Research literacy:
read studies and data well

Research literacy is the ability to find, read, evaluate and use research — academic studies, industry surveys, market reports, internal analyses and experiments. It includes recognising study designs (experiments, observational studies, surveys, case studies); judging samples and bias; distinguishing correlation from causation; understanding statistical significance, confidence intervals and effect sizes; checking funding and conflicts of interest; spotting cherry-picking and overclaiming; and weighing individual studies against systematic reviews and the broader evidence.

The need is real even in science. In 2015, the Open Science Collaboration replicated 100 studies from three psychology journals: 97% of originals reported significant results, but only 36% of replications did, and effects were about half as large on average.¹ A decade earlier, John Ioannidis argued in PLoS Medicine that many published findings may be false, particularly with small studies and flexible analyses.² Surveys have their own pitfalls; Pew Research Center's guidance on writing survey questions shows how wording changes answers.³ Employers want people who can reason with evidence — the WEF ranks analytical thinking the top core skill.⁴ Research-heavy roles include research analyst, listed by AmbitionBox at ₹5.3–5.9 lakh in India.⁵

Psychology replications with significant results (Science, 2015)

36%

Original studies with significant results (Science, 2015)

97%

Replication effect sizes relative to originals (Science, 2015)

~½

Employers calling analytical thinking essential (WEF, 2025)

7 in 10

Table: How research literacy is assessed (editorial summary)

Table: Types of evidence and their limits (editorial summary)

Table: How evidence practice developed (editorial summary)

• A critique of a real study or report
• A correlation-vs-causation example from your work
• A summary weighing several sources
• A survey you reviewed for bias
• An experiment you designed or analysed
• A verified research literacy certificate

Show evidence-based impact: "Reviewed vendor research claims; designed A/B test that showed true effect; prevented costly default change." Real examples only.

"How do you decide whether to trust a study?"

Check design, sample, who funded it, effect size, whether results replicate and what other studies show.

"What's wrong with 'users who do X retain better'?"

It's a correlation; those users may differ in other ways. An experiment is needed to show causation.

"Significant but tiny — does it matter?"

Statistical significance doesn't mean practical importance; look at effect size and cost.

"How do you summarise conflicting evidence?"

Weigh study quality and size, explain disagreements and state confidence levels.

"How do you spot a biased survey?"

Look at sampling, response rates, question wording and who commissioned it.

• "How does the team use research in decisions?" — culture.
• "Do you run experiments?" — rigour.
• "Which data sources are trusted?" — inputs.
• "How are findings shared?" — communication.

Table: Research-heavy roles and typical pay in India (AmbitionBox, October 2026; indicative)

Table: Common research literacy tools (editorial guidance)

• Analyst roles — sound conclusions.
• UX and product research — evidence-based design.
• Healthcare and policy — evidence-based decisions.
• Journalism and communications — accurate reporting.
• Leadership — better strategy.

1. Week 1: Designs — Learn study types and their limits.
2. Week 2: Practise — Critique three real reports with a checklist.
3. Week 3: Prove it — Write an evidence summary; take the Research Literacy Certification.

Related: Media Literacy, AI Information Verification and Critical Thinking. Guides: media literacy, AI information verification and statistical reasoning.

• Reports — reading critically.
• Vendor claims — checking evidence.
• Experiments — designing tests.
• Dashboards — interpreting changes.
• Presentations — stating confidence.
• Decisions — weighing evidence.

"What's a confounding variable?"

A factor that affects both the supposed cause and the effect, creating a misleading association.

"What's publication bias?"

The tendency for positive results to be published more than null results.

Table: A first-90-days plan (editorial guidance)

Learn which data and reports your organisation relies on. Critique them respectfully and suggest improvements. By the third month, design or run a test that answers a real question properly.

AI tools now search literature, summarise papers and extract findings in seconds. They can also misstate results, invent citations or miss limitations. Research-literate professionals use AI to find and skim, then read methods and results themselves, check citations exist and verify key numbers against the original.

Read the abstract for the claim. Jump to methods: who was studied, how many, what design. Check results for effect sizes and confidence intervals, not just p-values. Read limitations. Then ask whether the conclusion matches the evidence. Note what other studies say.

Watch for causal language ("causes", "proves") from correlational studies, small samples presented as general truths, percentages without baselines, and headlines that go beyond the paper. Check whether the study population matches the people the claim is about.

State the question and the bottom line first. Give a confidence level — high, medium or low — with reasons. Summarise the strongest evidence and key limitations. Note what would change your view. Keep detail in an appendix.

Define the question and the single change you want to test. Randomly assign people or units to the change and to a comparison group, so other differences balance out. Decide the main outcome measure and sample size before you start. Run the test long enough to capture normal variation, then compare results, including the size of the effect and its uncertainty. Simple, well-run experiments answer causal questions far better than looking back at existing data.

"Published means proven." Many findings don't replicate.

"Significant means important." Effect size matters.

"More data always means truth." Biased data at scale is still biased.

Check the title, axes, units and source. Look for truncated axes that exaggerate differences, and for missing baselines or time periods. Ask what comparison the chart implies and whether it's fair. Find the underlying numbers if possible, especially before quoting the chart in your own work.

• Treating correlation as cause.
• Ignoring sample bias.
• Relying on one study.
• Quoting p-values without effects.
• Trusting AI summaries blindly.

Replication data come from the Open Science Collaboration (2015); the false findings argument from Ioannidis (2005); survey design guidance from Pew Research Center; skill priorities from the WEF Future of Jobs Report 2025; and India pay from AmbitionBox (read October 2026). The dark mode scene, evidence table and history are editorial. Science of Apps publishes the free test linked here. Last reviewed October 2026.

• Study design — how research is structured.
• Correlation — two things moving together.
• Causation — one thing causing another.
• Effect size — how big a difference is.
• Replication — repeating a study to check results.