Build the measure before the machine.
Habitability is being developed as a territorial intelligence platform: measure how habitable a place is, understand how it changes through time, and eventually anticipate future trajectories.
Select the historical CORE.
The Grenoble historical audit has passed. The existing Eurostat layer is sufficiently rich to support a serious 10-year backtest attempt. The current task is now methodological: select a compact, defensible set of indicators before defining dimensions, normalization or any headline score.
First validation milestone
Indicators with Grenoble data over 2014–2024
Indicators with at least 70% historical coverage
Indicators with at least 90% historical coverage
Continuous series within their observed historical window
Decision: do not launch a new source-hunting campaign. Existing Eurostat history is sufficient to proceed to CORE indicator selection and test whether a defensible trajectory can be constructed.
What has already been built
What is not yet validated
Historical CORE
Coverage is sufficient, but the relevant non-redundant indicator set has not yet been selected.
Core methodology
Dimensions, directionality, normalization and aggregation rules are not yet frozen.
Headline index
No public Habitability score should be treated as validated until the historical backtest succeeds.
Cross-city robustness
The methodology has not yet been tested unchanged across several territories.
Forecasting
2030–2050 scenarios remain downstream of a trusted historical measurement model.
Automation at scale
Agents can maintain and explain the system later; they are not the current bottleneck.
No new agent development until the Grenoble backtest works.
New automation should only be added when it removes a demonstrated bottleneck in the validated methodology or data-maintenance workflow. Individual missing data points should not pull the project away from the core validation path.
Validation trajectory
Historical coverage audit
Validated for Grenoble over 2014–2024: 194 indicators with data, 101 with at least 70% coverage, 13 with at least 90% coverage, and 183 continuous series.
Core indicators
Reduce the broad indicator universe to a defensible core based on relevance, historical coverage, stability and geographic comparability.
Dimensions
Lock the conceptual structure and assign core indicators to a small set of stable Habitability dimensions.
Normalization
Define transparent normalization, directionality, missing-data and aggregation rules before producing any headline score.
Grenoble 10-year backtest
Reconstruct Grenoble through time using the same methodology each year and test whether the trajectory is interpretable and robust.
Methodology validation
Test face validity, sensitivity to weights, missing-data robustness and geographic consistency.
Cross-city validation
Apply the exact same framework to several additional territories before claiming general validity.
Website integration
Turn validated trajectories, drivers and provenance into the core public experience on habitability.eu.
Agents & automation
Use scouts, validators and explainers to maintain the system once the methodology is stable—not to define it.
Forecasting
Only after the historical model is trusted, extend Habitability from observed trajectories toward future scenarios.
The product thesis
Habitability should not merely rank places. Its distinctive value is to show whether a territory is becoming more or less habitable, which forces drive that change, how trustworthy the evidence is, and—later—how the trajectory could evolve.
Sequence: Data → 10-year backtest → Methodology → Cross-city validation → Public product → Agents → Forecasting.