See every check ATLAS runs on itself →  ·  Every dataset, publisher, date and licence →

How to use ATLAS

  1. Start with one area. Type a town or an area name into Compare my area. You get its published figures beside Ireland's, each one saying what it is and what it cannot tell you.
  2. Ask in plain words. Ask takes a question like "what is the population of Tralee?" and shows you the query it turned it into before it answers. Ask it to rank places and it will decline and say why.
  3. See one measure across the country. The map colours every area by whatever you choose, with a table of the same numbers underneath. Where a measure has several years, a Play button moves through them.
  4. Browse by subject. Themes groups every measure into ten subjects, each with a map, a chart and a trend.
  5. Check anything you would repeat. Every figure names its publisher and date. Sources lists all of them with their licences, and How ATLAS checks itself shows the tests behind them.
  6. Make it easier to read. Every page has Irish at one click, a reading font for dyslexia and low vision, and a dark appearance. On a phone, the installable app on the home page works offline once added to your home screen.

One thing ATLAS will not do at any step: tell you what an area needs, who qualifies for anything, or place areas in an order against one another. It shows what is published and leaves the conclusion to you.

A proposed pilot

We propose testing ATLAS with community representatives preparing for a local meeting, using housing and population evidence for one area. The test is whether they can find relevant figures, trace the sources and identify what still needs investigation. This user and task remain hypotheses.

A proposed four-week pilot would compare the same task with participants' usual sources and with ATLAS, measuring completion time, correct source identification and misunderstandings. No user benefit, commissioner or funding commitment has yet been validated.

Operation and maintenance

A local authority or public participation network is a candidate pilot partner. A named owner would oversee source updates, licence checks, corrections and support. Static hosting may use a free allowance, but maintenance takes staff time and hosted AI requests can cost money. A pilot must measure those costs before proposing a recurring service.

How to read a label

Every important statement in ATLAS carries one of eight labels. That way, smooth wording cannot hide the difference between what was measured and what was worked out from it.

Verified fact
Stated directly by a named official source, for the place and period given.
Calculated indicator
Worked out from named source figures by a fixed formula that anyone can repeat.
Inference
A conclusion drawn from the evidence given, with the reasoning shown.
Scenario
A “what if” result, worked out from stated inputs and assumptions. Never called a forecast.
Forecast / projection
The result of a named method, saying who is responsible for it, when it was made, how far ahead it looks and how wrong it has been before.
Assumption
A value or choice that the data does not establish, kept in view.
Unknown
Not available, not comparable, out of date, too uncertain, or never measured. Shown as no data, not zero.
Judgement required
A decision for a named professional, a legal process or a public body.

The data behind ATLAS

What is missing is said, not hidden. This list names every source ATLAS means to use, including ones not added yet. A theme showing mostly Being considered is one we have planned, not one we have delivered. Nothing is left out to make the list look better than it is.

Every source ATLAS uses or has identified, by theme, with where each one stands. Whether the checks on that data passed is on How ATLAS checks itself.

Measure and dataset names, and the registry detail describing them, are shown as their publisher issued them, not translated for this prototype.