A renewal letter in a shrinking city

Picture the decision as it actually happens, at a kitchen table in Fukui city this spring. A household has kept the same internet provider since the dial-up years; the e-mail address on the family's bank correspondence ends in mitene.or.jp, and it has ended that way since before the children left for Osaka. The renewal letter offers the local product, mitene光, at 4,950 yen a month for a detached-house gigabit line on a two-year commitment. Television advertising and the electronics shop by the station push the national alternative: docomo hikari on exactly the same NTT West fibre at 5,720 yen, with up to 1,210 yen a month shaved off each family smartphone bill as a bundle reward. A household with two docomo handsets would come out ahead by switching; a household with none would pay 770 yen a month more for the privilege of a Tokyo call centre. The wire into the house would not change in either case. Only the name on the invoice would.

That kitchen-table calculation, multiplied across a prefecture, is the entire economics of mitene internet Co., Ltd. — and the prefecture doing the multiplying is getting smaller at a record pace. The 2025 national census counted 729,386 residents in Fukui, down 37,477 people or 4.89 per cent in five years, the prefecture's steepest decline since modern counting began and its first drop below 750,000 since 1947. The prefectural statistics office's monthly estimate has kept falling since: 724,385 people and 296,188 households as of June 1, 2026, with natural decrease running at several hundred people a month. Every one of those 296,188 households is a potential broadband account, and each year there are fewer of them, and the ones that remain skew older — which cuts both ways, because older households are the least likely to chase a bundle discount and the most likely to keep an e-mail address they have printed on twenty years of new-year cards.

This essay works out what that loyalty is worth in yen, using only what the record supports: the company's own published tariffs, its parent's securities filing, the telecom ministry's wholesale statistics, the address and routing registries, and the operational traces a working network cannot help leaving. The short version: the consumer franchise is a slowly evaporating annuity on somebody else's fibre, priced with unusual honesty; the durable value sits in a second, quieter business — data centre, mail infrastructure and video distribution sold to other operators — that the consumer annuity financed into existence; and the local brand is the hinge that holds the two together.

Thirty years at the same address

Identity first, because the name has changed shape three times while the business never moved. The service called mitene began in March 1996 inside Mitani Shoji, a Fukui trading house founded on cement and coal in 1914 and listed on the Tokyo exchange, which by the mid-1990s already owned a computing subsidiary dating to 1969 and a cable-television company it had founded in 1983. The internet service ran as a division of the trading house for six years. In January 2002 the group incorporated a vehicle called Net Mitani Co. — the date, January 18, 2002, is recorded on the prefecture's official employment site, and the incorporation appears in the parent's own corporate chronology — and that July the ISP business was formally transferred into it, whereupon the vehicle took the name it holds today: ミテネインターネット株式会社, mitene internet Co., Ltd. The domain registry preserves the seam precisely: JPRS whois shows mitene.co.jp registered on May 28, 2002, and the service domain mitene.or.jp carrying a registered date of July 1, 2002 — the day of the business transfer — even though the Wayback Machine holds captures of mitene.or.jp from December 1996, a plain "Mitene Internet Service HomePage" with an info@ contact, six years before the company that now answers for it legally existed.

Ownership is a matter of public filing rather than inference. Mitani Shoji's securities report for the year to March 2025, filed with the Kanto finance bureau in June 2025, lists mitene internet as a consolidated subsidiary with 86.2 per cent of voting rights held by the parent, 77.4 points of that indirectly — which is to say, largely through Fukui Cable Television, the group broadcaster in which Mitani holds 46.4 per cent and de facto control, and its sister operator in Sakai city, held at 65.1 per cent. Capital is a modest 34 million yen. The corporate number, 4210001004801, resolves in the government's gBizINFO register to the head office at 1-3-1 Toyoshima, Fukui city — the same address APNIC's records attach to the company's IP address space — though the register carries no financials, no procurement awards and no certifications, a reminder that small unlisted Japanese companies can be fully real and almost entirely undisclosed at the same time.

Two verification attempts deserve honest reporting. The telecom ministry publishes its register of notified carriers as a downloadable spreadsheet; the current edition would not open cleanly in two different readers when checked for this essay, so mitene's carrier notification number could not be read from the primary list. Its standing as a carrier is nonetheless beyond doubt: NTT West's official directory of hikari-collaboration operators lists the company as a wholesale customer, the industry association JAIPA carries it in the national provider register at the Toyoshima address, and the consumer disclosure statement on its own site is drafted under the Telecommunications Business Act. Second, no standalone profit figure for mitene exists anywhere in the public record: the parent consolidates it into a "living and regional services" segment that also contains cable television, nursing homes, a Toyota dealership, ready-mixed concrete, petrol stations and LP gas — 142.1 billion yen of segment revenue and 3.94 billion of segment operating profit in the year to March 2025. The filing notes, usefully, that segment profit jumped 29.1 per cent partly because cable-television construction costs fell. mitene's own contribution is invisible inside that aggregate, and the analysis below says so wherever it matters.

The name, incidentally, is plain Japanese — mite ne, "take a look" — and it long predates the Tokyo photo-sharing app that now dominates the word in national search results. In Fukui, the collision runs the other way: the ISP had the name first, by two decades.

Renting the wire: the business of selling somebody else's fibre

mitene owns no access network. What it has sold, in every era, is a service layer on infrastructure rented from larger neighbours: NTT's copper in the dial-up and ADSL years (with a parallel ADSL product wholesaled from the Hokuriku power utility's carrier arm), NTT's fibre since 2000, and cable plant belonging to its own affiliates — the company has run internet service over Fukui Cable Television's network since 1999 and over the Tsuruga-area cable operator RCN since 2001. That structural choice, made when the alternative of building plant was at least conceivable, defines every margin the firm earns today.

The current consumer business has two rungs. The older rung is the pure provider contract: a customer keeps a FLET'S line billed by NTT West and pays mitene 1,628 yen a month — a figure confirmed on the company's fee table effective April 2023 — for authentication, addresses, mailboxes and support. On NTT West's discounted family tariff of 4,730 yen the stack totals 6,358 yen; on the undiscounted list price of 5,940 yen it totals 7,568. The newer rung is mitene光, launched in October 2015, eight months after NTT opened its wholesale scheme: mitene buys the whole line wholesale, sets a single retail price, and NTT vanishes from the customer's paperwork. The company's own price page states the family rate as 6,578 yen undiscounted, 4,950 yen with the two-year commitment; the apartment rate as 5,478 and 3,850; and it advertises the arithmetic against FLET'S without embarrassment — 5,940 plus 1,628 equals 7,568, against 4,950, "up to 31,416 yen a year cheaper." A ten-gigabit variant exists with a smaller two-year discount of 1,100 yen a month against the gigabit product's 1,628. Exit terms are mild by Japanese standards: 9,800 yen if a subscriber leaves within 24 months, nothing thereafter.

The national context for that structure comes from the regulator's own monitoring. As of March 2023 — the most recent full survey the ministry has published — Japan had 38.07 million fibre subscriptions, 23.57 million of them on NTT East and West plant, and 16.81 million of those, 71.3 per cent, sold not by NTT but by 838 wholesale-recipient operators under their own brands. mitene is one of the 838. The same report shows who its shelf-mates are: mobile carriers took 73.4 per cent of all wholesale lines, and the NTT group itself — docomo hikari above all — took 47 per cent. The prefectural provider competes in a wholesale channel that was designed for it and then captured by giants using a weapon it cannot forge: the mobile bundle discount that pays customers out of a phone bill the local ISP does not touch.

There is a third rung the invoices barely show. Since 1999 the company has operated internet service for its cable-television siblings — Fukui Cable Television in the capital, the Sakai system to the north, RCN around Tsuruga bay — so a further tranche of prefectural households rides mitene's authentication, addresses and mail without ever contracting with mitene. Within the group's consolidated accounts this is close to an internal transfer; economically it means the firm's true footprint in Fukui broadband is wider than its own brand's subscriber file, and that the group can arbitrate quietly between selling a household cable internet or collaboration fibre without either sale leaking outside the family. A 2001 partnership with BIGLOBE, one of the national ISP majors, shows the same instinct running in the other direction a generation earlier: when a national brand needed a local face in Fukui, the local firm rented its own layer out. Selling service layers on other people's plant, and renting your own layers to other people — the company has never really had a third mode.

One number in this structure is deliberately missing, and its absence is a finding rather than a gap. The rate mitene pays NTT West per wholesale line is not published — not by NTT, not by the ministry. The regulator's own review states that NTT East and West apply uniform wholesale terms to all comers but that the prices are confidential, so fairness "cannot be externally verified." Every hikari-collaboration operator in Japan, from docomo down to a 23-person firm in Fukui, therefore runs its core business on a cost line the public record cannot see. What the record does show is the spread the market will bear: identical fibre retails at 5,940 yen from NTT itself, 5,720 from docomo, and 4,950 from mitene. Someone in that chain is either accepting a thinner margin or betting on lower costs elsewhere, and a 23-person company with no television budget is the obvious candidate for both.

The tariff table prices loyalty

Read as a single document, mitene's price page is a confession of competitive position. The company cannot outspend the mobile carriers, so it prices below them: 770 yen a month under docomo hikari's two-year family rate, 990 under NTT West's undiscounted stack even before the provider fee. It cannot match the bundle, so it targets the households the bundle cannot reach: the price comparison on its own site is drawn against FLET'S plus a provider fee — the configuration of its own legacy base — rather than against docomo, because the realistic prize is converting its existing 1,628-yen provider customers into 4,950-yen whole-line customers before a bundle salesman gets to them. Each such conversion multiplies mitene's revenue per household by three while cutting the household's total bill by 1,408 yen a month against the discounted FLET'S stack. Both sides win; the loser is NTT West's retail arm, which is precisely how the wholesale scheme was meant to work.

The bundle asymmetry deserves to be stated in yen because it defines the ceiling of the franchise. A Fukui family with two docomo smartphones choosing between mitene光 at 4,950 and docomo hikari at 5,720 is not comparing 4,950 with 5,720; it is comparing 4,950 with an effective 3,300 or less, once as much as 2,420 yen of monthly handset discounts enter the ledger. Against that arithmetic the local brand cannot win on price, and does not try. Its economically rational base is the shrinking set of households outside the big-three mobile ecosystems, plus the households for whom the deciding asset is not price but the mitene.or.jp mailbox, the Fukui dialling code on the support line, and a counterparty whose engineers might be met at a school event. Retention engineering is visible in small print: when NTT West's final FLET'S ADSL lines were switched off on January 31, 2026, mitene's notice took care to say that customers' PPPoE identities and mailboxes would survive intact onto any future fibre contract — the addresses, not the copper, being the thing the company actually owns.

How many households answer that description is the question the record refuses to answer directly. mitene publishes no subscriber count; the parent's segment disclosure buries it; the ministry's per-operator wholesale table names only operators above 30,000 lines, and mitene does not appear in it — which is itself a datum, establishing an upper bound. The crowd-measurement site minsoku has accumulated 193 speed reports for mitene光 in total, against tens of thousands for the national brands, averaging a healthy 287 megabits down over the past three months. Single-sourced and self-selected, that figure proves nothing about subscriber numbers on its own; but a collaboration brand that has drawn fewer than two hundred crowd measurements in a decade of operation is consistent with a whole-line base in the four figures or low five figures, concentrated in one prefecture, skewing older than the self-measuring population. The honest statement is: somewhere between a few thousand and a few tens of thousands of paying households, with the provider-fee legacy rung likely still the larger share.

The arithmetic of a business that publishes no numbers

Assemble the unit economics from the primary documents, and flag what is evidence and what is inference. Evidence: a mitene光 family subscriber pays 4,950 yen a month under the two-year discount, 59,400 yen a year, per the company's published tariff. A legacy provider-fee subscriber pays 1,628 yen a month, 19,536 a year, with NTT billing the line separately. An additional mailbox — the classic multi-generation household product — adds 330 yen a month. Evidence, too: the wholesale rate mitene pays NTT West is confidential; the company had 23 employees at the last prefectural filing, average age 41.4, graduate starting pay of 300,000 yen a month; and its parent attributes to the whole seven-business segment an operating margin of 2.8 per cent.

Now the inference, stated as such. If the ISP service layer inside mitene光 is worth what mitene separately charges for it — 1,628 yen — then the line component embedded in the 4,950-yen price is about 3,322 yen, which sits 44 per cent below NTT West's own 5,940-yen list price for the identical line and 30 per cent below its promotional 4,730. That gap is only sustainable if the confidential wholesale rate sits comfortably below 3,322 yen, meaning NTT West sells the line to mitene for materially less than it charges its own retail customers; the collaboration model, in other words, converts NTT's retail margin into the reseller's gross margin, and mitene passes a visible slice of it — 1,408 to 2,618 yen a month, depending on the comparison — back to the household as the price of loyalty. The residual it keeps per whole-line subscriber cannot be computed from public documents, and this essay does not pretend otherwise.

What can be bounded is the wage bill against the consumer book. Take the ISP layer's 1,628 yen as a generous proxy for monthly gross contribution per household — generous because the whole-line product must also cover wholesale invoices, and the provider-fee product must cover authentication, transit and mail costs. Twenty-three employees at a fully loaded cost of, say, six to eight million yen each — an inference from the advertised salary bands, not a filing — imply a payroll near 150 million yen a year. At 19,536 yen of annual ISP-layer contribution per household, payroll alone would consume the entire consumer margin of roughly 7,700 households before a single router, exchange port or data-centre generator is paid for. If the true consumer base is in the four figures, the consumer business cannot be carrying this company. Either the base runs well into five figures — possible, given three decades of accumulation and the cable-affiliate service contracts — or, far more consistent with the observable evidence, the consumer annuity long ago stopped being the main event, and the people are paid by the other business. The network record, next, says exactly that.

Costs on the network side are partially visible in kind if not in yen. The routing table shows mitene originating just under 70,000 IPv4 addresses across five aggregate blocks — 124.241.0.0/17 alone is 32,768 addresses, registered by APNIC to the Toyoshima head office — plus an IPv6 /32. At the prices paid in recent years for transferred IPv4 space, an endowment of that size is an eight-figure-yen asset in its own right, acquired in an era when membership fees rather than markets allocated addresses; a shrinking consumer base, whatever else it does, frees addresses the company can redeploy to paying infrastructure customers rather than buy at auction. The exchange-port bill — eight ports across two countries, five of them at 20 gigabits — is a real fixed cost with no published price, flagged accordingly; it is the tell that traffic, not subscribers, is what this network is now sized for.

The demand curve is retiring

The consumer franchise's terminal mathematics are not a secret; they are on the prefecture's own statistics page, updated monthly. Fukui loses roughly 450 more residents to death than it gains by birth every month, and the 2020-2025 census interval turned in the worst five-year decline on record. Households — the billing unit that matters — are falling more slowly than population, because single-elderly households are rising, and that composition effect is the kindest thing demography is doing for the company: a fibre line is a per-household product, and an 80-year-old living alone keeps paying 4,950 yen just as a family of five would. But household formation runs through the young, the young are leaving — social flows hover near zero against a natural decrease that never pauses — and every February a customer file closes with an estate rather than a cancellation form.

Fibre coverage, meanwhile, offers no frontier to grow into. Japan's household fibre availability reached 99.84 per cent nationally in March 2023, and Fukui's build-out — accelerated for decades by prefectural programmes and the group's own cable plant — is effectively universal. There is no unserved hamlet left whose first connection mitene can win. Growth in Fukui is purely a matter of share, taken from national brands armed with bundle discounts, in a market losing about one per cent of its people every year. The January 2026 copper sunset finished the tidying: FLET'S ADSL's last Fukui lines died on the 31st, closing the technology the company had ridden since 2001 and pushing its final copper holdouts to choose, one last time, between the local name and a national one.

Against that, the company's demographic defences are real but wasting. The mailbox is the deepest: e-mail addresses are Japan's great unpriced switching cost, woven into bank registrations, municipal notices and school contact lists, and mitene's base has had thirty years to weave them. The extra-mailbox option at 330 yen a month quietly monetises the multi-generational version of the same lock. The cable affiliates channel another cohort: subscribers of Fukui Cable Television and the Sakai and Tsuruga systems ride internet service that mitene has operated for the group since 1999, so a slice of the prefecture's broadband runs through the company even when the customer never says its name — the group hosts FCTV's mail domains on mitene infrastructure to this day. And the portal habit persists: the consumer homepage still leads with Fukui shopping, food and event listings, a small daily reason to see the brand that a docomo subscriber never gets. None of these assets grows. All of them decay at the speed of the population pyramid, or a little slower. The rational strategy for such a book is exactly what the tariff shows: price to slow the leak, spend nothing on acquisition wars, and milk the annuity while building something else. Which the company, unusually for its size, verifiably did.

The network that outgrew its prefecture

Here the registries stop describing a prefectural reseller and start describing something stranger. Autonomous system 17961, acquired in January 2002 before the company itself was fully assembled, today advertises its routes at eight internet exchange ports: JPNAP Tokyo and Osaka at 20 gigabits each, JPIX Tokyo at 20, BBIX Tokyo and Osaka at 20, KINX in Seoul at 10, and BBIX Singapore and Hong Kong at a gigabit apiece. Its PeeringDB entry self-declares 100 to 200 gigabits of traffic, mostly inbound, with an open peering policy. Those are operator-entered figures and should be read as claims rather than measurements; but exchange ports are contracts with third parties, and nobody pays for 20-gigabit ports in two cities — let alone ports in three countries — to serve a few thousand pensioners' mailboxes in Fukui. The company's own incident log corroborates physical reach: a May 2026 fault notice reports an optical-level failure on "our trunk fibre segment" in Hyogo prefecture, which places company-operated backbone 200 kilometres beyond the prefectural border, on the route to Osaka's exchanges.

The chronology explains what all that capacity is for, one product at a time. A data centre in Fukui opened in June 2004, with security certifications the same autumn. An Equinix interconnection followed in 2007, the same year as a Tokyo sales office. In 2010 the company established a joint venture with the Tokyo radio broadcaster Bunka Hoso — a prefectural ISP co-founding a company with a national radio station is not a consumer-broadband act — and by 2014 it was selling white-label cloud mail to other operators; its mail-platform page today addresses ISPs, carriers, cable companies and municipalities as customers running their own brands on mitene's clusters. A cloud video-distribution product, MediaStorm, launched in February 2015, eight months before mitene光 did. In September 2018 the company ran a large-scale 4K IP-multicast distribution trial for the national sports festival hosted in Fukui. In 2021 it became an AWS consulting partner and opened what it calls Hokuriku's first interconnection data centre, whose specification page leans hard on geography: seismic-grade-7 tolerance in the lowest earthquake-insurance rating zone in Japan, nine metres of elevation, Hokuriku Electric feed with generator backup, closed-circuit on-ramps to AWS and Azure, and a connection to the academic network SINET. The pitch is Fukui-as-refuge — disaster-recovery floor space for Tokyo and Osaka firms that need their backups outside the Nankai Trough's reach — and the named case studies on the corporate site, Fukui Bank and the prefectural industrial support centre, show the local enterprise flank of the same business.

Follow the money through that sequence and the consumer story inverts. The dial-up and ADSL cohorts of the late 1990s funded an autonomous system, address blocks now worth eight figures, a data centre and a mail plant. As the consumer base aged, the plant was progressively resold — to other ISPs facing the same mail-economics squeeze without the scale to fight it, to broadcasters needing distribution, to municipalities needing compliant hosting, to enterprises needing a disaster site that is cheap precisely because Fukui is emptying. The Seoul, Hong Kong and Singapore ports and the mostly-inbound traffic profile are consistent with content distribution and cloud interconnection, not with eyeball access in Fukui. The consumer brand meanwhile pays a different kind of dividend: it is the reason a Fukui bank, city hall or school board treats a 23-person company as an institution rather than a vendor. In procurement terms, thirty years of mitene.or.jp addresses on the prefecture's correspondence is the moat around the data centre. The two businesses are not a diversified pair; they are one asset — trust plus plant — sold at retail to households and at wholesale to everyone else.

The risks of the second business are the mirror of its logic. Mail is a scale game refereed by three hyperscalers, and the incident log shows the referee's whistle: for part of May 18, 2026, mail from the company's sending clusters bounced at Microsoft 365 until deliverability was restored — a routine skirmish, transparently reported, but a reminder that a regional operator's flagship wholesale product survives at the pleasure of reputation algorithms tuned in Redmond and Mountain View. The data centre's disaster-recovery thesis competes with every hyperscaler region and with rival regional DCs making identical low-seismicity claims. And the whole edifice rests on 23 people — the prefectural filing counts fourteen men and nine women — whose average age matches the company's: a key-person concentration that would alarm any acquirer's diligence questions, and that the parent's deep group bench only partially offsets.

Signals from the edge of the record

What the official record leaves dark, the unofficial record lights unevenly, and the flickers are worth reading. The hiring picture first: the national job boards carry 17 open listings against the company's name this summer — a large number for a 23-person firm — and the advertised bands run from five to six million yen for infrastructure sales engineers up to seven to twelve million for senior technical leads, Tokyo-competitive money aimed explicitly at U-turn and I-turn returnees. A company harvesting a dying consumer book does not advertise twelve-million-yen engineering roles; a company scaling a wholesale infrastructure business does. The signal suggests the second business is where headcount pressure sits; sustained postings over the next year, or a visible headcount rise in the prefecture's next filing, would settle it.

Service-quality chatter is thin, and the thinness is itself the signal. The crowd-measurement record — 193 reports ever, respectable median speeds, latency ordinary for PPPoE-era plumbing — sits where a small, old, satisfied base would put it: too few complaints to trend, too few users to benchmark. The company's incident log, kept publicly since the 2000s with categories still listing PIAFS and other Heisei-era artefacts, reports faults in minutes-to-hours granularity with named causes, including the unflattering ones — a roaming supplier's failure that cut vgate.ne.jp logins for 25 minutes in May 2026, the Hyogo trunk fault, the Microsoft bounce. Radical operational transparency from a firm that publishes no financial numbers at all is a cultural datum: the engineering side of the house behaves like a common carrier even where the corporate side discloses like a family business. What would move this from signal to evidence is simple and unlikely: a subscriber count, a churn figure, or a data-centre utilisation number in any public document.

Employee-voice platforms hold a listing for the company but keep the reviews behind registration walls, so workplace sentiment cannot be read directly; the observable proxies point the same mild way as the hiring bands. The prefectural filing reports a 74.2 per cent paid-leave take-up and three planned graduate hires against one the year before — small absolute numbers that nonetheless represent a 13 per cent headcount expansion if delivered, in a prefecture where the working-age population shrinks every year. The competing interpretation, that the postings mask replacement churn in a stretched team, cannot be excluded from outside; the difference would be visible within eighteen months in whether the incident log's response quality holds and whether the same roles reappear on the boards.

The quietest signal is namespace hygiene. The company still holds mitene.jp (registered 2001), mitene.or.jp, mitene.ad.jp — the address-management credential of a first-generation Japanese network operator — mitene.co.jp and mitene.biz, and keeps legacy notice categories alive for services that died a decade ago. Institutions that intend to wind down do not maintain their 1996 furniture this carefully. Whoever is steering expects the name to be needed for a long time yet.

What would move this judgement

The reading offered here — evaporating consumer annuity, durable wholesale flank, brand as the bridge — rests on documents that could be contradicted by a handful of facts, and it is worth naming them. A published subscriber number from the company or a ministry disclosure placing mitene above the 30,000-line reporting threshold would overturn the assumption of a small base and make the consumer business the story after all. Disclosure of NTT East-West wholesale rates, which the regulator has so far declined to force, would replace this essay's central inference — a wholesale price comfortably below 3,322 yen — with arithmetic, and would reprice every collaboration operator in Japan in the same afternoon. A tariff event would matter in either direction: mitene abandoning the 4,950-yen family price upward would signal the loyalty premium has finally exceeded what a shrinking base will bear; a national brand launching a Fukui-targeted discount below it would signal the bundle carriers have decided prefectural books are worth strip-mining. On the corporate register, any change in Mitani Shoji's 86.2 per cent control — a buy-in to full ownership, a merger with the cable affiliates, or a sale to one of the consolidators rolling up regional Japanese ISPs — would mark the moment the group repriced the asset itself. The 2030 census will quantify the demand curve's slope with brutal precision. And the failure mode to watch is not commercial but human: a company of 23 whose engineering reputation is the product has a bus factor measured in single digits, and a bad retirement year would show up in the incident log's response times before it showed up anywhere else.

Absent those facts, the judgement stands as follows. mitene internet is a rare clean specimen of what Japan's wholesale fibre regime actually produces at the periphery: a firm that neither owns the wire nor sets the wholesale price, surviving on a 1,628-yen service layer, a 770-yen loyalty discount against the national default, and a prefecture's worth of accumulated trust — and that spent its demographic head start converting subscriber cash into infrastructure the subscribers will never see, in Seoul, Singapore and a seismically boring corner of Hokuriku. The loyalty of an aging prefecture turns out to be worth about 4,950 yen a month, for as long as the people paying it last. What the company built with thirty years of it may outlast them.

Evidence register