How a figure gets here

How our data is organised

Prism organises official statistics as a simple hierarchy. Understanding it makes the whole site easier to read: it tells you what a chart is measuring, where each number came from, and how any figure traces back to its printed source.

The site is browsed top-down through four navigation tiers — domain → subject → topic family → series. A domain is a broad grouping (for example "Economy, Work & Finance") that holds several subjects; a subject holds topic families — an entity or topic such as petrol prices; and a topic family gathers the individual series you actually chart. Beneath a series sit the two data-model terms: an observation is a single number in a series, and a dataset is the source table a number was copied from.

Observation — one number

An observation is the atomic unit: one measured value, for one place or entity, at one point in time. It is what you read off a single cell of a chart or table.

Prism keeps two forms of every value side by side: the figure exactly as printed in the source, and the parsed number used for charting. Where the source shows no figure, Prism records why it is missing rather than guessing or leaving a silent blank.

Series — one thing measured over time

A series is a named measure followed consistently across time and geography — the thing you actually chart. Every observation belongs to exactly one series. A series carries its own unit, frequency (monthly, quarterly, annual), and geographic level, and it is the unit Prism designs, names and reviews before any of its numbers are published.

Dataset — where a number came from

A dataset is one source table, captured exactly as it was published — a specific table on a specific page of a specific report. It answers "where did this come from?" rather than "what does it measure". Each dataset records its table number, its source file, and — importantly — who actually produced the figures, which is not always who published them (see A note on sources below).

The distinction between series and dataset matters because they are not the same shape. One series is often assembled from several source tables, and one source table often supplies several series.

Subject — the broad area

A subject is a top-level area of statistics. Prism organises Nigerian statistics into 23 subjects, each mapped to a UN CSA 2.0 domain and a governing international manual: Agriculture, Balance of Payments, Births, Deaths & Fertility, Communication, Education, Elections & Representation, Employment Programmes & Registrations, Financial Institutions, Foreign Trade Statistics, Geography & Reference, Health, Household Statistics, Inflation & Prices, Justice & Security, Labour Force, Manufacturing, Migration, National Accounts, Petroleum products, Population, Power, Public Finance, and Transportation. Every series and every dataset belongs to one subject, which is how the site is browsed from the home page down — see the A–Z subject index.

A second example, from a different subject

The same structure holds everywhere. Take inflation, in the Inflation & Prices subject:

• Subject: Inflation & Prices.
• Series: "Composite CPI headline inflation rate (all items), Nigeria" — a monthly national measure running January 2018 to December 2022, with 60 published observations on Prism.
• Observation: headline inflation in December 2022 was 21.34%. That one month, one figure, is a single observation.
• Dataset: the figures were compiled and published by the National Bureau of Statistics (NBS) itself and drawn from the Consumer Price Index tables of the Annual Abstract.

Served on Prism as at 2026-08-03.

This is also the clearest way to see how series differs from dataset: the series is the idea you want — "headline inflation, month by month" — while the datasets are the individual printed tables the numbers were lifted from. In the electricity case those tables were produced by NERC; in the inflation case they were produced by NBS. The series is what you measure; the dataset is where it came from.

A note on sources: who published ≠ who produced

Prism keeps two source roles apart for every dataset, and shows both:

• the publisher — who released the compilation (usually the National Bureau of Statistics), and
• the originator — the institution that actually produced the figures (for electricity, the Nigerian Electricity Regulatory Commission; for elections, the Independent National Electoral Commission; and so on).

Where NBS both produced and published a table — as with the Consumer Price Index — the two roles are the same. Where it only republished another body's figures, Prism credits the originator and never presents NBS as the source of data it did not produce.

Every figure on Prism is shown with its source and the date the data is current as of. The subject count and list above are read live from Prism's canonical registry; the worked example figures were read from Prism's public data at the time of writing (2026-08-03) and will move as the corpus grows. The structure they illustrate does not.

Where data comes from

Prism works from public statistical publications. A publication has a publisher, and individual tables may name a separate originating institution. Prism keeps those roles separate where the data model exposes them, and does not infer an originator when the source does not prove one. The Sources page explains the attribution boundary.

Processing

Source tables are extracted, normalised into datasets and series, and checked before publication. Prism preserves source values unless a documented correction is required to make the table's own structure coherent.

Validation gates

A batch is not published while a required validation gate is missing or failing. Gate names and requirements are maintained in Prism's validation vocabulary.

Publication

Published pages read from public catalogue views. Those views expose only public-facing fields, not internal batch files, scripts or prompts.

Labels and repairs

Prism aims to preserve source labels and values. Where a source label is repaired, the repair must be documented and the printed form retained in provenance fields. Labels can change when a documented repair requires it.

Revisions, corrections and withdrawals

If a source revision or Prism correction changes what should be public, Prism should update or withdraw the affected data and state what happened on the affected surface. Prism should not claim that public correction logs exist until a correction surface is built and linked.

Known limitations

Some source tables are withheld until Prism can read their structure reliably. Some published source tables contain contradictions or printing defects; Prism records those as source issues rather than silently smoothing them away.