How this directory works

Where the facts come from

Four kinds of source, in this order of trust: MEXT's records of certified schools; each school's own mandated disclosures and website (the 告示基準44号 outcome form, the 45号 inspection report, the 18号 self-evaluation, fee and intake pages); open government data about the ward or city; and public learner discussion.

Which schools appear, and each one's status, comes only from the official lists: the Immigration Services Agency's list of notified schools and MEXT's register of certified schools. A school's own claims never set its status.

When a school is certified, its MEXT record replaces crawled values for every field both cover. Otherwise the newer of the school's own document and a correction the school sent wins. When two of a school's documents disagree, the newest dated one wins; a document's date is the creation date printed on it (作成年月日), or its file's date when none is printed.

The year a school has been operating since has its own order, which the “Most established” lens below sets out: its MEXT opening date, then the oldest edition of the immigration list in its unbroken run of editions, then the year its own website gives.

The list's past editions, from 11 March 2009, are read from their copies in the Internet Archive's Wayback Machine; editions published since are kept by the directory itself. A year from them reads “at least 2016”, since older editions may simply not be archived, or “2009 or earlier” for a school already on the oldest one. Each such year links to the edition it comes from.

Every figure on a school's page links to its source and says when it was last checked. Anything missing reads “Hasn't published this”: it is never hidden, estimated or filled with a default.

Where a school is placed

Each school is placed from its address: its MEXT record's when it is certified, otherwise the one its own website prints. The Digital Agency's abr-geocoder matches that address against the Address Base Registry (アドレス・ベース・レジストリ) and gives the registry's point for the school's block, house or parcel, or for its town when the registry has nothing finer.

A school the geocoder can place only at its ward, city or prefecture isn't placed at all, because that area's centre can be kilometres from the school: it gets no map dot and no nearest stations. Station distances are straight lines from the school's point, not walking routes.

How often it's checked

A script checks MEXT's records every week and republishes the pages of certified schools whose records changed.

Everything else is re-read when the operator re-runs the pipeline over the schools' sites: at least once a year, ahead of the main application season. A figure last checked more than 13 months ago is marked as such.

Lenses

A lens orders schools by one published fact. It ranks only the schools that published that fact; the rest are listed below the ranking as “Hasn't published this”, never ranked last. There is no overall score.

University track

Students who went on to further study ÷ (completers + early leavers who met the standard), from the school's newest 告示基準44号 outcome form: the same base as the form's own headline rate. Highest first.

A school whose newest form reports 0 completers has no graduates yet, so it has no share to rank.

Source: each school's own 告示基準44号 outcome form, the standard form every listed school must publish.

Work track

Students who moved to a work visa ÷ (completers + early leavers who met the standard), from the school's newest 告示基準44号 outcome form: the same base as the form's own headline rate. Highest first.

A school whose newest form reports 0 completers has no graduates yet, so it has no share to rank.

Source: each school's own 告示基準44号 outcome form, the standard form every listed school must publish.

Best value

The headline course's total tuition ÷ its class hours. Lowest first.

The headline course is the 2-year course, or the longest course when a school has no 2-year course. Its tuition counts tuition-bearing items including the entrance fee, tax included, and leaves out the screening fee and housing.

Source: MEXT's course records for certified schools, and each school's own fee pages otherwise.

Most established

The year the school has been operating since. Earliest first. The year comes from the first of these that has it:

1. The day the school opened, from its MEXT record (日本語教育機関 開設年月日), for certified schools. A day MEXT lists that hasn't come yet is a plan, not an opening, so it isn't used.

2. The oldest archived edition of the Immigration Services Agency's list of notified schools in the unbroken run of editions that name the school, going back from the current one. A rename counts only once the directory has confirmed it. Older editions may simply not be archived, so the school has been operating since at least that year, and may be older. The past editions are read from the Internet Archive's copies.

The oldest edition held is 11 March 2009 (平成21年3月11日). A school already on it reads “2009 or earlier”: the archive can't tell those schools apart, so they all tie at 2009.

3. The year the school gives for its own opening (開校, 設立 or 創立) on its own website. The founding year of the company or school corporation that runs the school doesn't count, and a year before 1950 or still to come is taken as a misreading and left out.

Sources: MEXT's records of certified schools, past editions of the Immigration Services Agency's list of notified schools (法務省告示) from 2009, read from the Internet Archive's copies, and each school's own website.

Most international classroom

The share of students outside the school's largest nationality group, from the latest annual report the school filed with MEXT. Highest first.

It describes who is in the classroom, not how good the teaching is.

Source: the annual reports certified schools file with MEXT. Schools that are listed but not certified don't file them.

• Ties are broken by the directory's record order.
• A value a school sent without a document is shown on its page, labelled as the school's own, but isn't ranked.
• Schools that are no longer listed aren't ranked.

Build your own ranking

You weight each lens's fact from “Don't care” (0) to “Essential” (4). For each fact you weight, a school's position among the schools that published it becomes a score from 1 (first) to 0 (last); tied schools share the midpoint.

A school's place is the weighted average of its scores. A fact it hasn't published counts as the middle position, halfway between first and last: never the worst value and never the best, so a school can't top a ranking on the one fact it published. Its row says “Hasn't published this” for that fact and is marked “Ranked on 2 of 3”. A school that published none of the facts you weighted isn't ranked.

Weights that are multiples of each other give the same order, so they share one link: “A lot” on two facts ranks the same as “A little” on both.

The fit finder

Where, when, for how long, the budget and the course goal are hard filters: they rule a school in or out. A school that hasn't published a fact a filter needs is listed under “Can't confirm”, never quietly passed or dropped.

What matters to you is a preference. The schools that pass every filter are ordered with the build-your-own ranking, each fact weighted equally, and marked “Ranked on X of Y” when they published only some of them. A school that published none of them is listed after the ranked schools, in its own group.

Entry expectations are shown as each school states them, where a document states them.

What learners say

A school's page summarizes public forum discussion only when there is enough of it: at least 3 threads, one of them from the last 36 months. The summaries are in our own words, grouped into themes with the thread count and dates, and each links back to its threads. Nothing a poster wrote is quoted, no username is shown, and nothing is scored. Where posters disagree, both sides are shown; where an official figure exists, the summary points to it.

Corrections from schools

A school can correct or complete its listing. It proves it controls its domain by clicking a confirmation link sent to an address published on its own verified website; the sender's address alone is never trusted.

A corrected value appears on the school's page labelled “Provided by the school”, with its date. It enters lenses, rankings and the fit finder only when a document backs it. A correction never changes a lens's formula.

A correction states the school's courses as numbers only: each course's length, total fees, class hours and start months. The link lasts 72 hours, and a confirmed correction is reviewed before it appears. When the school later publishes a newer dated document, the document replaces the correction, and the page says so.

Correct a school's listing

The placement estimate

The fit finder can estimate your level from six short sentences you put into Japanese, from N5 up to N3. An AI model reads them once and picks out the grammar you used; the estimate is the highest JLPT level of grammar your answers used correctly. It needs a free account, gives one estimate every 7 days, and keeps only its result, never your answers.

That JLPT level is shown with the CEFR levels in the JLPT's own reference table, below. The table maps a level and a total score, never a level alone, so the estimate shows the range a passer of that level can be shown at, and a course's stated CEFR entry level is compared with the lowest level in that range.

The table is reference information only. It covers reading and listening, not speaking or writing, and not everyone who passes N3 is at B1. The estimate reads what you wrote, so take its CEFR level as a guide.

This estimate isn't visa evidence. A student visa application needs a test certificate, such as a JLPT result, or a record of your hours of study.