Performance & talent reviews

A review cycle yourmanagers don’t dread.

Most performance processes are a form nobody wanted to fill in, feeding a rating nobody trusts, producing a conversation nobody enjoyed.

Book a call
The problem

Everyone’s a four

The pattern is always the same. Managers avoid the hard ratings because they’ll have to defend them, so almost everyone lands in the middle. The genuinely excellent get the same score as the coasting, which insults one and reassures the other.

Why the ratings stop meaning anything →

Then pay decisions get made on something else entirely, because nobody actually believes the ratings. Underneath that is a simpler problem. If you haven’t agreed what good looks like at each level, a manager rating someone is just recording an opinion. You can’t calibrate opinions.

What is calibration in performance reviews?

Calibration is a session where managers compare and defend their proposed ratings against each other before anything is finalised. It exists to stop one team rating generously and another harshly, and it is the part of the process that produces real decisions about promotion, pay and underperformance.

How I solve it

Tie it to your levels, then make calibration real

Performance hangs off your levelling framework. Once a manager can point at what a senior engineer is expected to do, a rating stops being a personal judgement and becomes a comparison against something written down.

Then calibration, which is where the value actually sits. Managers in a room, defending their ratings to each other, with someone in the chair who will push. That session usually surfaces two things: a strong performer nobody outside their team had heard of, and a problem that’s been quietly tolerated for a year.

Talent review grid — output of calibration
GrowerHigh potential, still building. Invest.
High potentialStretch them before someone else does.
StarRetention risk. Know what keeps them.
InconsistentSupport, or an honest conversation.
CoreThe engine room. Most of your team.
High performerDeliver reliably. Reward properly.
UnderperformingAct now. This is what gets tolerated.
EffectiveSolid at level, no immediate move.
SpecialistDeep expertise, may not want to progress.
← PotentialPerformance →

Illustrative. Yours is built around your levels and language.

Timeline

Designed in four weeks, run in six

Performance framework build and first cycle
Week 1What’s broken now, what you need to know
Weeks 2–3Framework tied to levels, rating scale, cycle design
Week 4Manager training and comms
Weeks 5–8Cycle runs: self, manager, peer input
Week 9Calibration sessions
Week 10Outcomes, pay decisions, talent review
What’s included

What you end up with

  • Performance framework tied to your levels rather than floating free
  • Review cycle designed for the size you are, not the size you’ll be
  • Forms and questions short enough that people answer them honestly
  • Calibration sessions facilitated properly, at least the first time
  • Talent and succession view for the roles that would hurt to lose
  • Manager training on feedback that changes something
  • Clear link between performance outcomes and pay decisions
  • Configured in your HRIS so it runs without a spreadsheet
Questions

What founders ask

How often should we run reviews?

Twice a year works for most scale-ups, with lighter check-ins between. Annual is too slow at your growth rate. Quarterly usually collapses under its own weight.

Do we need ratings at all?

You need something comparable if performance is going to inform pay. It doesn’t have to be a number, but “we’ll just have conversations” tends to mean the loudest managers get the biggest budgets.

What about 360 feedback?

Useful once managers can already give direct feedback well. Introduced too early it becomes a popularity contest with extra admin.

Should performance be linked to pay?

Linked, not welded. Ratings should inform pay decisions alongside band position and market movement. When the two are welded together, people stop being honest in reviews.

We’re forty people. Do we need this?

You need something light. A full framework at forty is overkill, but “the founder knows everyone” stops being true somewhere around sixty and it’s better to have started before then.

Next step

Find out who’s actually doing well.

Book a call