Two Weeks Harder Than Expected After Going Nationwide — Where 'Just Multiply the Count' Stops Working

Hello from the MAPRISE team. In early-to-mid August, we pushed hard to take our data nationwide from Kyushu's 7 prefectures, and reported on it here — first on taking registry cadastral maps and PLATEAU nationwide, then on eight more layers going nationwide. This is the continuation: a record of what we actually spent the following two weeks doing.

After the fanfare of "nationwide rollout complete," the following two weeks were a quiet "polishing" phase. Much of the know-how we built up during the Kyushu-only era carried over to the whole country through simple multiplication of record counts. But a fair amount didn't work that way, and where that line falls is the subject of this post.

What moved over these two weeks

What scaled by simple multiplication

Appraisal report data grew roughly 11x, from 23,360 records across 8 prefectures to 258,214 records across all 47 prefectures nationwide. Even so, the API structure of MLIT's Real Estate Information Library doesn't change at prefectural borders, so we loaded the nationwide data without touching the ingestion logic at all. Large-scale fill land (A54), land-alteration regulation areas (A56), National Forest data, and FIT-certified solar facilities fall into the same category — each simply reused the "dry-run → count verification → production load → per-prefecture count diff" procedure we described in our previous post on the eight-layer rollout. The order of magnitude changed; the mechanism didn't. This is the ideal form of scaling by multiplication.

What needed individual checks

Most of this round's work, however, didn't fit that ideal pattern. The reasons break down into three categories.

1. The thing to be counted doesn't exist nationwide in the first place

Location-optimization plan areas are covered in only 45 of 47 prefectures, not all 47. The reason is simple: for Tottori and Okinawa, the source data itself — MLIT's National Land Numerical Information (A55) — hasn't been published yet. On top of that, since location-optimization plans are drawn up at each municipality's own discretion, area data simply doesn't exist for municipalities that haven't drafted a plan. The four prefectures excluded from land-alteration regulation areas (A56) — Aomori, Kagawa, Saga, Okinawa — hit the same kind of wall: either no zone has been designated there, or the data hasn't been published.

Historical topographic maps (1:50,000 scale, from the pre-war Showa era) ran into the same wall. We use a public-domain collection published by Stanford University Libraries, but the state of digitization varies wildly by region, and as of August 2026 some major cities — including Sendai, Kyoto, and Osaka — still aren't covered. We couldn't simply extrapolate "how many cities are left" from the national mesh count; we had to go through the collection sheet by sheet to determine coverage.

Telling "not yet covered" apart from "the target simply doesn't exist" takes checking case by case, and that's something no amount of multiplying record counts can substitute for. We learned how dangerous it is to get this wrong the hard way, in mid-August. We were about to write up a welfare-facility layer on the assumption it was "still Kyushu-only" — but a fresh measurement showed it had already expanded to 42 prefectures, and we had to rewrite the description at the last minute. Right after that, a separate audit found a licensing problem in part of the underlying data, and we switched to temporarily withholding the record count altogether. It was a hard lesson in the danger of trusting an old description instead of re-measuring.

2. Even within the same dataset, how far a layer can scale differs by layer

PLATEAU's 3D city model — the building geometry and attributes themselves, like position, height, and usage — has scaled to 265 municipalities nationwide. But the per-building seismic collapse probability, layered on top of that, requires a separate and heavy computation (deriving seismic-intensity meshes), and remains fixed at 27 Kyushu municipalities and roughly 1.96 million buildings.

We actually treated these two as the same scope once, and mistakenly wrote in the August 16 data-update entry that "the collapse-probability layer covers the same area." We caught it five days later, on August 21, and corrected it. It was a hands-on lesson that within a single data source like PLATEAU, "geometry data" and "derived secondary data computed from it" scale at completely different levels of difficulty. Rather than erasing the mistake and pretending it never happened, we chose to keep the correction, and the story behind it, in the update history.

Something similar happened with land-price data. Official land prices had expanded to all 47 prefectures nationwide, while the prefectural land-price survey remained limited to 2025 data for Kyushu's 7 prefectures only — a 2026 edition doesn't even exist yet. Bundling both under one "land price" line would have been misleading given how differently the two have scaled, so we split them into two separate entries.

3. Inconsistent wording — unglamorous work that never quite ends

The source attribution for city-planning decision GIS data had drifted into four slightly different wordings scattered across the underlying JSON file ("MLIT City Planning GIS Data," "Municipal city-planning decision data," and so on). Wording drift that would have been a one-line fix back in the Kyushu-only days now shows up scattered across many files and many sections as the scope of what we reference has grown. Some parts of the LP are also cached at deploy time, so fixing the surface alone would just get rolled back on the next deploy — we had to trace the fix all the way back to the source data file.

We also re-checked the displayed record count for farmland parcel (fude) polygons against the actual data and found it was wrong. The underlying coverage hadn't changed; we corrected the displayed count to the actual figure — approximately 26.47 million records across 41 prefectures. Every time we write down a number now, we've learned to be suspicious of it until we've re-verified it against the live data — a habit that matters even more after going nationwide.

When we get stuck, we ask the government

Back when we only had to think about Kyushu's 7 prefectures, most questions could be settled by reading published materials or checking with a municipal desk. Dealing with the whole country multiplies both the number of agencies to check with — the Geospatial Information Authority of Japan, MLIT's head office, the Ministry of Agriculture, Forestry and Fisheries, the Ministry of Justice, each prefectural government — and the layers of bureaucracy involved. This round, we emailed the relevant desks to ask whether A55 data for Tottori and Okinawa (location-optimization plan areas) has a publication timeline, and whether additional coverage is planned for areas still missing from the historical topographic maps, like Sendai, Kyoto, and Osaka.

Honestly, the answers we get back are usually less specific than we'd like — "no timeline at this time," "under progressive preparation," that sort of thing. Even so, that's far safer than guessing "it'll probably be published soon" and getting ahead of ourselves in the LP copy. Sometimes all an inquiry confirms is that we still don't know — but we think that's still worth recording. Asking first, so that we can honestly write "we don't know" instead of guessing, is unglamorous, but it's part of the sweat-equity work too.

The wider it gets, the more checking multiplies

Honestly, the further nationwide coverage progresses, the longer each individual fix clearly takes. Back in the Kyushu-only days, one bug meant fixing one spot in one file. Now the same wording is scattered across multiple surfaces — the per-prefecture data-source list, the update history, the announcements feed, industry-specific solution pages, two languages — and fixing it in one place while an old description lingers somewhere else has become a common occurrence. Regression checking has gone from "Kyushu's 7 prefectures times the number of layers" to "47 prefectures times the number of layers times the number of surfaces it appears on," and the time spent just checking keeps growing.

We won't deny that things feel slower as a team. But we see that as the natural consequence of the scope having grown, and as evidence that we're trying to expand without sacrificing quality. Whether you look at the sheer volume — record counts, prefectures covered — or at the quality and governance side — unified source attribution, a codified data sourcing policy — MAPRISE is unmistakably moving forward.

Who this actually helps, and with what

Rather than end on technical process alone, here's where we think the data we took nationwide or newly added over these two weeks could actually help.

Real-estate appraisers and brokers: With appraisal report data now at 258,214 records across all 47 prefectures, the "two independent appraisers' valuations for the same standard point" and "comparison / income / cost valuation methods" that used to be usable only for properties in a handful of prefectures can now be referenced for any property nationwide — useful as supporting evidence for valuation objectivity via comparable cases in other areas.

Loan underwriting and collateral-valuation staff at financial institutions: Now that registry cadastral maps and parcel polygons are nationwide, you can screen a remote collateral property's parcel boundaries and shape at your desk before visiting the site. Combined with nationwide appraisal reports, this could help streamline the first-pass check in collateral valuation.

Land and house surveyors: Overlaying registry cadastral maps with the newly added historical topographic maps (1:50,000, pre-war Showa era) lets you research a site's formation history — former riverbeds, reclaimed land, and so on — from your desk anywhere in the country before fieldwork. Using historical topographic maps as a clue to liquefaction risk is already common practice, and adding it as a nationwide layer is meaningfully useful.

Urban-planning consultants and municipal planning staff: With location-optimization plan areas now viewable across 45 prefectures, you can compare your own area's plan side by side with other municipalities' residential and urban-function inducement zones — useful for benchmarking compact-city policy.

Property and casualty insurance underwriters: Combining historical topographic maps with existing flood-inundation and borehole-log layers could serve as one input among others for earthquake and flood insurance underwriting (not a substitute for formal ground surveys or official hazard maps).

Renewable energy and forestry-related businesses: Overlaying nationwide FIT-certified solar facility data with nationwide National Forest data (sub-compartments and forest roads) could support an initial screening for idle-land or forest-land utilization potential.

Real-estate developers: Whether a prospective development site falls within a location-optimization plan's residential or urban-function inducement zone is a key factor in floor-area-ratio relaxations and subsidy eligibility. Nationwide coverage now makes that same screening possible for sites outside the areas previously covered.

Lawyers and judicial scriveners: For inheritance cases and boundary disputes where land rights need checking, nationwide registry cadastral map coverage could help streamline preliminary research regardless of where the case is located.

In every case, data on MAPRISE carries no certifying authority from any public agency and doesn't replace formal appraisal, survey, or registration procedures. It's positioned strictly as reference information to make desk-based preliminary research and screening more efficient.

Closing

Honestly, the day we can say "nationwide coverage, complete" with a straight face is still some way off. As this post shows, we sometimes correct things after publishing them, and we run into walls where the source data simply doesn't exist yet. Even so, we don't think there's any way to expand without sacrificing quality other than patiently telling apart, one at a time, what can scale by simple multiplication from what needs an individual check. Over these two weeks, our pace has genuinely slowed — but in exchange, the quality and quantity of our data, and our ability to explain it honestly, have unmistakably moved forward.

Questions or feedback are always welcome at info@maprise.jp.

The data introduced in this post is sourced from open data published by MLIT, the Ministry of Justice, the Geospatial Information Authority of Japan, the Ministry of Agriculture, Forestry and Fisheries, the Forestry Agency, METI, Stanford University Libraries, the OpenStreetMap Foundation, and others. Record counts and prefecture coverage reflect the time of this post's publication and may change as our verification work continues and as source agencies publish further data. Descriptions of potential industry use cases represent general illustrative scenarios and do not guarantee applicability to any specific business use.