The review that decides something

The problem it solves

The annual review arrives at year end as an administrative obligation. The supervisor remembers the last six weeks clearly, fills in the form with high marks to avoid an uncomfortable conversation, and the document is filed without anything changing in the operation. Whoever needed to hear a correction never heard it, and whoever held the team together all year gets the same score as someone who did not.

The form is not the problem. The problem is that nobody agreed in writing, at the start of the period, what doing that job well would look like. Without that prior agreement a review becomes an opinion about a person, and an opinion cannot be discussed with evidence. It gets discussed with feelings.

There is a second failure, quieter than the first. When every supervisor rates by their own yardstick, a four in one department does not mean what a four means in another. The organization believes it holds comparable data about its people; what it actually holds is several rating styles placed in the same table. The price comes due when a raise, a promotion or a separation has to be decided and the file says the opposite of what everyone knows. From then on the real decisions are made outside the system, the team notices, and the cycle lives on as a ritual that eats hours and decides nothing.

What it covers

The day builds the cycle in the order it happens: the expectations agreed at the start of the period, the follow-up through the year, the calibration between supervisors before anything is communicated, the review conversation, and the decisions that follow from it. The order is not a detail. Each stage rests on the one before, and no form, however well designed, can rescue a cycle that began without an agreement.

The work is done on the real roles of the organization in the room. Defining what acceptable performance means in one specific position is harder than filling in a generic competency matrix, and it is what holds the rest of the system up. On that footing the day draws the distinctions a form tends to blur: the work versus the person, results versus observable behavior, and what should stay out of the review even when it irritates the supervisor. It also looks at the biases that creep into any rating: recency, likeability, the halo effect, and the pull towards generous scores that spare everyone an argument. And it asks when self-assessment and multi-source input genuinely help, and when they only get in the way.

Most of the day is work in the room. Participants draft the expectations agreement for one of their own roles and write down, for every level of the scale, what would have to be observed to earn it. A full block goes to calibration, the piece that is almost always missing: supervisors comparing real cases, presented without names, until the scale means the same thing to all of them. The review conversation is rehearsed in pairs, including the version people dread most, in which the person does not accept the rating.

The last block is the question that decides whether the system survives. What will the company actually decide with the result? A review cycle that feeds no decision about development, assignment or pay becomes December's form again within two cycles. The answer does not fit inside a single engine, which is why ALUD works on it where People meets Finance, the point at which every decision about pay has to find room.

By then each participant has the pieces of the cycle fitted together: an expectations agreement that fits on one page, a scale in which every level states what has to be seen to earn it, the sequence of the conversation through to its close, and a calendar that puts calibration ahead of any announcement.

What participants learn

• Agree in writing, at the start of the period, what acceptable performance means in each role.
• Separate evaluating the work from evaluating the person.
• Calibrate criteria across supervisors so the scale means the same thing company-wide.
• Run the review conversation on evidence gathered through the period rather than on recent memory.
• Connect the result to concrete decisions about development, assignment or pay.

Agenda

1. 01Expectations written at the start of the period: the agreement everything rests on
2. 02What gets evaluated: results, observable behavior, and what should stay out
3. 03Rating scales and the pull towards generous scores that avoid conflict
4. 04Calibration between supervisors, worked on real cases
5. 05Recency, likeability and the halo effect in rating
6. 06The review conversation when the person disagrees
7. 07Self-assessment and multi-source input: when they help and when they get in the way
8. 08From result to decision: development, assignment and pay

What the organization leaves with

• A one-page expectations agreement format for each role.
• A rating scale with observable descriptors written for every level.
• A script for the review conversation, with its sequence and its closing points.
• A cycle calendar with the calibration session scheduled before any result is communicated.