You have four product concepts and one budget. Show each person a single concept and you need a lot of respondents. Show each person all four and you need fewer, but every rating is shaped by what came before it. That choice, monadic testing or sequential monadic, is one of the first design decisions in concept testing, and it decides how you sample, how you rotate and which significance test you run at the end.
Key takeaways
- Monadic testing gives each respondent one concept, so scores are clean and realistic but you need a full sample per concept.
- Sequential monadic shows several concepts in rotated order, so each person rates every concept and the total sample is much smaller.
- Four concepts at 200 completes each means 800 interviews monadic, against roughly 200 to 300 sequential monadic.
- Order and contrast effects are the price of sequential designs. Rotate, and check first-position data.
- Compare monadic cells with a two-proportion test. Compare sequential concepts within respondent, with a paired test.
What is monadic testing?
In monadic testing, each respondent sees one concept and rates it on its own. You split the sample into matched cells, one per concept, and the only difference between cells is the concept shown. Because nobody sees an alternative, the score reflects how the idea lands cold, much as it would on a shelf or in an ad.
The cost is sample size: every concept added means another full cell of interviews.
What is sequential monadic testing?
In a sequential monadic test, each respondent sees several concepts, one at a time, and answers the same questions after each, in rotated order. You get more ratings from fewer interviews.
The trade-off is that the second concept is never seen fresh. Respondents compare it, consciously or not, with the first.
Monadic vs sequential monadic: how do they compare?
| Factor | Monadic | Sequential monadic |
|---|---|---|
| Concepts per respondent | One | Several, usually three to five |
| Sample size needed | One full cell per concept | One sample rates all concepts |
| Cost | Higher, scales with each concept added | Lower overall, though interviews run longer |
| Realism | High, like meeting the idea in market | Lower, respondents see alternatives side by side in time |
| Order and contrast effects | None | Present, must be rotated and checked |
| Sensitivity to differences | Lower, cells vary by who is in them | Higher, each person is their own control |
| Fatigue | Minimal | Grows with each concept shown |
How many respondents do you need?
Take a worked example. You are testing four concepts, A to D, and want 200 ratings per concept, a common minimum for reading a percentage with a margin of error around plus or minus 7 points.
- Monadic: 4 cells × 200 completes = 800 interviews.
- Sequential monadic: each respondent rates all four, so 200 respondents already give 200 ratings per concept.
Teams usually field 200 to 300 for the sequential version, because you also want a usable read of each concept in first position, before any comparison could color it.
How do you read first-position data?
With four concepts rotated evenly, each concept appears first for a quarter of the sample. At 300 respondents, that is 75 first-position ratings per concept. At 200, only 50.
Those ratings are a small monadic test inside your sequential one. Here is concept A’s top-2-box purchase intent (the share choosing the top two boxes of a five-point scale) by the position it was shown in. The numbers are illustrative.
| Position shown | Respondents | Top-2-box | Score |
|---|---|---|---|
| First | 75 | 30 | 40.0% |
| Second | 75 | 28 | 37.3% |
| Third | 75 | 26 | 34.7% |
| Fourth | 75 | 24 | 32.0% |
| All positions | 300 | 108 | 36.0% |
The overall 36.0% is fair to every concept because each was rotated the same way. The 40.0% first-position score is closer to what a monadic test would show. On 75 respondents, its margin of error is about plus or minus 11 points, so use it to confirm direction and ranking, not as a precise score.
How do rotation and order bias work?
Order bias means a concept’s score changes with where it appears. In practice we see two patterns. Scores drift down as respondents tire, as in the table above. And a contrast effect pulls a concept down when it follows a strong one, or lifts it when it follows a weak one.
Simple rotation (ABCD, BCDA, CDAB, DABC) puts every concept in every position once, which handles the drift. It does not balance which concept comes right before which. A balanced design handles both. For four concepts, these four orders do it:
- A, B, D, C
- B, C, A, D
- C, D, B, A
- D, A, C, B
Each concept appears once in each position, and each concept directly follows every other concept exactly once. Assign respondents to the four orders in equal numbers, 75 each at a sample of 300.
What are proto-monadic and paired comparison tests?
A proto-monadic test is a sequential monadic test with a final forced-choice question: after rating each concept, respondents pick the one they prefer. You get both individual ratings and a direct preference.
A paired comparison test shows two concepts and asks which one people prefer, often on several attributes. It suits a final choice between two finalists, but says nothing about whether either is good enough on its own.
When should you choose each design?
Choose monadic testing when:
- You need a realistic absolute score to compare against an action standard or past norms.
- Concepts are complex or very similar, so one would contaminate the next.
- You can afford a full cell for each concept.
Choose sequential monadic when:
- You are screening early ideas and the ranking matters most.
- Budget or a hard-to-reach audience limits the sample.
- You want to detect smaller differences between concepts.
The questions you ask after each exposure stay the same either way. Our guide to concept testing questions covers wording and order.
How do you analyze each design?
Monadic cells are independent groups; sequential ratings come from the same people. That decides the test.
Monadic: a two-proportion test
Concept A scored 72 of 200 top-2-box (36%). Concept B scored 56 of 200 (28%). Is the 8-point gap real?
- Pool the two cells: (72 + 56) / 400 = 0.32.
- Standard error of the difference: square root of 0.32 × 0.68 × (1/200 + 1/200) = 0.0466.
- Test statistic: 0.08 / 0.0466 = 1.71.
That falls short of 1.96, the cutoff for 95% confidence. To detect an 8-point gap reliably (80% power, 95% confidence), you would need about 530 per cell.
Sequential monadic: a within-respondent test
Now the same 36% against 28%, but from 300 respondents who each rated both concepts. Because the ratings are paired, compare people, not groups. McNemar’s test uses only the people who disagreed with themselves: 45 rated A top-2-box but not B, and 21 did the reverse.
- Difference in discordant pairs: 45 − 21 = 24.
- Test statistic: 24 / square root of (45 + 21) = 24 / 8.12 = 2.95.
That clears 1.96 comfortably. Treated as two independent groups of 300, the statistic would be about 2.10, so the pairing adds real sensitivity on top of the larger base. For mean scores, the equivalent is a paired t-test. For a full crosstab of every concept by segment, use crosstabs with significance testing, but check that the test matches the design.
What are the common mistakes?
- Running independent-sample tests on sequential data. This throws away the sensitivity you paid for.
- Rotating position but not sequence. Simple rotation leaves contrast effects in place. Use a balanced design.
- Comparing scores across designs. A sequential monadic score is not comparable to a monadic norm. Keep benchmarks within one design.
- Showing too many concepts. Past five, fatigue flattens later ratings. Split them across respondents.
- Ignoring first-position data. If the ranking in first position disagrees with the overall ranking, investigate before you report a winner.
Before you field, write down whether the decision needs an absolute score or a ranking, and pick monadic testing or sequential monadic from that. Then size the sample to the smallest difference that would change the decision. When the test is done, connect the findings across studies and waves so the next concept round starts from what you learned, and see how the same discipline applies to message testing and creative testing.