Method
From a territorial question to a decision.
Five steps repeated in every case. We don’t produce maps to confirm a conclusion reached in advance: we frame questions, test them, verify the results and state what the analysis cannot tell us.
Frame the question
Everything starts with the decision someone has to make, not with the technique. The question sets the scale, the data and what a useful result looks like.
Gather evidence
Public data (census, business directory, crash records, street network, zoning)and the organization’s own records, brought to a common spatial unit, with their dates and their limits.
ExampleFootprints of 261,000 buildings set against the zoning of 164,504 parcelsZoning versus the built cityModel
Pedestrian networks, isochrones, spatial statistics, typologies and simulations with explicit rules. No opaque composite indices: every threshold is in plain view and open to discussion.
Test
The result is compared with other sources, with alternative scenarios and with what chance alone would produce. Each case states what it cannot prove.
ExampleZoning intensity matches proximity for 18% of the population; by chance, 20%Densify where the city already is?Decide
The analysis ends in a feasible decision: which land to occupy first, which crossing to improve, where to locate a service. The atlas and the technical note make it verifiable.
Principle
Data can also contradict the hypothesis.
When the data contradict the hypothesis
Every atlas begins with a question and, almost always, with a suspicion. What separates an analysis from an argument is what you do when the data don’t confirm it. In this series, the rule is to say so on the map itself.
Three cases
- Is Airbnb emptying the city? Listings don’t follow the emptying: they follow vacation housing. Where 40% or more of homes were for occasional use, there are 4.58 listings per 100 homes, eleven times more than where there were none.
- The walled city. We expected gated developments to lengthen the walk for people living outside them. Removing gates and private streets from the network, only 0.6% of the population outside would walk any farther. The effect falls on those who live inside, whose elementary school is 1.4 km away, against 572 m outside.
- Twelve years of the business directory. INEGI’s business directory does not register the pandemic: 2020 appears as its best year and 2021 dips just 1.4%. The directory is updated in waves, so it is useful for seeing structure, not short-term swings.
Why it matters to decision-makers
A policy built on a false hypothesis spends in the wrong place. Regulating vacation rentals where the city is emptying, opening gates to improve mobility for people living outside, or reading the crisis in the directory would all have been poorly aimed decisions. Knowing what is not happening also provides direction.
New data
When new data arrive, we don’t mix scales.
Note · 2026-09-27
2025 Intercensal Survey: what can be updated in the atlases, and what cannot
On September 22, 2026, INEGI published the results of the 2025 Intercensal Survey (EIC). It was carried out as a survey: some 7.3 million sampled homes were visited between October 6 and November 14, 2025, with a reference date of October 15.
The difference matters. A count covers every home and makes it possible to publish data by block. A survey estimates: its results reach down to the municipality and to localities of 50,000 residents or more. Cancún crosses that threshold, but its blocks do not.
What doesn’t change
Almost every atlas works by block or by 250-meter cell with the 2020 Census, because the question is where: which blocks are losing population, who lives behind a gate, how many children cross an avenue. That fine-grained reading cannot be redone with the 2025 EIC. Until the 2030 Census, the 2020 Census remains the block-level baseline.
- The emptying city and The value frontier compare blocks between 2010 and 2020; the EIC does not allow that comparison to be extended to 2025.
- The walled city, Crossing to get there and The care workday assign block-level population to the street network; they will keep using 2020.
- Housing without permanent residents comes from the Census; the EIC captures occupied private dwellings, so it cannot be used to measure vacant housing.
What can be revised
- The total population of the municipality and of the locality of Cancún in 2025: it shows whether the established city remained stagnant while the periphery grew, though without saying in which blocks.
- Everyday mobility, one of the survey’s topics: municipal-level travel times and modes can be used to test the accessibility model in One hour by bus.
- Household vehicles: nationally, the share of homes with a motorcycle rose from 12.2% to 19.9% between 2020 and 2025.
- Labor force participation: Quintana Roo recorded 61.6%, the second highest in Mexico; it is context for the employment and care atlases.
How it is used
The rule is not to mix scales. Where an atlas needs blocks, it stays with 2020 and says so. Where the municipality or locality is enough, the 2025 EIC comes in as an update or as a check, with its margin of error in plain view. Revisions are added to each case as they are made.
Source: INEGI. Results Report 37/26, 2025 Intercensal Survey (EIC) (September 22, 2026)
Cases affected: The emptying city · One hour by bus · The hotel boulevard · The value frontier
Tools
Tools stay in the background: they are chosen to fit the question.
- Python
- MapLibre GL
- D3
- OpenStreetMap
- Network analysis (Dijkstra)
- Spatial statistics (Getis-Ord Gi*)
- k-means