problem before it is a fairness problem. Run the calculation with your real distributions before the matrix goes to managers, not after they have promised numbers in one-on-ones.
How to build your own in five steps
- Fix the midpoints first. Every compa ratio is only as good as the midpoint under it. Benchmark each job and level, then group them into grades.
- Set the pool. Decide what percentage of payroll the merit cycle gets, and separate it from promotion and market-adjustment money.
- Draft the grid. Four compa columns and three to five performance rows. Increases fall as you move right, rise as you move up.
- Cost it against your real distributions. Weighted average of every cell. Adjust the Meets row first, since it carries most of your people and most of the money.
- Write the exception rule. Who can go outside the grid, by how much, and who signs off. Without it, the exceptions become the policy within two cycles.
Where merit matrices go wrong
The most common failure is the one nobody logs: the performance axis is not real. If 85 percent of your people are rated Exceeds, the matrix has collapsed into a flat increase with extra paperwork, and the money stops doing any work. Before you blame the grid, get an honest read on whether the review process itself is producing usable signal, because a matrix cannot fix ratings that all say the same thing.
The second failure is treating the matrix as the whole answer to underpayment. A person at 0.82 compa ratio does not get to midpoint on a 6.5 percent raise; they get to 0.87. Structural gaps that big need a market adjustment funded outside the merit pool, and pretending otherwise is how pay compression survives three merit cycles in a row.
The third is applying one grid to every job family in a year when one family's market moved and the others did not. If engineering salaries jumped 8 percent and finance did not, no merit matrix reconciles that. Move the ranges for the family that moved, then run the matrix inside the new ranges.
Frequently asked questions
What is a merit matrix?
A merit matrix is a grid that sets each employee's raise percentage from two inputs: their performance rating and their compa ratio, meaning where their salary sits against the midpoint of their range. High performers who are paid below midpoint get the largest increases, and people already above midpoint get smaller ones.
How do you calculate a merit increase?
Find the employee's compa ratio by dividing their base salary by their range midpoint, then find their performance rating. Read the percentage where those two meet on the matrix and multiply it by their current salary. A person earning $88,000 in a grade with a $100,000 midpoint has a compa ratio of 0.88, so they sit in the below-90 column.
What is a good merit increase percentage for 2026?
For US employers, 3.2 percent is the merit budget most surveys land on for 2026. Mercer's October 2025 QuickPulse survey of 1,013 organizations reports merit increases of 3.2 percent and total increases of 3.5 percent. WTW put overall US salary budgets at 3.4 percent in January 2026.
What is the difference between a merit increase and a cost of living adjustment?
A merit increase varies by individual performance and pay position and comes out of a fixed pool. A cost of living adjustment is a flat percentage applied to everyone regardless of performance. Most US employers under 200 people run merit only, because a flat adjustment spends the same money without changing anyone's behavior.
Should someone at the top of their salary range still get a merit increase?
Give them a small one or none at all, and pay the difference as a bonus instead. Raising base pay above the range maximum red circles the employee and quietly breaks the structure. If your best people keep hitting the ceiling, the range is set too low or the job has outgrown its grade.
How do you use compa ratio in a merit matrix?
Compa ratio becomes the horizontal axis, usually split into four columns: below 0.90, 0.90 to 0.99, 1.00 to 1.09, and 1.10 and above. The percentages decrease as you move right, so the same performance rating buys a bigger raise for someone underpaid against midpoint than for someone already above it.
Next steps: work out where your people sit today with the compa ratio calculator, check the same question from the other direction with range penetration, and make sure the percentile your matrix pays toward matches your compensation philosophy.