Projects / Power BI dashboard

Power BI dashboard of French historic monuments

The 46,760 protected monuments of the Mérimée database, looked at this time the way a data analyst would: cleaning, measures, interactive visualisation.

Period: October 2026
Type: Personal project
Data: Mérimée database, 46,760 records
Tool: Power BI Desktop

Overview page of the dashboard (in French): breakdown by protection type, ten most frequent building types, map and region filter — Overview filtered on Île-de-France: 3,988 monuments, 1,221 of them listed at the highest level, “classé” (30.6%).

Starting point

The Mérimée database records the buildings protected as historic monuments in France; the Ministry of Culture publishes it as open data. I usually work on this corpus from the NLP side, for information retrieval and RAG. Here the goal was different: learn Power BI and ask an analyst's questions of the same records — how many, where, from which period, protected when?

Data preparation

CSV 78 columns→Power Query cleaning→Model 14 columns→DAX measures→Report 2 pages

Protection: the source field mixes dates and statuses (“1980/01/25 : classé MH ; 1980/01/25 : inscrit MH”). It is split into a type (classé, inscrit, not specified) and a year.
Century: an M formula reduces “4e quart 15e siècle ; 1er quart 16e siècle ; 17e siècle” to the century of the first building campaign.
Coordinates: the GPS field is split into latitude and longitude, typed with the locale that matches the decimal separator.
Quality: missing values labelled “Non renseigné” (not specified), double regions reduced to the first one, and three mistyped dates excluded — they were enough to flatten the whole timeline chart.

Measures and visuals

Three DAX measures are recomputed under the active filters: the number of monuments (COUNTROWS), the number of “classé” monuments (CALCULATE) and their share (DIVIDE).

The report has two pages. The first gives the overview: breakdown by protection type, ten most frequent building types, map of the monuments and a region filter that updates every visual. The second is about time: protections per year since 1840 and the most represented centuries of construction.

Timeline page of the dashboard (in French): monuments protected per year and ten most represented centuries of construction — Protections per year and centuries of construction, whole of France.

What the data show

32.1%of monuments are “classé”, at least in part

2,038protections in 1926, the peak of the series

15.1%of records with no century of construction

Churches dominate: 10,118 records, ahead of houses (5,802) and châteaux (5,704).
534 records carry a protection dated 1840, the year of the first list of historic monuments.
Among known centuries, the 16th comes first (5,328 records), ahead of the 12th and the 18th.

Limits

Mérimée is mostly a textual and categorical database, so the measures remain counts and proportions. A monument that is both “classé” and “inscrit” is counted as “classé”, and a monument spanning two regions is attached to the first one. This is a first dashboard, built to learn the tool; above all it reminded me that half of the work lies in data preparation.

Power BI Desktop · Power Query (M) · DAX

← Semantic search