Organizations and Hiring in Japan: A Reading Guide

You approved the headcount in spring. By autumn the seat is still empty, or it was filled and the person already gave notice. HQ asks for a better recruiter or a bigger salary band.

Usually the recruiter isn't the problem, and neither is the band. Most hiring and retention problems in Japan are design problems. I put this column together for executives who own a team here, whether from HQ or on the ground. Define and score the role before you score anyone. Once people join, the reporting lines, the team size and the boss decide whether they stay. Pay matters less than most people expect. It removes complaints. It doesn't create commitment.

How do I find and judge the right people?

Start with the role. Hackman and Oldham score a job on five dimensions, and a role that scores low keeps nobody, however good the hire. Then use a talent map to decide which requirement to drop when the market can't supply all of them.

• Score the role before you score the person: the five dimensions, and what to redesign when a role scores low.
• A talent map is not a list. Here's the question it answers. Requirements multiply, and companies sit in three rings around your target.
• Audi's First Year in F1, Read Through Org Design and Talent Mapping walks through a public case: dual leadership, three talent pools and one very hard hire.
• Your resume is a request to be put in a box: the resume screen is mostly sorting, and the org boxes you draw shape who gets through.
• After the resume screen: what I look for in an interview. Evidence over adjectives: conditions, trade-offs, results and costs.
• Reference checks in Japan: what you can actually learn: former employers usually confirm employment and little else, so plan the check before the offer.
• Why do queues get long all at once? A simple queueing model, with a live simulation, explains why recruiting replies slow down suddenly and why slack isn't waste.
• Disney's 66 Character.AI hires: a talent map block: How Disney's deal for about 65 Character.AI staff reads on a talent map, and when hiring individuals, buying a team, building inside or waiting makes sense.
• Can foreign hires become site engineers in Japan?: Daito Trust is training foreign graduates as site supervisors in Japan, and whether they become core engineers depends on the path and the site manager.
• The two maps: who competes for your customers and people: the market map shows who competes with you and how they operate, and the talent map shows where the people are and which requirement to relax first.
• Does this person fit? Veto, weakest layer, five levels: break the role into the same four layers as the person, check the veto items, then the weakest layer, then give one of five verdicts. Look back at day 90.
• The Marshmallow Test Failed. So Does Hiring on One Trait: The 2018 replication shrank the marshmallow test's predictive power to a tenth of a standard deviation. A case for using single-trait judgments as a veto, never as a ranking.
• Fit or No Fit Is the Wrong Question in a Hiring Meeting: A composite hiring meeting stuck on fit or no fit, rerun with four-layer scores and five levels, shows the dispute collapsing into one fixable layer.
• Why Interview Scores Drift to the Middle, and How to Stop It: Interview scores pile up at 3 for four reasons, and each has a fix, from behaviorally anchored scales to calibration sessions that look at every interviewer's score distribution.
• Reverse Talent Maps: Read a Company by Where Leavers Go: Where a company's former employees went, whether they moved up and when they left reveals its internal bottlenecks and real talent competitors better than any exit interview.
• Five Reasonable Hiring Requirements Left Ten People in Japan: Five reasonable hiring requirements, multiplied through Japan's official statistics, leave about ten likely candidates a year. Count the pool before writing the job description.
• Why Recruiters in Japan Keep Sending the Same Candidates: Two thirds of people in Japan who want a new job are not looking, so agency shortlists overlap; a sharper brief and a talent map reach the rest.
• In-House Recruiters or Agencies in Japan? Split by Role: Whether to recruit in-house or through agencies in Japan is a per-role decision, set by talent scarcity, hiring volume and whether the search can be public.
• Can't Find the Right Hire? Which Requirement to Relax First: When the search stalls, relax requirements by how fast they can be learned and how costly a miss is: industry years and tools first, judgment and values never.
• Same Person, New Company: Why Performance Doesn't Travel: Studies of star analysts and cardiac surgeons show that part of a track record stays with the old employer, so interviews should compare the conditions behind past success with yours.
• No Relevant Experience? What Employers Should Check Instead: Experience predicts performance mainly in the first few years, so when a candidate has none, employers can check learning speed, how they take apart an unfamiliar problem, and the inner layers.
• Degree, Employer, Title: Three Easy Screens, Little Proof: Degree, last employer and title are the fastest resume screens and the least validated; this piece shows what each one removes, what it lets through, and what to check directly.
• Culture Fit Is the Laziest Reason to Reject a Candidate: Splits "not a culture fit" into a values clash, a working-style mismatch and plain unfamiliarity, using Rivera's study and a 172-study meta-analysis, and says what feedback should record instead.
• Why Being Overqualified Gets Candidates Vetoed: Employers reject overqualified candidates out of three fears: quick exits, boredom, and team disruption. Research supports parts of each, and only a verified Drive or Friction veto justifies a no.
• Why a Good Recruiter Sometimes Tells You Not to Move: Why a contingency recruiter in Japan would ever advise a candidate to stay: refunds, public six-month leaver counts and a two-year rule, plus how an employer can read that behavior.
• Real Failure or Safe Failure? Reading the Interview Answer: Most failure stories told in interviews are chosen to be safe. Listen for the situation, the candidate's own scope, what they would do differently, and whether the logic holds.
• How to Score Modesty in Japanese Candidates: Self-assessments sit on different scales across cultures. Score Japanese candidates on facts and scope, leave self-ratings out, and apply one ruler to everyone you interview.
• Hiring Side-Job Talent in Japan: Contract, Hours, Reviews: Hire a side-job worker in Japan as an employee and the first hour can be overtime; how to set the contract, hours, confidentiality and reviews.
• How Employee Referrals Work Inside Japanese Companies: How an employee referral moves through a Japanese company, why the bonus must be paid as wages, and what research says about referred hires staying longer.
• Is Their Japanese Enough? How Employers Judge Engineers: The JLPT N1 tests reading and listening only; the work situations employers in Japan actually judge, and how to test each in an interview.

Closing the hire, and the months after

Japan's 65.1% offer decline rate for new graduates counts students who hold offers, so it can't be compared with offers issued, and mid-career hiring has no common figure. Treat the close as its own stage. After the start date, remember that probation gives you less room than many foreign managers assume.

• Reading Japan's 65% Offer Decline Rate: what the number counts and how to run the close.
• Probation in Japan doesn't mean you can let someone go. Declining to confirm a hire is legally a dismissal, notice applies after day 14, and fixed-term contracts don't get you around it.
• “It was all me”: finding what a candidate really did: How to turn a candidate's claim that they did it all into scoreable facts by probing personal actions, decisions, numbers, failure and other people's roles.
• AI Can Tidy Your Interview Notes. It Can't Make the Call: Which parts of interview note-taking can go to AI, why judging what a candidate actually did stays with the interviewer, and what Japan's privacy rules require.
• Write Down Your Hiring Predictions. Check Them at Day 90: Why hiring judgment rarely improves with practice, and how to make it checkable: up to three observable predictions written at the offer and reviewed on days 30, 60 and 90.

Why didn't the raise keep them?

Herzberg separated the things that cause complaints from the things that create commitment. Pay belongs to the first group. Raise it enough and the complaints stop. Engagement doesn't follow, which is why counter offers so often fail.

• A 50% raise, and he still left: an engineer takes a 50% counter offer and leaves anyway, read through Herzberg, self-determination theory and a pay meta-analysis.
• Pay removes complaints. It doesn't create commitment. The two-factor theory itself, and what it means for retention budgets.
• Motivation is a design decision: autonomy, competence and relatedness each map to something you draw, such as reporting lines, the grade ladder and team size.
• Four kinds of relationships, and why pay design goes wrong. Fiske's models explain why pay fails when the rules of one relationship are applied to another.
• Why exit surveys never tell you why people left: 48% of leavers say family, 13% mean it, and the survey only reaches the outer layers of a person.
• Universe 25: The Mice Had Food. They Ran Out of Roles: Calhoun's Universe 25 reread from the original paper and its critics, and what it suggests about organizations where headcount outgrows real roles.
• Three-track pay: set grade, allowance and bonus separately: traditional grades mix how mature a person is with how heavy the role is. Split them, and grade sets base pay, modules set the allowance and results set the bonus.
• Lifetime Employment Is Not a Japanese Tradition: Lifetime employment was named by an American in 1958 and built by large firms to keep skilled workers. By a 2006 estimate, under one in ten had it.
• Manager and Specialist Tracks in Japan Need One Grade Scale: Why specialist tracks in Japan turn into holding pens when pay steps sit on titles, and how one grade scale with a leadership allowance keeps both tracks equal at the top.
• First-year performance reviews: don't chase fairness: In year one a new company has no facts to grade. Write expectations, record facts monthly, hold one conversation, and leave grades and curves for year two.
• Evaluating Foreign Staff in Japan: Fix Expectations First: Foreign employees in Japan are often rated against an unwritten picture of a good employee. Writing expectations as behaviors and calibrating raters fixes more than a new form.
• Reading Engagement Scores in Japan: What to Check First: Gallup puts Japan at 8% engaged, partly a matter of answering style. When your own score falls, check who answered and what changed before debating causes.
• Layoffs at a Japan Subsidiary: Where the Exit Process Breaks: Foreign companies cutting jobs in Japan rarely fail at the first meeting. They fail after the employee says no. Case law, exit papers and steps that hold up.
• Parental Leave Returners in Japan: Setting Rating and Role: Japanese law bans penalizing parental leave but sets no method for rating the months away. How employers should write return, review and promotion rules.

What shape should a Japan entity take as it grows?

A new entity is a simple structure. Run it with head office's machinery and it suffocates. Redraw the chart at 5, 15, 50 and 150 people, because at each of those sizes the way relationships hold together changes.

Japan adds its own twist. Nakane observed that Japanese teams group by place more than by qualification, so global functional lines that pull people out of their place tend to break.

• Don't run a startup-stage entity like head office: Mintzberg's configurations, and what happens when HQ imposes its own.
• 5, 15, 50, 150: when to redraw the org chart. An entity growing to 150 redraws its chart three times.
• Why global matrix structures break in Japan: fix the place first, then draw the functional lines.
• Conway's law: draw the team before the product runs the law backwards: decide the product shape, then draw the team that will build it.
• Two people at the top: when does it actually work? Co-led companies tend to win big or lose big; here is what decides which.
• Add people, and each one pulls less: eight people on a rope pull half of what they would alone, and the fix is one owner per module.
• Fifty meters apart, conversations stop. The Allen curve still holds for remote teams, which makes seating a design decision.
• Pointing and calling: a check nobody can see isn't a check: A Japanese railway experiment cut errors to one sixth by making checks visible. Why engineering teams in Japan insist on steps that someone else can verify.
• The 1.4-minute Shinkansen and what Japanese customers expect: The Shinkansen averages 1.4 minutes of delay a year, disasters included. Where Japanese customers' expectations for SLAs and incident handling come from.
• The Oriental Witches, Read as an Organization: Japan's 1964 volleyball gold, won by textile workers, read as an org case: a closed talent pool, designed training, one decision left to the players, and a planned end date.
• Seven org symptoms and the layer each comes from: Seven organizational symptoms you can check against your own company, each mapped to the layer it usually comes from and the first question worth asking.
• Grades Are Not Titles: A Minimum Grade Table for Japan: A six-step first grade table for a small Japan entity, why grades should follow people and titles follow seats, and why a section-head title does not exempt overtime.
• After a Foreign Acquisition: The Order of HR Integration: After a foreign acquisition in Japan, integrate evaluation and bonus first, converge grades slowly, and change retirement allowances last and only for future service.
• Yakushoku Teinen: Designing Roles for Managers After 55: About one in five Japanese firms removes managers from posts at a set age. The years before 60 need a designed role, with pay following the responsibilities left.
• Transfers in Japan: How Employers Decide Who Moves: Japanese employers can usually order a transfer without consent. Disputes start when they skip the weeks between the naiji and the formal order. Cases and data.

Is there one method behind these articles?

Yes. I start with two maps, one of the market and one of the talent, then work through product, sales and organization, and set grades and pay last. People and companies get read with the same four layers. The pages below write the method down, including the parts I haven't proven yet.

• No guessing: how I read markets, people and organizations: the whole method on one page. Market and talent maps come first, then product, sales and organization, and grades and pay last, with a note on what is still unproven.
• Ground rules: where facts come from and who decides: the ground rules for where facts come from, how conclusions get written and how systems get set. AI never makes the hiring call.
• Four layers: one ruler for people and organizations: core, patterns, equipment and context, used for a person and for a company alike. Core can veto a fit, and patterns set the ceiling.

Where does influence actually sit?

The org chart shows how a company wants information to flow. Who owes whom, who carries messages between HQ and the local team, and whether bad news reaches the top explain more of how decisions really get made.

• The people who aren't on the org chart: reporting line versus influence line, and how to find the people who sway decisions.
• The favor ledger is more honest than the org chart. Mauss's give, receive, repay, applied to the unwritten ledger every Japanese office keeps.
• Your bilingual staff are not translators: multilingual staff stand on the biggest structural hole, between HQ and local, and that seat deserves a formal position.
• When bad news stops traveling up: Edmondson's psychological safety, measured by whether bad news reaches you, with resignation-proxy data as its inverse.
• Fujitsu stopped setting graduate pay by degree: Fujitsu now sets graduate starting pay by job, and the real lesson is that the role definition has to come before the pay table.
• Cool Biz: How a Nudge Changed Japan's Office Dress Code: Japan dropped the office tie without a law. Cool Biz read as a rollout: a name, a role model, numbers, written exceptions and clients, and where company rules in Japan usually fail.
• CoDMON Shut Everything Down Before the PHP Patch Landed: CoDMON took every service offline 30 minutes after learning the PHP patch would come too late. Values are only tested on the day they cost revenue; decide in advance who can hit stop.
• Before a reorg, count what each person loses: Reorgs usually fail in execution. A five-part loss ledger for each affected person, and the order that keeps resistance low: talk, announce, then switch permissions.

If you only have time for two, read the role-scoring piece and the one on why the raise didn't work. Together they cover the two moments when teams most often lose people: before the hire, and at the first counter offer. Much of this research comes from large Western companies. For a ten-person Japan office, I hold the findings more loosely.

Definitions for talent map, the three rings and the four-layer model are in the glossary. If you are still setting up, start with Launching in Japan.