Skip to content

Decision-ready data.

Atmart Data

The platform of reconstructed data on Haiti. We do not republish public files: we build datasets that do not exist yet, by linking scattered sources to shared identifiers, to a history and to documentation that hold up in front of a committee. During its first year, Atmart Data publishes data on Haiti only — one country, treated in depth, rather than ten skimmed over.

11 707territorial entities identified, from department down to locality
5 levelsdepartment › arrondissement › commune › section › locality
2 949indicator values computed across the 140 communes
1 standardthe Fèy Done Atmart, applied to every dataset

What are you looking for?

Three ways in, depending on what you need today.

National benchmark

Public spending, per inhabitant

The Ministry of Economy and Finance publishes budget execution on budget.gouv.ht — but only at the national level, ministry by ministry: no breakdown by commune or department exists to date. Here, then, is what can honestly be published, taken from the official documents of the ongoing 2025-2026 fiscal year (cumulative from 1 October 2025 to 30 June 2026, that is 9 months out of 12).

186.4 bn HTGauthorized spending executed over 9 months — 51.75% of the revised budget of 360.3 bn
≈ 15,670 HTGper inhabitant over these 9 months (population: 2024 UNFPA projection — an order of magnitude, not a measurement)
176.0 bn HTGactual disbursements according to the TOFE (cash basis) — a different reference frame; the gap is normal and documented

Every figure is extracted from an official PDF sealed by a SHA-256 hash, with its internal checks (sum of the months, execution rate). Download the national benchmark CSV — documented columns: concept, period, source, document hash, limitations.

The problem

In our markets, the data exists. It is using it that costs.

The raw figure is rarely the real obstacle. What blocks you is the time lost between the source and the decision. That interval is exactly what Atmart Data works on.

🧩Fragmentation

Sources are scattered across institutions, donors, portals and PDFs. Nobody holds them together.

📐No standard

Commune names, units, years, codes: nothing joins without being harmonized by hand.

Methodological uncertainty

Where does the figure come from? How old is it? What does it not cover? Without an answer, it is indefensible.

Cleaning time

Two to three days of preparation per series. That cost is what we remove — not the figures, which are often public.

Our answer

A backbone, not a file warehouse

What is truly missing in Haiti is not the data: it is common identifiers. Without a shared key, a school, a market, a clinic, a price and a budget cannot be tied to the same commune or the same period. The Backbone is that foundation — and every new dataset that connects to it increases the value of all the previous ones.

1️⃣Reconstruction

PDFs, HTML tables, unstable APIs and badly structured Excel files become clean tables.

2️⃣Harmonization

A series becomes comparable again despite changes of nomenclature and method.

3️⃣Cross-referencing

Sources that do not talk to each other are linked by the same territorial and temporal key.

4️⃣Derived indicators

Measures published nowhere else: access to services, price volatility, real coverage.

5️⃣Decision-ready product

CSV, Excel, GeoJSON, Power BI, dictionary, methodological note and a starting reading.

The territory reference set is published: 11,707 entities, from department down to locality, arrondissements included. See the Backbone and its method →

Where to start

Six audiences, six ways in

The same catalogue, but not the same need. Choose the door that concerns you.

The offer

Four levels, not one more

An offer you cannot grasp in a few seconds does not sell. Here is the whole architecture.

LevelFor whomWhat you getPrice
Free General public, discovery, basic academic use Previews, samples, visualizations, learning datasets, teaching material $0
Dataset One-off need, tight budget One complete dataset, its Fèy Done, its variable dictionary, an explicit licence $15 – 34
Decision pack SMEs, consultants, NGOs, business teams Several datasets assembled around a business question + a starting reading ("decision-ready") on quotation
Institutional licence Companies, universities, public administrations, structured NGOs Multi-user access, extended catalogue, updates, methodological support annual, on quotation

The price follows the value of the transformation — scarcity, preparation, updating, time saved — not the number of rows. Licence details →

The method

Every dataset arrives with its signed data sheet

This is the Fèy Done Atmart: nine sections, the same for every dataset, in four languages. It states what the data can do — and what it cannot do.

1 · Identification

Version, date, next update, licence, contact.

2 · Description

What a row represents, volume, geographic and temporal coverage.

3 · Source and method

Official source, extraction date, cleaning performed, enrichments added.

4 · Representativeness

What the dataset represents and what it does not. Who is missing.

5 · Quality

Missing data, known limitations, possible lag in the source.

6 · Uses

Intended uses and forbidden uses, explicitly.

7 · Ethics

Personal data, legal basis, applicable rules.

8 · Versions

Change log, dataset by dataset.

9 · Commitment

A reported error is corrected in the next version.

Read the trust and method charter → · See a complete product sheet →

The programmes

Three programmes that hold the model together

Where the catalogue really stands

We apply a simple rule, to ourselves included: we only put on sale what we can deliver today. The catalogue therefore shows three states — Available Sponsorable In preparation — and never a "Buy" button on a file that does not exist yet.

As of 14 August 2026: 5 datasets on sale, 23 free resources — including the four master reference sets of the Backbone and the source register — and 5 datasets in preparation, open to sponsorship. All of them cover Haiti: regional and African datasets are withdrawn from the catalogue until the second year.

Trajectory

Haiti first, the region afterwards

Regional expansion is not a diversification: it is the replication of a method already proven on difficult ground.

  1. Phase 1 — lay the backbone (0–6 months). Territory reference set stabilized ✓, time and indicator reference sets, data-sheet standards and versioning rules, source register and legal matrix ✓, academic access framework ✓, sponsorable offers ✓.
  2. Phase 2 — the pillar products (6–12 months). Profile of the 140 communes, first geocoded school register, first versioned health register, monthly market panel, harmonized macro-financial base, first licences and first sponsorships.
  3. Phase 3 — industrialize, and only then extend (12–24 months). Multi-source cross-referencing, access to services, sector packs, university agreements. Replicating the method to the Caribbean and then to Francophone Africa only begins after the first year, once Haiti is covered.

The order is not decorative: publishing a school register before stabilizing the territory means condemning yourself to redo it. The portfolio in 5 layers →

Going further

The four method pages

They are no longer in the menu — six entries are enough to work with. They stay here, for anyone who wants to check how all this is built.