1 / 15

Session 3 of 8

Designing Better Surveys

How you ask a question shapes the answers you get. Today we learn to write survey questions that actually work.

📊 Data Science for Young Minds · Grade 4

2 / 15

Opening Hook

What's Wrong With This Survey?

"Don't you agree that our terrible, boring school lunches should be completely changed immediately?"

Try to answer this question. What problems do you notice?

3 / 15

Today's Plan

What We're Doing Today

🔍 What makes a survey question bad — leading, vague, double-barreled
📏 Likert scales — measuring opinions on a 1–5 scale
☑️ Multiple-choice vs. open-ended questions
🔬 Piloting — testing your question before launch
✏️ Redesign 3 bad surveys + pilot with a partner

4 / 15

Lesson 1

What Makes a Bad Survey Question?

🚩 Leading: "Don't you think recess is too short?" — tells you how to answer
🚩 Loaded: "Do you still forget to do your homework?" — assumes something bad
🚩 Vague: "What do you think about school?" — too broad to answer usefully
🚩 Double-barreled: "Do you like math and science?" — two questions in one
🚩 Confusing: "Do you sometimes maybe occasionally read?" — unclear and awkward

A bad question gives you bad data — even if 100 people answer it.

5 / 15

Lesson 2

The Likert Scale

A Likert scale lets people show how strongly they agree or disagree — not just yes or no.

1

Strongly Disagree

2

Disagree

3

Neutral

4

Agree

5

Strongly Agree

Example: "I enjoy doing homework. (1–5)"
This captures nuance — someone who circles 2 feels very differently from someone who circles 4.

6 / 15

Lesson 3

Multiple-Choice vs. Open-Ended

☑️ Multiple-Choice

Respondents pick from options
Easy to count and graph
Fast to analyze
May miss unexpected answers
Best for: categorical data

📝 Open-Ended

Respondents write anything
Rich, detailed answers
Hard to summarize or graph
Best for: exploring new topics
Use sparingly in surveys

7 / 15

Lesson 4

Piloting Your Survey

Before you send a survey to 30 classmates, pilot it — test it on 1–2 people first.

🔬 Read each question aloud to your pilot person
👂 Listen to how they interpret it
❓ Ask: "Was that question clear? Did any word confuse you?"
✏️ Revise any question they found confusing

Real researchers always pilot. Finding a confusing question before you collect data is much better than finding out after.

8 / 15

Activity Time!

Survey Redesign Challenge

You'll see 3 badly written survey questions. For each one:

🔍 Identify the problem (leading? vague? double-barreled?)
✏️ Rewrite the question to fix it
🔬 Pilot your new version with your partner
📝 Record whether your partner found it clear

⏱ Redesign: 15 minutes · Pilot: 8 minutes

9 / 15

Bad Survey 1

Redesign This Question

❌ Original (Bad)

"Don't you think recess should be longer?"

Problem: This is a leading question — it pushes the respondent toward "yes."

✅ Your Mission

Rewrite this as a neutral question. Consider: should it be multiple-choice or a Likert scale?

Write your redesign on your worksheet. Then pilot it with your partner.

10 / 15

Bad Surveys 2 & 3

Two More to Fix

❌ Bad Survey 2

"What do you think about food?"

Problem: Too vague — could mean anything. No useful data.

❌ Bad Survey 3

"Do you sometimes maybe occasionally read books at home?"

Problem: Confusing wording — three vague time words in one question.

Fix both. For Survey 2: make it specific. For Survey 3: make it clear with answer choices.

11 / 15

🧠

Brain Break — Thumbs Survey!

Your teacher will read a statement. Show your Likert score with your fingers (1–5)!

"I enjoy working in groups." · "I feel confident doing math." · "I like learning about data."

Look around — everyone gave different numbers! That's what Likert captures.

12 / 15

Debrief

Share Your Redesigns

"Read your redesigned question aloud. Did your partner find it clear when you piloted it? What did you learn?"

Which original question was hardest to fix? Why?
Did any group write a question that was still a little confusing?
What's the most important rule for a good survey question?
When would you use a Likert scale vs. multiple-choice?

13 / 15

Rules for Good Survey Questions

The Survey Quality Checklist

✅ One idea per question — no double-barreled questions
✅ Neutral wording — no hints about how to answer
✅ Clear and specific — readers know exactly what you're asking
✅ Answer choices that cover all options (include "Other" if needed)
✅ Format matches the data type (Likert for opinions, multiple-choice for categories)
✅ Piloted on at least one person before launching

14 / 15

Vocabulary Review

Words to Know

Likert scale

A 1–5 (or similar) rating scale measuring strength of opinion

Pilot

To test a survey on a small group before full launch

Revision

Improving a question based on feedback from testing

Neutral

Wording that doesn't push toward any particular answer

Leading question

A question that hints at or pushes toward a specific answer

Open-ended

A question where the respondent writes their own answer

15 / 15

Wrap Up

Session 3 Complete!

✅ Bad survey questions give you bad data — even with many respondents
✅ Leading, vague, and double-barreled questions are the most common problems
✅ Likert scales capture nuance that yes/no questions miss
✅ Piloting reveals confusing questions before they ruin your data
✅ Good survey writing takes practice — and revision is always worth it

🔮 Coming up — Session 4: Now that we can collect data well, how do we organize numerical data? Line plots and stem-and-leaf plots!