From 7 prefectures in Kyushu to nationwide - a day spent checking data one sheet at a time

Hello from the MAPRISE development team. This post is a little different from usual — instead of introducing a specific new feature, we're writing up a record of one day's work, as it happened.

From Kyushu's 7 prefectures to nationwide

MAPRISE started out covering Kyushu's 7 prefectures — Fukuoka, Saga, Nagasaki, Kumamoto, Oita, Miyazaki, and Kagoshima. Narrowing the scope let us build up data accuracy and usability first. Since then, we've heard from real estate, financial, and homebuilder users looking beyond Kyushu, and we've been working to expand our coverage area alongside the range of data we offer. Today was one of the busiest days yet for that expansion.

What we expanded, one layer at a time

Over the course of today, the following data layers were either expanded from Kyushu's 7 prefectures to all 47 prefectures, or added as brand-new layers:

• Elderly care service providers (Ministry of Health, Labour and Welfare; new layer, 223,107 records)
• Disability welfare service providers (WAM NET; new layer, 209,452 records)
• Census small-area statistics (MIC Statistics Bureau; 231,931 records)
• Economic census mesh (MIC Statistics Bureau / METI; 325,522 records)
• Foreign resident counts, nationality breakdown, and ratio heatmap (MIC; 718,457 records combined)
• Crime occurrence data (prefectural police; 292,391 records, including the three prefectures held back until today by data-format differences)
• Prefectural Land Price Survey, L02 (MLIT; 98,684 records spanning 2023-2025)
• Welfare facilities (MLIT; 42 prefectures, 93,232 records — five prefectures, Iwate, Yamagata, Gifu, Wakayama, and Ehime, are not yet covered)

We also connected elderly care and disability welfare service providers to our AI chat. They were already on the map, but you couldn't ask about them in chat — now you can, and the chat answers using the same underlying data as the map.

None of this is a flashy new feature. Expanding each layer nationwide meant re-reading the source statistical tables, cross-checking the record counts prefecture by prefecture, and confirming nothing in the existing Kyushu data had shifted — steady, unglamorous work, repeated layer after layer.

What careful checking turned up

When you check this carefully, you inevitably find things — some good, some not so good. Here are two we found today.

The first was in an economic statistics dataset. While preparing the Kyushu data for comparison ahead of the nationwide rollout, we found that records that should have shown the most recent year's values had actually been carrying figures from nearly a decade earlier — the record counts looked completely normal, so this is exactly the kind of discrepancy you'd never catch from a glance. We only found it because we were chasing down a small mismatch in the number of records processed. We corrected it to the right year's values.

The second was in the layer that color-codes foreign population on the map. We found that one prefecture's population had been displayed at roughly 30% below its actual figure. The cause was that the source statistics contain two different counts — one for the whole municipality, and one for just the densely-inhabited part of it — and our data had been overwritten with the smaller of the two. We found this during the verification work for the nationwide rollout, and corrected it.

Honestly, finding issues like these is never a great feeling. But catching them and fixing them, the way we did today, is a genuinely good outcome. The alternative — publishing without ever noticing — is far worse.

Why we re-check the old data every time

Every time we add a new layer, we also check that the existing Kyushu data — record counts, contents — hasn't shifted. It's tedious, honestly a little boring, work. But skipping it is exactly how issues like the two above would slip through unnoticed. It's not glamorous, but it's the most reliable approach we know of, so we keep doing it.

Where things stand

As of today, MAPRISE now covers all 47 prefectures for land price publication data, elderly care and disability welfare service providers, census small-area statistics, small-area boundaries, economic census mesh, the three foreign-population datasets, crime occurrence data, and the prefectural land price survey. Some datasets, like welfare facilities, still have prefectures we haven't been able to cover yet, and several others remain Kyushu-only for now.

What's next — aiming for nationwide by end of August, honestly still a work in progress

We're aiming to finish nationwide coverage for our major data layers within August. That doesn't mean every feature will work identically everywhere, though. Seismic collapse-probability simulation and features like the PLATEAU 3D building models, which depend on how far each municipality has built out its own underlying data, are out of scope for this push — the source data itself varies too much municipality by municipality for the same verification process we've built for Kyushu to simply carry over.

To be honest, everything described in this post, including today's work, is still a work in progress. New irregularities keep turning up, and the more we check, the more we find that needs fixing. Even so, we intend to keep doing this the same way — checking one sheet, one dataset, at a time.

Closing

Today we wrote up a record of the work itself, rather than introducing a feature. It's not flashy, but we believe this kind of steady accumulation is what gets MAPRISE to a place that's more accurate and covers more ground. We'll keep at it, one careful check at a time.

The data introduced in this post is sourced from open data published by Japan's Ministry of Health, Labour and Welfare; the Welfare and Medical Service Agency (WAM NET); the MIC Statistics Bureau; the Ministry of Economy, Trade and Industry; the Ministry of Land, Infrastructure, Transport and Tourism; and prefectural police departments. Record counts and prefecture coverage reflect the time of publication and may change as verification work continues.