Running your first appraisal cycle without it going wrong

A business ran its first formal appraisal cycle. Forms were distributed in March, returned in April, and collated by HR. Of forty-one employees, thirty-four were rated above expectations. Two of the seven rated as meeting expectations resigned within two months, both saying in their exit conversation that they had not realised there was a problem until the form.

Everything about the process had been followed. The form was well designed. The failure was not in the cycle; it was that the cycle was the first time anyone had said anything, and that the ratings meant nothing because nobody had defined what they were being compared against.

Decide what the cycle is for, before designing it

There are three plausible purposes and they pull in different directions. Trying to serve all three in a first cycle is the most common design error.

  DEVELOPMENT
    What should this person work on next?
    Needs: honesty, two-way conversation, no stakes.
    Ruined by: linking it to pay.

  DIFFERENTIATION
    Who is performing better than whom?
    Needs: comparability, calibration, evidence.
    Ruined by: managers protecting their own teams.

  PAY AND PROGRESSION
    Who gets what increase or promotion?
    Needs: a defensible basis, consistency, a budget.
    Ruined by: being decided before the ratings.

  A FIRST CYCLE SHOULD PICK ONE. USUALLY THE FIRST,
  UNLESS PAY DECISIONS ARE ALREADY BEING MADE
  ARBITRARILY - IN WHICH CASE START WITH THE THIRD.

The tension between development and pay is unavoidable rather than a design flaw to be solved. An employee who knows the conversation determines their increase will not volunteer weaknesses, and a manager who knows it will not be candid about them. That is rational on both sides.

Businesses that want both usually separate them in time — a development conversation in one part of the year, a pay decision in another, using but not conducted as the same conversation. That works, and it is more effort. A first cycle is usually better served by choosing one and being explicit about it.

If nothing was said all year, fix that first

The single most damaging thing a first appraisal can do is deliver news. An employee learning in March that their performance has been a concern since August has a legitimate complaint, and the business has lost the eight months in which something could have been done.

The rule that prevents this is worth stating as an absolute: nothing in an appraisal should be a surprise. If a manager is about to write something the person has not heard, the conversation happens before the form, not through it.

For a business with no habit of ongoing feedback, this means the first cycle should be preceded by managers having a conversation with each of their people — not a rating, just an honest current position. It delays the cycle by a few weeks and it prevents the two resignations in the opening example.

Ratings will inflate unless something stops them

Left alone, ratings cluster at the top. The reasons are entirely predictable and none of them are dishonesty.

• A low rating requires a conversation the manager does not want to have. A high one requires nothing.
• Managers advocate for their own teams, particularly where ratings affect pay, and a manager who rates strictly disadvantages their people against a colleague who does not.
• Nobody has defined what the middle rating means. Where 'meets expectations' is unspecified, it reads as a criticism, and managers avoid it accordingly.
• There is no comparison across teams, so each manager is rating against their own internal standard rather than a shared one.

The remedy is calibration: managers in a room, before ratings are final, discussing specific people against each other. It is uncomfortable, takes a couple of hours, and does more for the credibility of a performance process than anything else available.

It also works as training. A manager who has to justify a rating to peers, using evidence, learns quickly what evidence is needed — and rates differently the following year without anyone having to correct them.

What to train managers on

Most first cycles are supported by a briefing on how to complete the form. The form is not the difficult part.

1. 1

   Delivering a rating that is lower than expected

   The conversation almost every manager is dreading, and the one that determines whether the cycle damages anything. It is a practisable skill and half a day of role-play changes it materially.

2. 2

   Using evidence rather than impression

   "You've been unreliable" is an impression and it is arguable. Two specific instances with dates are evidence and they are not. Managers who have not gathered evidence during the year discover this at the worst moment.

3. 3

   Separating the person from the period

   An appraisal covers a period, not a character. A strong performer who had a difficult six months is not a weak performer, and treating a rating as a verdict on the individual is what makes these conversations so charged.

4. 4

   Handling disagreement without conceding the rating

   An employee who disagrees should be heard and the rating should not move because the conversation was difficult. A manager who revises under pressure teaches everyone how to get a better rating.

Where several managers are having their first difficult conversations in the same fortnight, it is worth treating as a cohort rather than individually — that is the population leadership and people management development exists for, and the timing before a first cycle is better than the timing after it.

A workable first-cycle sequence

Compressed into roughly eight weeks, assuming you are starting from nothing.

Weeks one and two: decide the purpose, define what each rating means in words, and confirm that every person has a manager who knows they are responsible for their review. That last check catches more problems than expected in businesses where reporting lines have drifted.

Weeks three and four: managers have an unrecorded honest conversation with each person about where they stand. No form, no rating. This is the step that gets cut for time and should not be.

Weeks five and six: forms completed, then calibration across managers before anything is shared. Ratings are provisional until calibration is done.

Weeks seven and eight: the review conversations, then a short written record of what was agreed for the next period. Two or three objectives, not eleven, and something the person can actually influence.

What makes the second cycle easier

The main thing is that objectives exist. A first cycle typically assesses a year nobody set expectations for, which is why it leans on impression. The second cycle has something to assess against, provided the first one ended with objectives written down.

The other thing is a running record. A manager who notes three or four specific observations per person per quarter — good and bad, one line each — arrives at the next cycle with evidence rather than recall, and recall is heavily weighted toward the last six weeks. This costs a manager about ten minutes a quarter per person and it is the highest-return habit in the whole performance management process.

Where the first cycle reveals that roles are not clearly enough defined to assess anyone against — which is common, and not a failure of the cycle — the prerequisite is role design and job descriptions rather than a better form. And where it reveals that ratings and pay have no consistent relationship, that is the finding to act on before the second cycle, because a performance management process whose outcomes do not visibly connect to anything will not survive two years of being taken seriously.