B7 · Data Literacy and Visualisation

Primary and Secondary Data

Primary data is data you collect yourself, for your own question, through surveys, measurements or observation. Secondary data is data that someone else collected for another purpose, such as a census, a government report or a published study.

6 min Beginner Jump to the activity

What you need: Nothing. This page starts from the beginning, with no programming and no formulas.

Definition

Should you count the cars passing your school yourself, or look up the city's traffic report? Which is better?

Primary data is data you collect yourself, for your own question, through surveys, measurements or observation. Secondary data is data that someone else collected for another purpose, such as a census, a government report or a published study.

In simple words

Cooking with vegetables you grew yourself is like primary data: you know exactly where they came from, but growing them takes time. Buying them from the market is like secondary data: quick and plentiful, but you must trust the seller and check the quality yourself.

Why it matters

Choosing the right kind of data saves time and avoids errors. Most AI systems are trained mainly on secondary data collected for other purposes, which is one reason they can carry hidden gaps and biases.

Key terms

Primary data
Data collected first-hand for your own purpose.
Secondary data
Data collected by someone else, for another purpose, that you reuse.
Census
An official count of every person in a country, usually every ten years.
Open data
Data published for anyone to use freely.
Metadata
Information about a dataset: who made it, when, how and what each column means.
Reliability
How far a source can be trusted to be accurate and consistent.

Explained in detail

Primary data: control at a cost

When you collect your own data, you decide exactly what to ask, whom to ask and how to record it. The data fits your question perfectly, and you know its weaknesses because you created them. The cost is effort and time, and usually a small size: a student can survey a few hundred people, not a few million.

Secondary data: scale with questions

Secondary data can be enormous and free. India's census counts every resident; the National Family Health Survey interviews hundreds of thousands of households about health; data.gov.in publishes government datasets on rainfall, schools, crops and much more. But you did not design it. The categories may not match your question, the data may be several years old, and you depend on the original collectors having been careful. Always read the metadata: when it was collected, how, from whom and what each column means.

Primary dataSecondary data
Who collects itYouSomeone else
Fits your questionExactlyPartly
Time and costHighLow
SizeUsually smallCan be huge
Knowing its flawsEasy: you made itMust read the metadata

Using both together

Good projects often combine both. Secondary data gives the big picture, and primary data fills the gap for your local situation. A student studying water use might take national figures on water supply from a government report, then survey families on their street to see how their use compares.

Secondary data and AI

Most AI is trained on secondary data: photos posted online, text from websites, past records from hospitals or banks. None of it was collected with the AI in mind. This is efficient, but it explains many problems. The photos online over-represent some countries; the records reflect past decisions, fair or not; the text contains the opinions and mistakes of its writers. Asking "what was this data originally collected for?" is one of the most useful questions about any AI system.

Most real projects use both kinds. A student studying local water problems might start with secondary data from the census and the district water department, then run a short primary survey on their own street to see whether the official picture matches what neighbours experience. Comparing the two is often where the most interesting findings appear.

How it works

ExampleType
You count birds in your garden each morningPrimary
You use the census figure for your district's populationSecondary
You interview shopkeepers about sales during a festivalPrimary
You download rainfall data from a government websiteSecondary

Try it yourself

Planning a park using both kinds of data

A ward committee wants to decide where to build a children's park.

Primary or secondary?

Who collected the data, and for what?

  • A dataset of online photos used to train an AI

  • Reading cricket statistics on a sports website

  • Using national literacy rates from a government report

  • Asking classmates their favourite sport

  • Measuring the temperature in your classroom every hour

Common mistakes

  • Using secondary data without checking when, how and why it was collected.
  • Thinking primary data is always better. Small or careless primary data can be worse.
  • Forgetting that most AI training data was collected for other purposes.

Check yourself

Question 1

1. Which is primary data?

Question 2

2. What should you check before using secondary data?

Question 3

3. Why can AI trained on secondary data carry hidden problems?

Summary

  • Primary data is collected first-hand; secondary data is reused from others.
  • Primary fits your question; secondary offers scale but needs checking.
  • Combine them, and always ask what data was originally collected for.

Finished reading?

Sources: Open Government Data Platform India (data.gov.in) · Ministry of Statistics and Programme Implementation (India)