PersonaChoice case studies
A thorough comparison between real survey data (humans) and PersonaChoice (AI).
See three validation results that showed a high agreement rate with real surveys.
The results on this page were produced with the latest PersonaChoice v3.0. Evaluation uses a unified standard: “how highly PersonaChoice ranks the items that performed best in the real survey” (Top-K agreement, Top-1 match).
This page presents validation based on published LINE Research cases.
By comparing how closely PersonaChoice’s simulation matches the real survey results, we assess its usefulness and reliability as a marketing tool.
-
This validation uses PersonaChoice’s AI personas as respondents. What is an AI persona?
An AI persona simulates human decision-making based on profile data such as age, gender, values, and personality traits. - All simulation result reports run during the validation are public. They can be viewed from the “task report” link in each simulation summary. Reports include the number of respondents per option and the reasons for each choice (within 100 characters), and are in Japanese.
Case 1
The 20s’ “aesthetic” vs. the 60s’ “concerns.” Fully reproducing the generational difference in appeal.
Research overview
- Source: LINE Research
- Date: January 2024
- Respondents: 1,053 women aged 15–69
- Format: Multiple choice
Question
From 9 toner ad banners, choose the ones you would want to click.
Survey result
Source: LINE Research
Top 3 for teens & 20s
1st
2nd
3rd
Top 3 for 50s & 60s
1st
2nd
3rd
PersonaChoice simulation result (Case 1)
Simulation 1 overview
- Run by: PersonaChoice
- Date: January 2026
- Respondents: 100 women aged 18–29
- Format: Single choice (LINE Research was multiple choice)
- Result: View the detailed report (JA)
Result: Top 3 for teens & 20s
1st 24.0%
2nd 14.0%
3rd 13.0%
Simulation 2 overview
- Run by: PersonaChoice
- Date: January 2026
- Respondents: 100 women aged 50–69
- Format: Single choice (LINE Research was multiple choice)
- Result: View the detailed report (JA)
Result: Top 3 for 50s & 60s
1st 28.0%
2nd 16.0%
3rd 13.0%
- ✓Reproduced the flip of #1 by generation: the 20s chose self-focused, aesthetic banners (e.g., “skin-type diagnosis”) #1, while the 50s–60s chose “dark-spot/dullness” appeals #1.
- ✓50s–60s result: matched the #1-ranked “dark-spot/dullness” appeal that older respondents most chose in the real survey.
This validation confirmed that PersonaChoice can reproduce the selection tendencies of women by generation with a high degree of agreement.
Episode: 300-person, population-faithful overall simulation (Case 1)
Simulation overview
- Run by: PersonaChoice
- Date: January 2026
- Respondents: 300 women aged 18–69
- Format: Single choice (LINE Research was multiple choice)
- Result: View the detailed report (JA)
Result
1st 19.3%
2nd 15.7%
3rd 12.7%
4th 12.0%
5th 9.7%
- ✓Matched the #1-ranked “dark-spot/dullness” appeal older respondents most chose in the real survey
- ✓Reproduced all top-3 banners published in the real survey (clarity, skin diagnosis, dark spots) within its Top 3
Because PersonaChoice is faithfully based on population distribution data published by the Japanese government, a 300-person sample naturally has a higher proportion of 50s–60s, matching the real population make-up. As a result, this simulation reproduced all top-3 banners published in the overall (ages 15–69) real survey within its top 3 (Top-3 match 100%, Grade A).
On the other hand, when you want to see the reaction of a specific age group, running with a narrowed generation is effective to avoid the influence of age composition. Based on this idea, we ran the individual simulations above for women in their teens/20s and women in their 50s/60s.
Case 2
The words that land differ by age. Spotting the phrasing that resonates with your target.
Research overview
- Source: LINE Research
- Date: April 2024
- Respondents: 1,053 women aged 15–69
- Format: Multiple choice
Question
When planning a new yogurt-flavored carbonated drink, choose the ad tagline (of 4) you would want to click.
- Gut care done in one bottle! Easy daily health support
- A drinkable beauty serum — beautiful from within by balancing your gut
- The power of 100 billion lactic acid bacteria! Balance your gut with fiber and feel great
- The fiber of 3 heads of lettuce plus the benefits of lactic acid bacteria support your gut!
Survey result
Source: LINE Research
Top 4 for 20s
- 1st: Gut care done in one bottle! Easy daily health support
- 2nd: A drinkable beauty serum — beautiful from within
- 3rd: The power of 100 billion lactic acid bacteria! Balance your gut with fiber
- 4th: The fiber of 3 heads of lettuce plus lactic acid bacteria support your gut!
Top 4 for 60s
- 1st: The power of 100 billion lactic acid bacteria! Balance your gut with fiber
- 2nd: A drinkable beauty serum — beautiful from within
- 3rd: The fiber of 3 heads of lettuce plus lactic acid bacteria support your gut!
- 4th: Gut care done in one bottle! Easy daily health support
PersonaChoice simulation result (Case 2)
Simulation 1 overview
- Run by: PersonaChoice
- Date: January 2026
- Respondents: 100 women aged 20–29
- Format: Single choice (LINE Research was multiple choice)
- Result: View the detailed report (JA)
Result: Top 4 for 20s
- 1st 43.0%: Gut care done in one bottle! Easy daily health support
- 2nd 20.0%: The fiber of 3 heads of lettuce plus lactic acid bacteria support your gut!
- 3rd 19.0%: A drinkable beauty serum — beautiful from within
- 4th 18.0%: The power of 100 billion lactic acid bacteria! Balance your gut with fiber
Simulation 2 overview
- Run by: PersonaChoice
- Date: January 2026
- Respondents: 100 women aged 60–69
- Format: Single choice (LINE Research was multiple choice)
- Result: View the detailed report (JA)
Result: Top 4 for 60s
- 1st 34.0%: Gut care done in one bottle! Easy daily health support
- 2nd 29.0%: The fiber of 3 heads of lettuce plus lactic acid bacteria support your gut!
- 3rd 24.0%: The power of 100 billion lactic acid bacteria! Balance your gut with fiber
- 4th 13.0%: A drinkable beauty serum — beautiful from within
- ✓20s result: matched the #1-ranked, most resonant copy (“gut care — easy & healthy”)
- ✓60s result: reproduced the older tendency to prefer the “lettuce (fiber, practical benefit)” appeal more than the 20s
- • It hit the copy younger respondents most prefer (#1) and reproduced the top selection tendencies
- • It was also confirmed that older respondents’ preference tendencies — including the direction of the generational difference — can be reproduced appropriately
Episode: the same result even with 50 respondents (Case 2)
Simulation 1 overview (50 people)
- Run by: PersonaChoice
- Date: January 2026
- Respondents: 50 women aged 20–29
- Format: Single choice (LINE Research was multiple choice)
- Result: View the detailed report (JA)
Result: Top 4 for 20s
- 1st 40.0%: Gut care done in one bottle! Easy daily health support
- 2nd 28.0%: The power of 100 billion lactic acid bacteria! Balance your gut with fiber
- 3rd 26.0%: A drinkable beauty serum — beautiful from within
- 4th 6.0%: The fiber of 3 heads of lettuce plus lactic acid bacteria support your gut!
Simulation 2 overview (50 people)
- Run by: PersonaChoice
- Date: January 2026
- Respondents: 50 women aged 60–69
- Format: Single choice (LINE Research was multiple choice)
- Result: View the detailed report (JA)
Result: Top 4 for 60s
- 1st 40.0%: Gut care done in one bottle! Easy daily health support
- 2nd 32.0%: The fiber of 3 heads of lettuce plus lactic acid bacteria support your gut!
- 3rd 26.0%: The power of 100 billion lactic acid bacteria! Balance your gut with fiber
- 4th 2.0%: A drinkable beauty serum — beautiful from within
- ✓20s result: matched the #1-ranked, most resonant copy (“gut care — easy & healthy”)
- ✓60s result: reproduced the older preference for the “lettuce (fiber, practical benefit)” appeal over the 20s
• The #1 “gut care (easy & healthy)” matched with both 50 and 100 respondents, and the top ranks were almost identical
Why: AI personas have extremely little “noise.” They lack the day’s mood, immediate prior behavior, and environmental fluctuations found in human responses, so the main selection tendency is clear even at a scale of 30–50 people.
Case 3
Ideals or practicality? Simulating even how “values shift” with economic conditions.
Research overview
- Source: LINE Research
- Date: June 2025
- Respondents: 4,734 men and women aged 15–24
- Format: Multiple choice
Question
If you were to choose a company to work for (in the future), what would you especially value? (Up to 3)
- Good relationships / a good workplace atmosphere
- Work that is rewarding / fulfilling
- Convenient commute
- A company in a stable industry
- Being able to take proper time off
- A large company
- Work that is what you want to do / are good at
- Work that matches your abilities / aptitude
- Good compensation (salary, bonuses)
- Substantial benefits and support programs
Survey result
Source: LINE Research
PersonaChoice simulation result (Case 3)
Simulation 1 overview (no environment set)
- Run by: PersonaChoice
- Date: January 2026
- Respondents: 100 enrolled students, men & women aged 18–24
- Format: Single choice (LINE Research was multiple choice)
- Simulation context: none
- Result: View the detailed report (JA)
Result
- 1st 16.0%: Work that matches your abilities / aptitude
- 2nd 14.0%: Being able to take proper time off
- 3rd 13.0%: Work that is what you want to do / are good at
- 4th 11.0%: Convenient commute
- 5th 10.0%: Good relationships / good atmosphere
- 6th 10.0%: A company in a stable industry
- 7th 8.0%: Good compensation (salary, bonuses)
- 8th 8.0%: Substantial benefits and support programs
- 9th 7.0%: Work that is rewarding / fulfilling
- 10th 3.0%: A large company
Simulation 2 overview (with environment set)
- Run by: PersonaChoice
- Date: January 2026
- Respondents: 100 enrolled students, men & women aged 18–24
- Format: Single choice (LINE Research was multiple choice)
- Simulation context: “Prices have risen sharply in recent years, and a weak yen continues.”
- Result: View the detailed report (JA)
Result
- 1st 16.0%: Being able to take proper time off
- 2nd 16.0%: Good compensation (salary, bonuses)
- 3rd 13.0%: Work that matches your abilities / aptitude
- 4th 10.0%: Work that is what you want to do / are good at
- 5th 9.0%: Good relationships / good atmosphere
- 6th 9.0%: A company in a stable industry
- 7th 8.0%: Work that is rewarding / fulfilling
- 8th 7.0%: Convenient commute
- 9th 7.0%: Substantial benefits and support programs
- 10th 5.0%: A large company
What is the “simulation context”?
The “simulation context” is a setting available in PersonaChoice’s Advanced mode.
By setting a “simulation context,” you can run a decision simulation that reflects a specific social situation or premise.
For example, under a specific environment such as ongoing price inflation or right after an earthquake, you can simulate what choices people tend to make, enabling more realistic validation.
- ✓With the environment set, “salary/compensation” rose from 7th to 2nd
- ✓Matched the real survey’s #1-ranked top item, “being able to take proper time off”
- ✓Reproduced all of the real survey’s Top 5 (5/5)
By appropriately setting the context (social environment), it was confirmed that PersonaChoice can reproduce tendencies very close to the real survey.
PersonaChoice’s strengths and limits
Strengths
Fast and accurate
Get high-accuracy research results in minutes
Low noise
Statistical convergence is fast; tendencies are clear even with small samples
Context adaptability
Results can be adjusted via the “simulation context” setting
Limits
Surveys asking for specific product/service names
e.g., “What manga are you reading lately? Which subscription services do you use?”
Real-time current-affairs surveys
Approval ratings, latest news, etc.
* Since PersonaChoice v2.0 (February 2026), tasks requiring real-time current affairs and up-to-date information are also supported.
Because AI personas don’t actually live day-to-day, they are not suited to surveys that depend on real-time experience
Validation scoreboard
Evaluated by how highly PersonaChoice ranks the items that performed best in the real survey (Top-K agreement, Top-1 match). To rule out chance matches, K is adjusted by the number of options (10+ options = Top-5, 6–9 = Top-3, 4 = Top-1 + Top-2).
86%
Achievement of “#1 hit or top-item match”
(6/7 cases)
78%
Top-item agreement
(overall average)
100%
Rated B or above (7/7)
| Case | Segment | Top-1 | Agreement | Grade |
|---|---|---|---|---|
| Case 1: toner banners | Overall (18–69, 300) | Top-3: 100% | A | |
| Case 1: toner banners | Women 50s–60s | ✓ | Top-3: 67% | B+ |
| Case 1: toner banners | Women 20s | Top-3: 67% | B | |
| Case 2: taglines | Women 20s (100) | ✓ | Top-2: 50% + Top-1 match | B+ |
| Case 2: taglines | Women 20s (50) | ✓ | Top-2: 50% + Top-1 match | B+ |
| Case 3: job values | No environment | Top-5: 80% | B+ | |
| Case 3: job values | With environment (inflation, weak yen) | ✓ | Top-5: 100% | A |
Grade distribution: A × 2 / B+ × 4 / B × 1 (B or above: 7/7 = 100%)
Evaluation criteria
| A | Agreement ≥ 80% + Top-1 match, or agreement 100% |
| B+ | Agreement ≥ 80%, or agreement ≥ 60% + Top-1 match |
| B | Agreement ≥ 60% |
| C | Agreement ≥ 40% |
| D | Below the above |
* The toner banners are evaluated on a Top-3 basis because the real survey published only its top 3 banners. The women-60s case in Case 2 is presented in the narrative section above as a contrast group for the generational difference.
Conclusion
PersonaChoice’s reliability and value
The high accuracy proven by three case studies
Case 1: toner ads
Reproduced how the chosen banner flips by generation. Matched the #1-ranked “dark-spot/dullness” appeal most chosen by 50s–60s.
Case 2: yogurt soda
Matched the #1-ranked copy most resonant with the 20s (gut care). Reproduced how the resonant appeal changes by generation.
Case 3: job values
With context (inflation, weak yen) set, hit the real survey’s #1 and fully reproduced the top 5 (Top-5 match 5/5).
Reproducing the real survey’s “rank tendency and distribution”
at an agreement rate of 80–90%
Specifically, it achieved “hitting the real survey’s #1, or matching 80%+ of the top items” in 6 of 7 cases (86%). Top-item agreement averages 78% overall, 100% in the best case, with a grade of B or above in 7/7.
Three innovative values PersonaChoice delivers
Fast, high-accuracy simulation
Run in minutes a survey that traditionally takes days to weeks — with accuracy on par with real surveys.
Remarkably low noise
Without the human “mood of the day” or “immediate prior behavior” fluctuations, even a 50-person sample grasps tendencies close to a large-scale survey — as demonstrated in the Case 2 episode.
Advanced simulation via context adaptability
As shown in Case 3, by setting the social environment or era, you can accurately reproduce consumer psychology under a specific context, such as “job values during the pandemic.”
Overwhelming advantages over traditional surveys
⏱️ Speed
Traditional: days to weeks
PersonaChoice: done in minutes
💰 Cost reduction
No recruitment fees, incentives, or operations costs. Dramatic cost reduction
🎯 Flexibility
Retry with different conditions any number of times. Accelerate the hypothesis-testing PDCA cycle
📊 Accuracy
Proven 80–90% agreement against real research
Where it shines in marketing
✓ Pre-testing ad creative
Quickly identify the most responsive appeal from multiple variations
✓ Preference analysis by target segment
Accurately simulate selection tendencies by generation, gender, and life stage
✓ Concept testing for new products
Predict consumer reaction before market launch and minimize risk
✓ Fast hypothesis testing
Instantly test an idea you just came up with and accelerate decisions
PersonaChoice is an
innovative decision-support tool
that changes the future of marketing
Proven accuracy, overwhelming speed, flexible usability.
With PersonaChoice, which combines all of these,
take your marketing strategy to the next level
Why not try it for free first?
Experience new marketing possibilities right now with 3,000 bonus points (free) for first-time sign-ups.