Performance for administrators
Setting up Performance so that the numbers coming out of it mean something. The configuration decisions here are, in effect, your company's definition of performance — most of them are made once and are awkward to change later.
Where: Performance › Settings You need: the admin right
The order to set it up in
- Skills list — define the capabilities you recognise
- Feedback topics and templates — what feedback is given against
- Assessment cycles — the questions
- Goal cycles — the targets
- Score card definitions — the weights that combine them
- User rights — who can do what
- Recurring cycles and automations — so it happens on schedule
Score cards come after cycles because they select cycles explicitly. Rights come late because you need to know what exists before deciding who can touch it.
The four decisions that determine everything
The score card weights. Self-assessment, reviewer assessment, goals and feedback each carry a weight, and together they produce every person's overall rating. This is the single most consequential configuration in the module — see Score cards.
Assessment question types. Multi-select questions award maximum to everyone and cannot discriminate. Any scored assessment containing them is inflated and flattened at the same time. Audit for this before every cycle — see How assessment scores are calculated.
Goal priority discipline. Critical counts four times Minor. If managers set everything Critical, the weighting is inert. This is a briefing problem, not a settings one.
Feedback confidentiality defaults. Confidential feedback is invisible to the person it is about. Decide your company's position and tell people, rather than leaving it to whoever fills the form — see Giving and receiving feedback.
Rights worth restricting
| Right | Why |
|---|---|
| Goal KPI override | Changes the measure a goal is scored against — after the fact, it changes the answer |
| Goal delete | Removes the record of what was agreed |
| Read/write all | Performance data is opinion about named individuals |
| Feedback allow public | Changes who sees what is said about someone |
Making it happen on time
Review processes fail by drifting, not by being badly designed. Recurring cycles create assessment and goal cycles on a cadence — weekly through to yearly — so the cycle happens whether or not anyone remembers.
Automations fire from an event rather than a date; the available trigger is a probation feedback request, which suits probation precisely because everyone's date is different. See Recurring cycles and automations.
Reviewing the output
Before anyone sees their rating, look at the distribution. If nearly everyone lands in one band, the cause is almost always the assessment design rather than your workforce. Fix it before the ratings are published, not after.
Then look at variation between reviewers. Two managers rating the same quality of work differently is the most common integrity problem in performance data, and it is invisible to the people affected.
Tips
- Pilot with one team before rolling out. Every design flaw appears in the first ten responses.
- Write down the weights and why they are what they are. You will be asked, by someone unhappy with their rating.
- Brief reviewers on the scale before the cycle closes, not after. A 4 means Exceeds, and most reviewers do not know that.
- Keep weights stable between periods. Changing them makes year-on-year comparison meaningless.
- Decide what the output is for — development, pay, promotion — and say so. People answer differently depending on the answer, and they are entitled to know.
- Do not let the module be the whole process. It records a conversation; it does not have one.
Troubleshooting / FAQ
Q: Everyone scored in a narrow band. Check the assessment for multi-select questions, and check whether goal priorities are being used.
Q: An employee has no score card. They were not in the selected cycles, or those cycles did not complete.
Q: A cycle ran but does not count toward score cards. Cycles are selected explicitly on the score card definition, not by date.
Q: Managers rate very differently from each other. Calibrate before the cycle closes. Nothing in the configuration fixes this.
Q: What does the verify skills permission do? Nothing at present — it is listed but not enforced.