Evaluating Foreign Staff in Japan: Fix Expectations First

When a Japanese company says its appraisal form doesn't work for foreign employees, the form is rarely what's broken. The usual trouble is that nobody wrote down what the company expects, and the managers doing the rating never compared notes. A new form fixes neither.

How are foreign employees evaluated in Japan today?

Mostly with the same system as everyone else, while their numbers climb fast.

According to employer filings collected by the Ministry of Health, Labour and Welfare (MHLW), 2,571,037 foreign nationals were working in Japan at the end of October 2025, up 11.7% in a year. Of those, 865,588 held "specialized and technical" residence statuses, the visas behind most engineering, sales and planning jobs. That group grew 20.4%.

On the company side, the most detailed public survey is older. In February and March 2018 the Japan Institute for Labour Policy and Training (JILPT) surveyed firms with 100 or more employees (2,260 valid responses) and asked how they manage highly skilled foreign staff. Among firms that hire or plan to hire such staff, 89.6% said they manage them like any other employee, with no special arrangements. Among the 10.5% that did have special arrangements, only 35.1% put the job content in writing at hiring.

Treating people the same is the right default. Japan's Labor Standards Act bans different pay or working conditions based on nationality. The catch is that a lot of the "same" evaluation in a Japanese company was never written down.

What does a cooperation score actually measure?

Often it measures how closely someone matches the company's unwritten picture of a good employee.

Japanese appraisal forms lean heavily on attitude items: cooperation, initiative, communication. The words are fine. What's missing is the behavior behind them. A cooperative person, in that unwritten picture, listens to the end before speaking, never contradicts the boss's plan in front of others, and raises objections one to one after the meeting. Japanese employees absorb this from their first week. A mid-career engineer from abroad has never seen the picture, and the manager rating him is holding it.

So the engineer who argues in meetings and ships on time gets a C for cooperation. The C is about meeting style.

MHLW's guidelines for employers of foreign workers say this fairly directly. Employers should clarify the qualities and abilities the workplace expects, and make the way evaluation, pay and placement decisions are made transparent. The guidelines don't say how. They point at the right spot, though: expectations that exist only in people's heads.

Japanese staff feel this too. In the worker part of the same 2018 JILPT survey, 6,028 employees said they were unhappy with their appraisals, and 67.0% of them picked "the criteria are unclear and arbitrary." That figure isn't split by nationality. Japanese employees fill the gap with years of reading the room. Foreign employees can't yet, so they hit the wall first.

Foreign residents' own answers point the same way, if faintly. In the Immigration Services Agency's FY2025 survey of 8,874 foreign residents, 5.4% of the 7,706 who were working said they were disadvantaged against Japanese colleagues in hiring, placement or promotion, 29.1% said pay was low, and 54.2% reported no particular problem at work. Across all respondents, 17.5% said they had faced discrimination while working. The survey asks about unfair treatment. It doesn't ask whether people know what is expected of them.

With expectations unwritten and raters uncalibrated, a foreign employee's low score starts to look like a nationality problem

Four states, by whether expectations are written as behaviors and whether raters share a standard. The framework is the author's judgment, drawing on MHLW's guidelines for employers of foreign workers and the JILPT survey (2018)

Across: expectations written as concrete behaviors (left: no, right: yes). Down: raters calibrated to a shared standard (top: yes, bottom: no)

Rated by feel, but consistently

Raters agree with each other, yet the standard lives off the page. Japanese staff fill the gap by reading the room. Foreign staff can't.

Every B can be explained

"Hear the plan out, then push back" is written down, and raters have practiced on examples. A low score points to a specific event, and the person knows what to change.

Low scores read as nationality

Nothing written, everyone rating by instinct. "Too direct" and "doesn't fit in" end up in the comments, nobody can say what to fix, and the person usually leaves.

Clear on paper, uneven in practice

Behaviors are written, but raters never practiced. Fluent Japanese lifts the whole score, and blunt speech gets docked under cooperation.

A new form doesn't move the bottom left. Write the expectations, calibrate the raters, touch the form last.

What raters need to learn first

The one skill that matters most is telling "this person is weak" apart from "this person does things differently from how we do them here."

I look at people in four layers: the core (drives and lines they won't cross), patterns under pressure, equipment (skills and experience), and context (this boss, this team, this stage of the company). As I wrote in the four-layer method, if you skip the context layer you can't separate a bad hire from a good person in the wrong seat.

Speaking bluntly in meetings is usually friction in the context layer. That's how meetings ran at his last company. Score it under cooperation and a context problem gets filed as a personal one. A year later, at the promotion meeting, it becomes "not sure he can lead a team." A habit one conversation could have fixed is now part of his record.

Two other biases show up a lot with foreign staff. One is a fluency halo: people with smooth Japanese get rated higher across the board, and people with halting Japanese sound as if they haven't thought things through. The other is similarity: the more someone resembles the rater, the better they score. Both happen between Japanese colleagues too. Nationality widens the gap.

Rater training can be short. Take three or four real appraisals, anonymize them, have several managers score them separately, then put the scores side by side. For every C, ask which event it refers to. A C with no event behind it is bias.

So should you change the form?

Keep the structure. Change two things.

Add a one-page expectations sheet per person: three to five items, each written as a visible behavior, ideally with an example. "Cooperation" becomes "when you disagree with a proposal, first confirm what the other person is trying to achieve, then offer an alternative." The Japanese habit of constant status updates becomes "if you fall two days behind, tell your lead the same day." Japanese staff would benefit from this page too.

Then pull Japanese language out of the other items. If the job truly needs Japanese, give it its own line and state the level: reading specs, or pushing back in a client meeting. If it doesn't, stop letting fluency leak into cooperation and communication scores.

The cost logic is simple. Replacing a white-collar hire who left because the appraisal made no sense means an agency fee, plus a stretch with the seat empty and the work landing on someone else. As I wrote in After you hire foreign staff in Japan, the hard part begins, most of the difficulty arrives after the start date. The expectations page takes an hour or two.

Some Japanese norms deserve to stay. Keep them, and say so in writing during the first month. What tends to lose good people is the opposite: a company that claims to treat everyone the same, then pulls out an unwritten standard once a year at review time.

---