Choose a public dataset before you build the chart

You have an idea for a chart. A search turns up several promising files, but their numbers do not quite agree. Before choosing a chart style, check whether the files describe the same thing.

Your outcome: choose a sensible starting source for one question, then explain what one row means, which dates and units it uses, and where its reuse terms are recorded. You can practise the reasoning here without an account or a download.

This is an expanded version of Anz Makes Sense’s GD01, “8 global data sources for your next project,” first published on social media on 21 September 2026. The original Facebook guide preserves that release. This website adaptation adds a worked example and practice; its source descriptions were checked on 22 September 2026.

Start with one question

“Find interesting data” gives you too many choices. Try a question with a measure, place and period: “How did the population of two countries change between 2015 and 2020?”

For that question, start with a population indicator from an international statistical publisher. For spreadsheet practice with rows and categories, a small, well-documented research or community dataset may be more useful. The eight destinations below serve different purposes; this is a starting list, not a popularity ranking or a quality guarantee for every file.

Choose a starting point

• World Bank Open Data: development indicators, including population. Start here for a country-and-year question; check the indicator definition and available years.
• Our World in Data: data and explanations about topics such as health, population and climate. Follow the underlying source notes. Its reuse guidance distinguishes its own work from third-party material.
• UN System Data Commons: a shared entry point to public statistics across the UN system. Trace the result to its contributing source and terms. The UN’s platform announcement explains that relationship.
• OECD Data Explorer: statistics on economic, education and other topics. Check which member countries and partner economies appear in the selected dataset. The official overview explains the service.
• UCI Machine Learning Repository: research datasets for machine-learning practice, including small tabular examples. Start with the dataset description and variable definitions before treating a column as a measurement.
• Kaggle Datasets: community datasets across many topics. Check who supplied the file, its original source and its licence. The dataset documentation describes the platform; a listing alone does not establish data quality.
• NASA Earthdata: Earth-observation data and discovery tools. Check the product, file format and processing requirements before downloading. Some access routes use Earthdata Login; choose a simpler table first if you are still learning spreadsheets.
• Google Dataset Search: a way to discover datasets hosted elsewhere. Its help page explains the search service. Open the actual provider’s page to inspect the data and terms.

You only need one suitable source to begin. A familiar logo does not make different measures directly comparable.

Read the notes before the numbers

For your chosen dataset, write a short source note:

1. Question: what comparison are you trying to make?
2. Row: does one row mean one country and year, one survey response, or something else?
3. Measure and unit: people, thousands of people, a percentage, or a total? What definition does the publisher use?
4. Coverage: which places and periods are included? Keep the observation period separate from the page’s update date.
5. Origin and reuse: save the dataset title, publisher, source link, version or update date, your access date, and the specific licence or terms link. Record the attribution it asks for.

Check missing values and any footnotes about estimates or changes in method. A blank value is not automatically zero. If the source or permitted reuse is unclear, keep that uncertainty visible and choose another practice file when needed. Download access and permission to redistribute are separate questions.

Keep an untouched copy of a downloaded source and do your calculations in a separately named working copy. The current-file guide shows how to distinguish a working file from an approved one.

Try a comparison with made-up data

Imagine two fictional places, Cedar and Bay. These invented records are for learning only; none comes from a linked publisher.

• Cedar: year 2024; population 12; unit thousand people.
• Bay: year 2024; population 15,000; unit people.

Comparing the raw numbers 12 and 15,000 would be misleading. Cedar’s figure means 12 × 1,000 = 12,000 people. With the same year and unit, the supplied figures put Bay 3,000 people higher. Before using real records, you would also check whether both publishers mean the same thing by population.

Now change the case: Bay’s figure of 15,000 belongs to 2022, and its 2024 value is blank. Can you still report that Bay’s population was 3,000 higher in 2024? Write one sentence explaining your decision before reading on.

Check your reasoning

You cannot establish that 2024 comparison from these records. Bay’s 2024 value is missing. Its 2022 figure cannot silently stand in for 2024, and a blank cannot become zero. You can describe the two dated observations separately, or look for a comparable period with available data for both places.

Your first useful result is a comparison you can explain: a clear question, matching definitions and units, an appropriate period, and a traceable source. Make that check before choosing colours or asking AI to write the conclusion.

About this adaptation

Written for Anz Makes Sense with AI assistance. The source list comes from our original GD01 release; the worked example and exercise are original synthetic teaching material. This page links to publishers without copying their datasets or charts. Check your selected file and its terms before using it.