How often should you run a performance review?

One cycle a year with a mid-year check is enough for most. In between, a monthly conversation between each person and their manager.

Should it be tied to the pay rise?

The result can inform the decision, but handle it in another meeting and at another time. If the two are mixed, the development conversation is lost.

How many competencies should you assess?

Five or six per person, chosen to suit the role. With more, those answering tire and the marks start to look alike.

What is the difference between performance and potential?

Performance describes what the person has done in the period. Potential is a hypothesis about what they could take on, and it is reviewed apart, on the 9-box grid.

Can the person see their review?

Yes. Article 15 of the GDPR gives them the right to access the data held about them. In a 360 review that does not include knowing who wrote each answer: paragraph 4 protects the rights of those who assessed them.

Performance review: methods, cycle and common mistakes

By Deibson Silva · Reviewed on 5 November 2026

A performance review is the process by which a company assesses, against written criteria and on fixed dates, how each person has worked and what they need to do it better. It works when it keeps the development conversation apart from the pay decision, calibrates ratings between managers and ends in a plan agreed with each person.

What is a performance review, and what is it for?

It is a conversation with a method. Once or twice a year, each person receives an assessment of their work based on criteria they knew in advance, and leaves with a plan for the months ahead.

It serves three different purposes, and it is wise not to mix them. So that the person knows how they are doing and where they can grow. So that the company decides with data who to prepare for other roles. And so that pay rises and promotions are shared out in a way that can be defended.

Where it is missing, those decisions are taken anyway, but from each manager's memory and from whatever happened last.

Which methods are there?

Methods differ in who assesses and in what is looked at. The table lists the six most widely used; nearly every company combines two or three.

Method	Who assesses	What it is for	Risk
Manager's assessment	Whoever manages the person	Having one clear yardstick and one conversation	It depends on a single point of view
Self-assessment	The person themselves	Preparing the conversation and seeing differences in perception	Used alone, it tells the company nothing
360 review	Manager, peers, direct reports and the person themselves	Seeing behaviour the manager does not see	Without anonymity and calibration, nobody answers frankly
By objectives	The manager, against agreed goals	Assessing measurable results	It rewards what is easy to measure and forgets the how
Continuous feedback	Manager and colleagues, through the year	Correcting in time, without waiting for the close	With no formal close, it leaves no record for deciding
Calibration	The management team, on ratings already given	Making the same mark mean the same in every area	Without rules, whoever insists most wins

360-degree feedback: how to design it so that it is useful

What is assessed: results, behaviour or potential?

All three, separately. Results say what the person achieved. Behaviour says how they achieved it, and explains whether they will be able to repeat it. Potential is a hypothesis about what they could take on later.

Behaviour is assessed through competencies, and each competency through a few indicators that someone has been able to observe. 'Shows initiative' cannot be assessed. 'Puts forward more than one option when raising a problem' can.

Potential is handled apart, on the 9-box grid, with its own conversation. If it is mixed with the year's performance, a person with one bad quarter is written off for everything, and another with one good year is given a role nobody prepared them for.

The 9-box grid: what it is and how to run it

What does a full cycle look like?

It has six steps, and the first two are done before anyone assesses anyone.

One, prepare: choose the competencies, the scale and who assesses whom, and fix the timetable. Two, tell: explain to staff what will be assessed, who will see what and what it will be used for. Three, collect the assessments in a short window of two or three weeks.

Four, calibrate: the management team reviews the ratings before anyone sees them. Five, give them back: each manager sits down with each person. Six, agree a development plan and close the cycle.

One cycle a year with a mid-year review works for most companies. What happens between cycles, the monthly conversations between each person and their manager, weighs more than the form.

Why do you have to calibrate?

Because two managers do not use the scale alike. One gives a four to whoever meets expectations and another keeps the four for whoever stands out. Without calibration, the mark depends more on who assesses you than on how you work.

In the session, managers of similar areas put their ratings side by side and explain the highest and the lowest with examples. When an example does not hold the mark up, the mark changes.

Three rules make it fair. The methodology is published beforehand. Every change is written down with its reason. And whoever runs the session does not adjust their own rating.

9-box grid template and how to calibrate it as a team

How do you give the result to the person?

In a one-hour conversation, with the report in front of you and no surprises. If the person hears for the first time in that meeting something the manager has known since March, the failing is the manager's.

Order helps. First, what the person did well, with two specific examples. Then one or two things that would change their work if they did them differently. Last, what each of the two is going to do: the person and the manager.

The development plan comes out of that. The usual mix is practice on the job, learning from other people and training, in that order of weight, signed by both.

The pay decision belongs in another meeting. When money and development are discussed together, the person only hears the figure.

What does the law ask for?

In Spain, Article 20.3 of the Workers' Statute allows the company to adopt measures to verify that work obligations are met, with due regard for the person's dignity. Article 64.5 entitles the works council to issue a report before systems for organising and monitoring work are introduced or revised, and Article 64.4.d to know the parameters of algorithms that affect employment decisions.

On data protection, each person must know what is assessed and why (Article 13 of the GDPR) and has the right to access the data held about them (Article 15). Article 22 gives them the right not to be subject to a decision based solely on automated processing that significantly affects them.

The Spanish data protection authority's list places the assessment of performance at work under its first criterion, and expects an impact assessment in most cases where two or more criteria are met.

And Annex III to the EU AI Act classifies as high-risk the AI systems used to evaluate people's performance at work. Its obligations apply from 2 December 2027. Other jurisdictions have their own employment rules.

Which mistakes should you avoid?

Assessing only what is recent. Without notes through the year, a twelve-month review reflects the last six weeks.

Using a scale with no labels. If a three has no written definition, each manager gives it their own.

Asking for too much. Fifteen competencies and forty questions produce the same mark for all of them. Five or six well-chosen competencies give you more.

Skipping calibration for lack of time. It is the step that makes results comparable, and without it promotions go back to being decided by closeness.

Closing the cycle without a plan. An assessment that does not end in something to do is remembered as paperwork.

Sources

Frequently asked questions

How often should you run a performance review?: One cycle a year with a mid-year check is enough for most. In between, a monthly conversation between each person and their manager.
Should it be tied to the pay rise?: The result can inform the decision, but handle it in another meeting and at another time. If the two are mixed, the development conversation is lost.
How many competencies should you assess?: Five or six per person, chosen to suit the role. With more, those answering tire and the marks start to look alike.
What is the difference between performance and potential?: Performance describes what the person has done in the period. Potential is a hypothesis about what they could take on, and it is reviewed apart, on the 9-box grid.
Can the person see their review?: Yes. Article 15 of the GDPR gives them the right to access the data held about them. In a 360 review that does not include knowing who wrote each answer: paragraph 4 protects the rights of those who assessed them.