Competency levels and pay bands are two different ladders

The sentence arrives forty minutes into a pay calibration meeting. "He is a Level 3 now, so he should be in band 4." Heads nod. It sounds like consistency. It is the moment a competency framework turns into a second job evaluation system, one with no factors, no weights and no audit trail.

A level is a statement about a person. A band is a statement about a job. The moment the first decides the second, you have stopped paying for work and started paying for a rating.

What breaks when a level picks a band

Grade drift. The CIPD's factsheet on pay structures notes that in narrow-graded systems, ten grades or more, most people reach the top of their range quickly, which "can lead to demands for upgrading and cause 'grade drift' (jobs being ranked more highly than justified)". When a level can move a band, every level review becomes an upgrade request. Nobody decides to inflate the structure. It inflates one defensible case at a time.

Pay equity exposure. Section 3 of the Code on Wages, 2019 bars gender discrimination in wages between employees doing the same work, or work of a similar nature, for the same employer. Section 2(v) defines similar work by the skill, effort, experience and responsibility it requires under similar conditions, not by title. So if two people do the same work and one sits a band higher because a manager argued a level, the band is the artifact you will be asked to explain. "The competency level said so" is not an answer about job value. The law does not currently compel a pay audit, which is why nobody finds the problem before a claim does.

The development conversation dies. Once a level is worth a band, no manager rates honestly downward and no employee hears feedback as feedback. "Sometimes works without supervision" is now worth four lakh a year, so everyone claims it, and your assessment evidence becomes advocacy.

You built the framework to describe capability. Within two cycles it describes claims.

The two ladders measure different things

The failure in both directions is the same: one structure answering the other's question.

Price the job before you place the person

The band comes from the work, not from whoever does it. That means a job evaluation you can produce on request: factors, weights, benchmark roles, and the date you last looked. Point-factor scoring is the common form. A simpler factor comparison is fine if the factors and weights are written down and applied the same way to every role.

Two numbers matter more than the method.

The span is the percentage difference between a band's minimum and maximum. The CIPD's example: a range of £30,000 to £45,000 has a 50% span. Wider spans carry more progression potential, which is what lets you hold a level inside a band instead of inventing a grade for it.

Overlap between bands is normal. The top of one band can sit above the bottom of the next, because an experienced person at the top is worth more than a new appointee learning the grade above. Bands stacked end to end with no overlap force every level change into a band change, and you have designed the problem into the structure.

The CIPD's 2019 reward management survey found 67% of employers use a pay structure. If you have none, build the band before the descriptors.

Make a level a position, not a promotion

The band is fixed by the job. The level decides where inside the band the person sits, and nothing else.

Write that mapping before anyone is in the room, as a range penetration or compa-ratio ceiling per level. Level 1 up to 85% of the band maximum, level 2 up to 95%, level 3 up to the maximum, anything above it needing a market exception with a review date. Then a level change is a pay move inside a band with a hard ceiling, not a move to a new band with a new ceiling.

North Carolina's state career-banding model runs three competency levels per broad job class: Contributing, Journey and Advanced. The class fixes the pay range. The levels describe what you would see, hear or measure at that level. The state's glossary is explicit that competency-based pay "revolves around the journey level, and compensation for a banded class is built off of the journey level". A person's level sorts them within a range. It does not reclassify the job.

Levels defined in observable terms can be argued with evidence. A range set by the class cannot be moved by a level review.

Write the test for a band move

A band move needs a job-content change. The test is three questions, all weighted equally, and all three must be answered yes.

1. Scope. Does the role now own more of something countable: people, budget, accounts, machines, sites, a service window?
2. Accountability. Has a decision moved into this role that used to sit above it, or has an approval threshold changed in this role's favour?
3. Consequence. If the role fails now, is the loss larger than a year ago, and can you name it?

If all three are no, the person changed and the job did not. Pay can move inside the band. The band cannot.

Write the answer down with a date and a name, because a "no" that lives only in conversation is a "yes" by the next cycle. Update the job description in the same sitting: a band move that leaves the JD untouched is a pay decision dressed as a design decision.

If your levels were built as one result per level, the evidence question is already settled and the review is short. If every level claims four outcomes, managers rate on whichever protects the person. The mechanics are in one result, one level.

Keep the two conversations in separate rooms

Structural rules lose to meeting room dynamics, so separate the meetings.

The level review has managers and L&D, dated observed examples, and the level descriptors. No salary sheet, no band table. Its output is a level per competency and a development action with an owner.

The compensation cycle has reward and finance, band ranges, position in band, budget and performance. Competency ratings are not an input to the number. If a manager wants to argue a level, the level was settled elsewhere with different evidence.

Space the two by at least six weeks. Put the level review two weeks before salary letters and every manager knows what the level is for, whatever the agenda says.

Tooling matters less than the separation, but it helps when one system holds the framework, the evidence and the band range and can still restrict what each forum sees. That is what the mapping in the platform is for. If your spreadsheet cannot, use two.

Worked example: 240 people, six bands

Illustrative numbers for an equipment services company. Take the shape, not the values.

Ravi is a B2 technician, six years in, rated level 3 on fault diagnosis. The old practice slots him into B3, raising his ceiling by ₹2.2 lakh for work he already does. Here his level 3 puts him at 93% of the B2 maximum: ₹5.6 lakh against a 4.2 to 6.2 range. To reach B3 he takes the warranty escalation queue and signs off other technicians' work orders. Scope and accountability change first, then the band.

Meera is a B4 service lead hired at ₹8.4 lakh, rated level 2, with the market for her role now at ₹9.6 lakh. Her manager proposes level 3 so she can be paid more. That fixes nothing and misdescribes her work. Re-price B4 on current market data instead: the ceiling lifts for everyone in B4 and shows as a budget line, not a favour. Pay problems from a mispriced band get fixed in the band.

The honest limit

Someone will eventually sit at the top of their band and the top of their levels. Three options are honest: redesign the job with real scope and move the band, pay a one-time award that does not consolidate, or accept that they will leave and plan for it. Adding a fourth level to retain them is the option everyone takes; it works for about twelve months and leaves a definition problem you cannot withdraw.

The mapping also assumes jobs differ in content. In a forty-person firm where everyone does a bit of everything, job evaluation is partly fiction and six bands is invented precision. Fewer bands, wider spans, market pricing with a written rationale and a log of exceptions is the more honest design at that size. Precision you cannot defend is worse than a range you can.

And it assumes the market pays for jobs. Where it pays for scarce skills, one PLC platform or one certification, you will sometimes pay above band for someone who has not changed job. Record that as a market exception with a reason and an expiry, because left as a level it becomes permanent, unowned, and the first thing a new HR head finds in an audit.

It costs a meeting, too. Two forums is real load for a three-person HR team. If you cannot staff both, run one annual band review and keep the level review as a written record that never touches money.

Do this before your next cycle

Pull every band change from the last twelve months. For each, write one line: what changed in the job. Do not accept "he grew into the role" or "she is a level 3 now". Count the lines where the only thing that changed was the person.

Divide by the total. That is your grade drift rate, and it is the number for the next compensation review. Not the number of levels, not the average rating: the proportion of band moves you cannot defend on job content. Above one in five, stop writing level descriptors until the bands are priced, because the framework is doing work it was never designed for.

The framework does not need rescuing. It needs the right meeting, the right evidence, and a clear statement that it has no authority over the salary range. Then the levels do what you bought them for: telling people what good looks like, in a room where nobody is negotiating.