Performance Reviews: The Methods and How to Choose

Most performance review processes fail before they start, at the moment someone picks a method based on what they saw at another company rather than on the question they need answered.

The methods are not interchangeable. Each answers something different, and using the wrong one produces an annual ritual that eats weeks and changes no decision.

The question first, the method second

Three different questions get crammed into the same process, and they should not be.

Did they hit the targets? An outcome question, answered with data that already exists.

How do they work? A behavior question. Two people can hit the same target and leave opposite situations behind them.

How far can they go? A potential question, and current performance does not answer it. An excellent analyst is not automatically a good leader of analysts.

Putting all three into one form and averaging produces a number that serves none of them.

The methods, and what each answers

Competency rating scales. The manager scores a defined set of behaviors. Answers the “how do they work” question well. Its known weakness is halo effect, where a general impression bleeds into every dimension. You mitigate it with written behavioral anchors, not adjectives.

360 feedback. Adds peers, reports, and internal clients. It is the best available method for observable behavior, because a manager only sees a slice of how someone works. It is expensive in time and degrades fast when tied to raises, because at that point people start scoring strategically. It works better decoupled from compensation.

OKRs or management by objectives. Answers the outcome question and nothing else. Precise when targets are measurable and honest. It breaks when the target depended on factors outside the person’s control.

Forced ranking. Requires distributing people along a curve. It solves rating inflation, which is a real problem. The cost is high and well documented: it erodes collaboration, because a colleague’s performance now affects your own rating. Most large companies that adopted it later dropped it.

Nine-box grid. Crosses performance with potential. It is not an evaluation method but a decision method, and it feeds off the others. Useful for succession and development planning, not for justifying a raise.

The most expensive mistake

Using one instrument to decide compensation and to develop people.

These are two conversations with opposite incentives. In the compensation conversation, the person needs to show the best of their year. In the development conversation, they need to talk about what they are bad at. Nobody does both in the same meeting, on the same form, in front of the same person who decides their raise.

Companies that separate the two get far more useful information out of the second. The first stays what it is, a compensation decision with explicit criteria.

Where potential comes in

Potential is the piece measured worst, and it usually ends up being the direct manager’s opinion.

Part of potential can be measured with an instrument. Stable personality traits predict performance in roles the person has not held yet, which is why they answer a question current performance cannot. A profile high in openness and emotional stability says something about how that person will handle a more ambiguous role, even if they are not in one today.

That does not replace the manager’s judgment. It adds a second data point, measured the same way for everyone, that can sit next to the first. When the two agree, the decision is easy. When they diverge, the conversation that opens is usually the most valuable one in the process.

Potential mapping covers how that second reading gets built, and succession planning with data covers how it gets used afterward.

A minimum process that works

If you are building this for the first time, four decisions are enough.

Separate compensation from development, into two different moments of the year. Pick four or five competencies with written anchors, not twelve. Add peer input only for the development conversation. And bring potential from an instrument, not an impression.

Everything else can come later. Starting with the full process of a ten-thousand-person company is the surest way to have nobody finish it in year one.