Methodology: How the data on maunawai.info is produced

Draft for the public methodology page · As of 8 July 2026

maunawai.info shows you the quality of your supplier's drinking water: with real analytical values, limit-value classification and a link to the original document. On this page we explain in full where this data comes from, how we collect it and where its limits lie.

Where the data comes from

Every water supplier in Germany is legally obliged to publish current analytical results for its drinking water on the internet (§ 46 Trinkwasserverordnung 2023). These mandatory publications — usually PDF analysis reports or tables on the supplier's website — are our primary data source. In addition, there are open datasets from the federal states, such as the supply area geometries of LANUV NRW (Datenlizenz Deutschland Zero) and, where licensed, state information systems such as TWIST in Rheinland-Pfalz.

The problem we solve: this data is scattered across thousands of websites, in a wide variety of formats, without any uniform structure. We bring it together and make it searchable, by postcode and by supply zone.

Every value on maunawai.info carries a link to the supplier's original document and the date on which we retrieved it. You can verify every value yourself.

How we collect the data

We work with a three-stage pipeline and state openly that software and AI are involved:

1. Automated collection. A crawler retrieves the suppliers' publications. It complies with the technical rules of the websites (robots.txt, low request frequency) and identifies itself with a name and contact address. Every document is archived with its source, retrieval date and checksum.
2. AI-supported extraction. The measured values from the analysis reports are transferred into a uniform structure by an AI language model. The AI reads off, it does not interpret: values are taken over verbatim, including entries such as "< 0.001" or "not detectable". Unit conversion and parameter assignment are then handled by deterministic program code, not by the AI.
3. Human review. No automated procedure is error-free, ours included. That is why trained staff check the extraction: new document layouts are initially checked in full against the original, while conspicuous or uncertain values are placed in a review queue and are only published after approval.

Quality assurance

• Double extraction: Documents are extracted twice independently; discrepancies between the two runs force a human review.
• Plausibility rules: Every value passes through automatic checks (physically possible ranges, unit consistency, comparison with limit values). Anomalies are not published but reviewed.
• Measured accuracy: Before launch we assessed the procedure against manually verified analysis reports. In the validation with 26 documents from 15 suppliers, 98.6 % of the more than 1,100 spot-checked fields were correct; the error classes found have been incorporated into the pipeline rules. We repeat this measurement on an ongoing basis with a growing test dataset.
• Confidence per value: Internally, every measured value carries a review status (automatically validated, human-reviewed). Values that reach neither of these states do not appear on the site.
• Traceability: Measured values are never overwritten. Corrections create a new version with a reference to the old one; the original document remains archived.

Timeliness

Most suppliers publish their analyses annually, typically in the first quarter; some quarterly or more often. We check the sources for new editions accordingly at regular intervals. For every value you can see the examination period or edition status of the underlying report and our retrieval date. We take the edition status from the document content itself, not from the file name or web address, because in practice both are often out of date.

We show limit values with versioning: where limit values change (for example PFAS-20 since 12 January 2026, PFAS-4 from 12 January 2028, lead from 2028), we state the value applicable in each case together with its date of validity and legal basis.

Limits of this data

Honesty is part of the methodology. You should be aware of these points:

• Not official. maunawai.info is a private information service. Binding monitoring of drinking water is the responsibility of the public health authorities (Gesundheitsämter); binding information is provided by your supplier and your Gesundheitsamt.
• Values apply from the waterworks or supply zone. What happens in the building's plumbing (for example old lead pipes in buildings predating 1973) is not covered by these analyses. For questions about plumbing, an examination by an accredited laboratory will help.
• We show what has been published. If a supplier does not publish a parameter, we cannot display it; we then mark it as "no published values found" with the retrieval date.
• Errors are possible. Despite multi-stage checking, transcription errors can occur. In case of doubt, the linked original document applies.
• Correction channel: Have you found an error, or are you a supplier whose data is shown incompletely? Write to kontakt@maunawai.info. We review every report and correct it promptly.

Licences and source citations

• The postcode ↔ supply zone mapping table is based in part on OpenStreetMap data (© OpenStreetMap contributors) and is openly available under the Open Database License (ODbL), including as a download via our API. For re-use: attribute "maunawai.info" and observe the ODbL conditions.
• Zone geometries for individual federal states come from open state data (e.g. LANUV NRW, dl-de/zero) and are labelled with their respective licence for each dataset.
• The CAP ↔ Comune mapping for Italy comes from GeoNames (CC BY 4.0) and is computed against the municipal areas from ISTAT. It serves search purposes only: the values themselves are tied to the Comune named by the Gestore in its analysis document — not to the postcode.
• The analytical values are facts from the suppliers' mandatory publications; the respective source is linked at every value. Our API terms apply to use of our processed database via the API.
• Citation welcome: If you use data from maunawai.info, we appreciate the source citation "maunawai.info" with a link and a reference to the supplier's original document.

Who is behind the portal

maunawai.info is a service of MAUNAWAI Schweiz GmbH. MAUNAWAI manufactures water filters. So that this interest does not distort the presentation, firm rules apply here: no product advertising in the data section, no purchase recommendations derived from measured values, and the honest overall assessment stated up front: German tap water is among the most strictly monitored foodstuffs, and well over 99 % of all samples comply with the limit values. This portal exists because the published data is hard to find, not because the water is poor.

Questions about the methodology: kontakt@maunawai.info