Ireland's Dashboard

Trust

Methodology

Ireland's Dashboard brings official statistics into a common place × measure × period model. Maps, charts and exports all read the same first-party store — not live upstream agency APIs at request time. This page explains what is official, what we calculate, and how we correct mistakes.

Value status

Every number carries a status. Explore shows Official for observed source values; Calculated, Estimated and Projected are never presented as if they were published that way by the agency.

Official (observed)
Directly sourced from an agency table (or a documented official revision). This is the preserved official truth in our store.
Calculated
Deterministic transform from official rows (percentage change, county sum of local-authority parts, per-100,000 overlay). Method id is shown where used. Official inputs remain stored separately.
Estimated / Projected
Modelled or forward-looking values. Not used in the current 12-indicator vertical slice. When they appear, they will be labelled and methodology disclosed — never as a silent replacement for an official observation.

Standardisation

  • Source tables are ingested via Python ETL into one analytical cube: indicator × geography × time → value + status + provenance.
  • Labels are mapped through a controlled taxonomy (“the Bible”). Unmapped labels are quarantined — we do not invent mappings at request time.
  • Index series may apply an always-on rebase to a canonical base period; raw published values can be retained for audit where recorded.
  • LLMs are not used row-by-row on data payloads at runtime. Offline batch metadata help only.

Geography

Geography is a platform capability: each indicator declares which levels it supports and a finest defensible ceiling. Explore offers only declared ∩ ingested places. Map zoom is not the same as statistical depth — grey counties mean not available, not missing colour.

County outlines on the map are first-party Tailte Éireann geometry (CC BY 4.0). Statistics always come from Ireland's Dashboard API, never Tailte or CSO at request time. Garda divisions and stations are overlapping operational areas: they appear in the Explore place list when an indicator publishes them, and are not painted onto the county map.

Methods registry

Versioned method ids used in the current product. Later estimation will attach to the same registry pattern.

geo.county_from_parts.v1

County from local-authority parts

When a CSO table publishes several official local-authority (or city/county) figures that were historically mapped onto one county code, Explore sums those official parts for the county map unit.

How:
At one period and one attr cut (for example Both sexes), collect every distinct official observation on that county code, add the headcounts, and label the result Calculated with this method id. A single part stays Official.
Not official:
The summed county total is not a figure the CSO published for that historic county polygon. Cite the official local-authority rows, or cite Ireland's Dashboard as the source of the sum.
Persistence:
Computed for display from observed rows; not a separate stored series

rate.per_100k.v1

Rate per 100,000 population

Client-only overlay: (count ÷ observed population) × 100,000 at the same place and overlapping period.

How:
Uses the selected count series and pop.total observed population after place and dimension filters. Never writes to the fact store. Official counts and population rows are unchanged.
Not official:
Not a CSO-published rate. Do not cite it as an official crime rate, birth rate, or similar unless the CSO publishes that exact series.
Persistence:
Display-only overlay — not stored in the fact table

pop.pct_change.v1

Population percentage change

Inter-period percentage change derived from FY003B population rows and stored as calculated facts.

How:
For each place and period, compute percentage change from the previous comparable observation. Rows are written with value_status=calculated and this method id. Observed population rows are never overwritten.
Not official:
The percentage-change series is Ireland's Dashboard calculation, not a CSO observation, even when the inputs are official.
Persistence:
Stored as calculated rows alongside observed facts

rsa.provisional_review.v1

RSA provisional fatality extract

Human-checked extract from a named RSA Provisional Review of Fatalities publication, stored as an estimated series that is not spliced onto CSO ROA12.

How:
A reviewed JSON fixture lists the RSA public-road death total, road-user splits and published speed-limit band aggregates for one calendar year. Rows are appended with value_status=estimated and this method id. Official ROA12 observations are never overwritten or extended by this extract.
Not official:
Not a CSO observation. Do not cite it as the next year of persons killed on roads (ROA12). RSA figures are provisional and subject to change. Garda totals that include non-public-road places are a different count.
Persistence:
Stored as calculated rows alongside observed facts

garda.ytd_fatalities.v1

Garda year-to-date fatality extract

Human-checked extract from An Garda Síochána Roads Policing year-to-date traffic fatality statistics, stored as an estimated series that is not spliced onto CSO ROA12 or the RSA annual extract.

How:
A reviewed JSON fixture stores the Garda YTD count, as-of date and source URL. Rows are appended with value_status=estimated and this method id. Official ROA12 observations are never overwritten.
Not official:
Not a CSO observation and not RSA public-road-only figures. Garda totals include some non-public-road public places. Provisional and subject to change.
Persistence:
Stored as calculated rows alongside observed facts

ytd_linear_annualize.v1

Linear year-to-date annualisation

Display-only full-year projection: year-to-date count × days in year ÷ day-of-year of the as-of date, rounded to the nearest integer.

How:
Uses the Garda YTD count and as-of date. Does not write to the fact store. Does not model seasonality, weekends or later months.
Not official:
Not an official forecast. Do not cite it as a Garda, RSA or CSO year-end figure.
Persistence:
Display-only overlay — not stored in the fact table

Indicators in this release

Vertical slice of 60 approved indicators. Each page deepens definition, licence, geography ceiling and caveats.

Revisions and corrections

Official statistics change when agencies revise tables. Ireland's Dashboard may also fix mapping or calculation errors. We keep official observations distinct from our own calculations.

Official source revisions

When CSO (or another agency) revises a published table, we re-ingest. ETL appends by default: new observed rows are added; we do not silently overwrite an official observation with a calculated or estimated value.

Dashboard methodology fixes

If we discover a wrong geography mapping, attr filter, or calculation (for example an incomplete local-authority sum), we fix the method or mapping, document the method id/version where relevant, and correct the public Explore view. Calculated series may be recomputed; observed rows stay as official facts.

What we will not do

We will not invent missing geography detail to make a map look complete, hide value status to look cleaner, or present Ireland's Dashboard calculations as if the CSO published them.

How to report an error

If a figure looks wrong, note the indicator, place, period, and Sex/other dimensions shown in Explore, and contact the operator via the site domain (irelandsdashboard.ie). There is no public comment board — this is a push-only civic resource.

Subjects · Explore all measures