Numbers deserve
better than a PDF

The short version

Imibare is Kinyarwanda for numbers or statistics. It is also a small, independent project that takes published African government data out of PDFs, spreadsheets and awkward portals, then makes it usable as data.

The catalogue currently has 216 datasets from Rwanda and Kenya. You can browse them here, download them as ordinary files, query them over HTTP or load them in Python. More importantly, you can see where they came from and what happened to them on the way here.

Why make this?

The data is often already public. That is not the same as being usable. A useful table may be buried in a scanned report, split across twenty Excel sheets or sitting behind a portal that makes a simple download feel like a small expedition. If you need one figure for a story or a paper, you can lose an afternoon getting it into a shape you can trust.

That seemed wasteful. The work of publishing a statistic should not have to be repeated by every person who needs it. imibare does the repetitive part once, carefully, and leaves the trail behind it visible.

What happens to the data here?

The original institution is still the source. imibare finds the publication, extracts the data, tidies it into a consistent shape and publishes it with the source link, retrieval date and notes about the work. If a table is incomplete, confusing or changed between editions, that belongs in the record too.

Each published change creates a version. Old versions stay available. This is useful when a source revises a release, but it is also basic courtesy to anyone trying to reproduce work that used the earlier data.

Some pipelines run on a schedule. Some need a person because the source is a PDF or a one-off publication. Both are labelled plainly as automated, curated or stale. A label is more useful than pretending every dataset was updated this morning.

Who is behind it?

One person, for now. I do not publish my real name because I would rather the project be judged by things you can check: the sources, the extraction notes, the versions and the corrections. The work is done alongside everything else, mostly in evenings and weekends.

imibare is independent. It is not funded by, endorsed by or affiliated with the governments and institutions whose data appears here. Those institutions remain responsible for what they publish. I am responsible for how imibare extracts, documents and presents it.

Can you trust it?

You should not trust it blindly. Government publications can contain errors. Extracting them can introduce errors too. A clean-looking dataset is not proof that every number in it is right.

For work that matters, check the linked original source and cite both the publishing institution and the specific imibare version you used. The point of this project is not to ask for faith. It is to make checking possible.

How can I use it?

Download any dataset as CSV, JSON or Parquet. Use the REST API if HTTP suits your workflow. Or install imibare in Python and write imi.load("rw.nisr.cpi.monthly"). There is no account, special client or proprietary format between you and the files.

Reuse depends on the source institution's terms, not on a permission from imibare. Every dataset page records what its publisher says and links back to the original release. The terms page explains the full position.

If something is wrong, say so

Please report an error. A dataset ID, the value in question and a link or citation for the original source are plenty to get started. You do not need to work out whether the problem came from the source or the extraction.

Reports are public because the response should be public too. If you would rather write privately, email hello@imibare.org.

What about other countries?

Rwanda came first and Kenya is live too. New coverage is added slowly, on purpose. The source has to be reachable, its licence position has to be understood, and the pipeline has to be something I can keep looking after. A long list of countries would be less useful than data that quietly stopped being maintained.

If you know a country's data landscape well, or want to support durable coverage, get in touch.