Count Lab day. Keeper Ines lends the crew the Commons sign-in book.
"Every member writes down when they usually visit," she says. "All 120 of them."
Comet sets the book next to the survey totals. "Evening got 15 answers. Morning got only 8."
Wren runs a finger down both columns. "What do you notice? There are more evening members, but not that many more."
Nova hovers over the table and projects a dotted line. "Would you like a hint?" she asks. "Forty answers from 120 members is one in three. What if every group answered at that rate?"
Comet picks up a pencil. "Then we can see who was over-heard and who was under-heard."
You will compare who answered with who belongs. Then you will plan fixes for the groups the survey under-heard.
A group's fair share is the number of answers it would give at the overall rate. Here that rate is one in three.
| Usual visit time | Members (sign-in book) | Answers (Q2) | Fair share (members divided by 3) | Difference |
|---|---|---|---|---|
| Morning | 30 | 8 | (you work it out) | (you work it out) |
| Afternoon | 30 | 10 | (you work it out) | (you work it out) |
| Evening | 36 | 15 | (you work it out) | (you work it out) |
| Weekend | 24 | 7 | (you work it out) | (you work it out) |
| Total (made-up data from the Commons Count story) | 120 | 40 | 40 | 0 |
| Audit finding | True or false? |
|---|---|
| The four fair shares add up to 40. | ? |
| Weekend members gave 1 fewer answer than their fair share. | ? |
| The audit shows the survey heard every group equally. | ? |
Wren reminds the crew: if the people picked for a sample could simply drop out, the sample would quickly become less valid.
So the fix is to reach the under-heard groups, not to collect extra evening answers.
Collaboration that includes diverse perspectives helps avoid bias. Ask morning and weekend members how they would like to be asked.
Careful, fair work, data analyst. Tomorrow you will meet projects that gather many voices at once.