Lukas Wallrich · Birkbeck, University of London
Visiting Lecturer · SWUFE · May 2026
Today, in three parts
What counts
What "good performance" means, and how we measure it.
Take it apart
Redesign one real US-tech appraisal form for another country.
Is it fair?
What we (should) actually reward: the result, or how it was achieved.
Chunk 1 of 3
What counts as good performance?
Before you can measure performance, you have to decide what you are even looking at.
Meet three engineers
A
Highest release and sales numbers on the team. A weak collaborator: others avoid working with them.
B
The team's strongest integrator. Holds projects together, but has weaker visible output of their own.
C
Technically brilliant, erratic. One innovation the firm adopted; one missed deadline that hurt.
Same firm, same China subsidiary, different performance profiles. You decide what it is worth.
Group Task小组任务05:00
· 2-3 people
Rank A · B · C — twice
First, for this year's bonus. Then again, for who you would invest development in. Give evidence for each ranking.
How we'll do it
Agree both orders as a group. They do not have to match.
Discussion only. Be ready to call your order out.
Rank them
A
Top release and sales numbers. A weak collaborator.
B
The strongest team integrator. Weaker visible output of their own.
C
Brilliant but erratic. One adopted innovation, one missed deadline.
05:00Two orders — bonus, then development. One line of evidence each, not adjectives.
Report-back反馈
Call it out
Your two orders: bonus, then development. Did they diverge? Why?
You could also share
A ranking your group argued about before agreeing.
Whether you were focused on output, behaviour, or potential.
There is no culture-neutral "performance"
The criterion problem. A rating quietly selects a mix of outputs, behaviours, potential, and values. Choose the mix and you have half-decided the score.
Subjectivity invites noise. The more subjective the criterion, the more a score reflects impression management, reputation, and politics rather than the work itself (Cardinaels & Feichter, 2021).
Your rankings differed because you chose different criteria — not because anyone was wrong.
Two jobs, one form
Control
Sort and allocate
Decide pay, promotion, and who exits. Looks backward; needs a defensible number.
Development
Help people grow
Build learning and skill. Looks forward; needs candour and honest weaknesses.
Candour drops the moment feedback sets pay (Lockyer & Sargeant, 2022). "Did it help me grow?" and "what was my bonus?" lead to different readings of a form - and different expectations.
Who rates whom?
Self
Rate yourself. Assumes you will speak up for your work.
Manager
The default. Their view usually sets your pay.
Peer
Colleagues judge people they depend on.
Upward
Reports rate the boss. Risky where hierarchy is steep.
360°
All sources at once. More views, more cost.
Numbers
Hard system data: sales, output, tickets. Counts, but no context.
Each source has its own bias, blind spots, and its own social risk.
Forced ranking is a tournament, not a scale
The defining feature. A required distribution: say 20% top, 70% middle, 10% bottom. Someone must land at the bottom.
It raises stress and gaming, worst where performance is subjective (Cardinaels & Feichter, 2021). Downgraded staff are 34%+ more likely to quit, so curves push out middle performers too (Bond, 2026).
The arc. GE's vitality curve made it famous → Microsoft drops it (2013) → Amazon tightens again (2025).
↗ ByteDance has a forced curve. More in that group's presentation.
360° feedback travels badly
Popular, for good reason. Ratings from every side promise a fuller picture than one boss alone.
High power distance. Rating a superior candidly feels improper, so those ratings inflate or go missing .
Collectivism and face. Naming a colleague's weakness in writing threatens harmony, so candour drops .
Appraisal that misfits the culture raises turnover . The hardest feedback to give runs upward.
Next: you take a real US-tech review process apart and rebuild it for a specific country.
Chunk 2 of 3
Take the process apart
One real review process, one country. Which features survive the move?
On the table: Google's review process
A real, documented system. Self-assessment, peer reviews, a manager assessment, calibration committees, a five-point impact rating, a "Googleyness" values rating, and a strong link to pay and promotion.
Built on individualist defaults. It rewards visibility, compares people against each other, and asks everyone to speak up for their own work.
Which features travel badly within this industry? How should they be adapted?
Every feature is a cultural bet
Self-promotion assumes people are comfortable advocating for their own work.
Calibration assumes ranking colleagues against each other reads as fair.
Peer comments assume candour about a colleague is socially safe.
"Living the values" assumes the values are shared and understandable for everyone.
The process makes an argument about what good work is and who can safely judge performance.
Redesign Lab小组任务18:00
· one country per group
Your country: China · Germany · Brazil
Open the worksheet. Read each of Google's features, then redesign the process for your country: change what does not fit, keep what does — and give a clear reason for every call.
Worksheet — the process is inside it; mark each feature in place:swufe-cchrm.pages.dev/s7lab
Report-back反馈60–90s per group
Share your thinking
Name your country, then your most important change and one feature you kept. Give the reason in a line.
Then let's count it up
Which features survived everywhere?
Which were rejected almost everywhere?
Was it culture — or institutions?
Some changes are culture. Directness, face, comfort with self-advocacy.
Some are institutions, not culture. Works councils, labour law, dismissal rules. Germany's co-determination is law, not "German culture."
The real bind. HQ wants comparable data → comparability needs standardisation → standardisation erases local meaning.
No feature is culture-free, but some travel better than others: clear criteria, a right of reply, transparent calibration.
You have redesigned how performance is judged.
Next: what we actually reward — and how much of "good performance" the person even controlled.
Chunk 3 of 3
What do we reward, and is it fair?
The result, or the way it was achieved? And how much of it was luck?
Behaviours or outcomes?
The "what"
Outcomes
Results, KPIs, the number. Easy to count, so pay usually follows it.
The "how"
Behaviours
Collaboration, mentoring, judgement. Harder to count, so it slips out of pay.
Many coaches have a no a*hole rule - but what does that mean in the workplace? And is it worth having?
Let's put a bounty on cobras
Goodhart's law
What happened. Colonial Delhi paid a bounty on dead cobras. People bred cobras to claim it; the scheme was scrapped; the now-worthless snakes were released, leaving more cobras than before.
The principle. "When a measure becomes a target, it ceases to be a good measure" .
It happens in firms. Wells Fargo's cross-selling targets drove staff to open about 3.5 million accounts customers never asked for (2017 estimate).
An outcome metric does not just measure work — it reshapes the work to hit the metric.
Pair-share配对分享05:00
· ~2 min each
Discuss in pairs
Can you think of a measure you have watched distort behaviour — gaokao prep, internship KPIs, app "engagement", step counts, citations, sales targets. What did people optimise instead of the real goal?
Three people, one bonus pool
Wei
The biggest number on the team: a ¥10M feature. But was handed the flagship project, a senior mentor, and an easy market.
Lan
The glue. Caught two failures before launch, mentored juniors, held the team through a reorg. No headline metric of her own.
Jun
The worst territory on the team. Solid numbers against a collapsing market, the most growth, and flagged a risk leadership ignored.
Same team, same level, one finite pool. Every yuan to one is a yuan from another.
Group Task小组任务10:00
· in threes
Divide ¥300,000
Split the pool across Wei, Lan, and Jun — three numbers that add to ¥300k, with one line of reasoning per share.
How we'll do it
Note down a clear split, with reasons.
Be ready to defend your split to the employees.
Report-back反馈
Read your split
Three numbers, and why. How much did you reward the result, the invisible contribution, and the effort against a bad hand?
One more question
Did you discount luck — the handed project, the easy vs collapsing market?
Luck egalitarianism: the idea
What you chose
Effort, ambition, the gambles you took. A fair share should track this.
What you didn't
The team, the territory, the passport (as well as talent, upbringing). A fair share should not hang on this.
"Ambition-sensitive, but endowment-insensitive" . The critique: taken literally it turns harsh, and effort is partly unchosen too .
…so what does it mean for reward?
How much of Wei's ¥10M did Wei earn? Pure outcome-pay rewards good brute luck (Wei) and punishes bad brute luck (Jun).
Should we weight what people control more? Effort and behaviours, not outcomes alone.
Build process people can see. Calibration, context adjustment, a right of reply. Communicate the logic of reward, e.g. benefit-sharing vs appreciation.
The goal is not a perfect formula — effort is partly luck too. It is to stop pretending the number is pure merit.
The limit case: the expat premium
We have met it before. In Session 5 we designed expat packages; in Session 6 we asked "equal pay — equal in what?"
Through this lens. A 4× pay gap driven partly by which passport you hold is brute luck, hard to defend as desert (应得 · what you have earned). (Chen et al., 2002)
What rescues it is not the number. It is a process people can see is fair (procedural justice), as that allows people to accept (some) distributive "injustice".
Reward is more than money
Recognition & growth
Public praise lands very differently across face and hierarchy norms. What counts as recognition varies.
Stability & wellbeing
Chinese Gen Z rates health and wellbeing highly, alongside pay and job security .
Ask yourself: what reward would actually satisfy you? What would make you stretch further?
Looking ahead
Next
Block 8: Labour Relations, Ethics & Institutions. Fairness is set by law and unions too, not only HR.
Reflection
An easy start: a way you have actually been evaluated or rewarded — did it help you grow, would it transfer? Full prompt on the assessment page.
Presentations
This session is relevant for most cases - where we will learn more from you.
Worksheet — the process is inside it; mark each feature in place:
swufe-cchrm.pages.dev/s7lab