0 of 15 checkpoints
Fact-check desk · about 12 minutes
You’ve joined the fact-check desk at The Harbour Ledger, a fictional community news site. Three data claims have landed on your desk, then a fourth arrives. Your job isn’t to prove them wrong — it’s to work out what the data can and can’t support.
Everything in this lesson — people, organisations, places, headlines and data — is fictional and was invented for teaching.
The aim is scrutiny, not blanket distrust: good data can be presented badly, and imperfect data can still tell us something.
Useful to know already: averages, percentages, and reading bar and line charts.
Case 1 of 3 · Graph framing
From The Tidings, Harbour High student newsletter Fictional
New ordering kiosks SLASH lunch queues!
“Our chart says it all,” writes the newsletter. The canteen timed the queue every lunchtime for 6 weeks before the kiosks and 6 weeks after.
Step 1 Look · worked example
Both charts show the same 12 weeks of measurements. Only the vertical range is different.
In View A the Before bar is drawn 2.25 times as tall as the After bar. The real averages are 9.4 and 8.9 minutes: a drop of 0.5 minutes, or 5.3%.
Step 2 Try · guided practice
Make the drop look as dramatic as you can, then as modest as you can. Watch what changes — and what doesn’t.
Try at least one setting starting at 1 or lower and one starting at 8 or higher.
Step 3 Think · context
Starting an axis above zero isn’t automatically dishonest. What matters is what the chart is for, whether the reader can see where the axis starts, and whether the size of the change matters in context.
Step 4 Apply · new case
The Harbour High garden club’s greenhouse should stay between 18 °C and 24 °C for seedlings. The club wants a chart to check yesterday’s hourly readings.
Case 2 of 3 · Sampling
From The Tidings, Harbour High student newsletter Fictional
Most students want the library open on Saturdays
“We asked students in the library at lunchtime,” the newsletter explains.
Step 1 Look · the population
Real fact-checkers rarely know the true answer. Here we do: a synthetic school roll of 600 students, the . In it, 282 students (47%) want Saturday opening — not a majority.
Students who use the library a lot are much keener on Saturday opening (80%) than those who rarely go (30%).
Ask students who happen to be in the library at lunch. In this model, a frequent library user is 30 times and an occasional user 6 times as likely to be picked as a student who rarely visits.
Pick names at random from the full school roll. Every student has the same chance of being chosen — a .
Step 2 Try · draw samples
No sample drawn yet. Choose a method and size, then press “Draw a sample”.
Draw at least one sample with each method.
Step 3 Try · does bigger fix it?
One sample can be lucky or unlucky. Repeating each method 40 times shows its typical result. Run it at a small size, then a large one.
No repeats yet.
Run repeats at two different sizes.
Step 4 Apply · new case
Gaming Club web page Fictional
88% say: build an esports room!
An open poll on the club’s own page asked, “Should the school build an esports room?” It received 1,240 votes.
Case 3 of 3 · Correlation
Harbour Seafront Traders’ summer report Fictional
Ice cream is making visitors lose their belongings
“On days we sell more ice cream, the lost-property office fills up,” say the traders, based on 60 summer days.
Step 1 Look · the association
Step 2 Try · compare like with like
The data record each day’s weather and visitor count. the days: if the link is really about ice cream, it should still show up when you compare similar days.
Busy days have more of everything. Compare the 15 highest and 15 lowest ice-cream days, first as raw counts, then per 1,000 visitors.
Split the days by weather and view the fairer comparison.
Step 3 Decide · justify a conclusion
Final report · apply everything
Advert from LeafyLoud Ltd Fictional company
TALK TO YOUR PLANTS: the LeafyLoud speaker makes plants grow TWICE as tall!
The advert includes three pieces of evidence. Examine each, then write your verdict for the Ledger.
LeafyLoud emailed 5,000 customers who had bought the speaker. 450 replied, and 412 of them said their plants “grew better”.
92% agree!
Response rate: 450 ÷ 5,000 = 9%.
LeafyLoud compared 200 customers’ gardens after 6 weeks.
Step 1 Investigate
Step 2 Write the verdict
A good verdict has three parts: a claim the data can support, an observation that backs it up, and a limitation.
Write your verdict as a short paragraph. It stays on this device and is not marked by the computer — use the checklist to review it yourself.
Your version doesn’t need to match. Different wording — and different limitations — can be just as good.
Case closed
This week, find one chart or statistic in an advert, article or video. Ask the four questions above and write a one-sentence cautious headline for it. If you can, compare notes with a friend: did you spot the same limitations?