What does this research actually support?
Choose a useful next test without turning a small study into a market estimate.
The question
ParcelNest helps people book parcel collection. Eight first-time users attempted the same booking task. An AI summary recommends making the map larger because seven said they liked it.
| Reason | participants | Share |
|---|---|---|
| Missed the last collection time | 5 | |
| Could not resolve their address | 2 | |
| Finished without help | 1 | |
| Total | 8 |
Invented, mutually exclusive outcomes from one small diagnostic study. Seven of these same eight participants also said they liked the map; this preference is not an additional group.
Seven of eight users praised the map. The main opportunity is a larger map and a more prominent location pin.
“What does this research actually support?”
Not saved by this site. Your browser may restore text when you navigate back.
A strong answer, step by step
Open each step in turn. This is one strong answer, not the only one.
Step 1: Separate observation from preference
What would you do here? Open the move you’d make.
Treat map praise as proof that a larger map fixes booking.
Tempting, but it misses something. A preference does not identify the cause of a failed task.
Distinguish observed failures from map preference.
The strong move. This preserves both findings without inventing causation.
Discard every preference because it is not behavioral data.
Tempting, but it misses something. Preferences can help explain behavior; they are not interchangeable with it.
See how a strong candidate says it
“Five people missed the collection time; liking the map does not explain task completion.”
Why it works: The observed behavior points to a task barrier. A stated preference can still inform later work.
Step 2: Limit the claim
What would you do here? Open the move you’d make.
Tell leadership that 62.5% of all users miss the cutoff.
Tempting, but it misses something. The study cannot support a population estimate.
State that five of eight study participants missed it.
The strong move. The finding stays attached to its actual sample and task.
Say the study is useless because it has eight people.
Tempting, but it misses something. It can diagnose problems without estimating market prevalence.
See how a strong candidate says it
“This study helps locate possible barriers. It does not estimate the share of all customers affected.”
Why it works: The sample is small and task-specific; a percentage would imply precision it does not have.
Step 3: Choose the next prototype
What would you do here? Open the move you’d make.
Change all three areas and test general satisfaction.
Tempting, but it misses something. You may not learn which change resolves the task barrier.
Ask an AI panel which layout it prefers.
Tempting, but it misses something. Synthetic responses would be hypotheses, not observed task evidence.
Test visible collection time with the same booking task.
The strong move. A targeted change can reveal whether the time information resolves the failure.
See how a strong candidate says it
“Test a visible collection cutoff and an address-confirmation step before increasing the map.”
Why it works: The prototype should help distinguish the causes that stopped participants.
Step 4: Revise with new evidence
What would you do here? Open the move you’d make.
Keep the cutoff change first for every audience.
Tempting, but it misses something. That ignores the new evidence about returning users.
Test address confirmation for the returning-user task.
The strong move. The next prototype follows the current uncertainty.
Pool both studies into a single market percentage.
Tempting, but it misses something. Different samples and tasks need context before aggregation.
See how a strong candidate says it
“If returning users know the cutoff but cannot confirm the address, I would test address confirmation first for that group.”
Why it works: A new audience and failure pattern can change the next decision.
The answer, in short
- Decision
- Test a visible cutoff for first-time users and an address-confirmation step for returning users.
- Because
- These are specific responses to observed barriers, not predictions from map praise.
- What would change it
- A controlled task test shows the map itself causes the time or address failure.
Common traps
- Calling eight participants a representative market sample.
- Presenting AI-generated user responses as research observations.
Check your answer
- Can I separate behavior, preference and interpretation?
- Did I state the sample and task?
- Does my prototype test a specific uncertainty?
- What evidence would change the next test?
Last reviewed Nov 10, 2026