Pervasive Insights™ · Synthetic customers
Build a panel from your own segments, then field a survey, run a moderated group, or interview a single panelist in depth. Every result carries its calibration state, and the whole panel is graded on a report card you can open any time.
Pervasive Insights is a synthetic research panel built on a client's own human research: a searchable library of every study the client has run, and synthetic respondents calibrated to how that client's real customers answered, with the error reported topic by topic. It is a City Research Solutions product. City Research has run consumer research since 1979.
Mode
Field a survey to 50 to 500 calibrated synthetic panelists, sized to each client. See results by total sample or by segment.
Mode
Run moderated focus group discussions with synthetic panelists from any segment. Group count and participants per group are set for each client; delivered as a transcript and a report.
Mode
Interview a single panelist in depth about their category behavior. Chat with a panelist in their own voice.
Panel builder
Every panel is drawn from calibrated synthetic panelists built on your fielded research: the segments your surveys found, the language your focus groups used, the trade-offs your trackers measured. A Market Representative Panel mirrors your buyer population; a Custom Panel lets you set the segment mix, gender mix and age group, and add recruiting criteria in plain language.
Where you have tested, you get your exact human results. Where you have not, you get a calibrated read that respects everything you have tested, with realistic person-to-person variation rather than a flat average. When something is genuinely new, the panel flags it low-confidence and points to a real test.
Attach stimulus images to any question. Results download as CSV or PDF. Plain-language recruiting criteria narrow any panel further.
Sample presets
Panel size
Segment mix
Gender mix
Age group
Segment names are illustrative. A live portal shows the client's own segments.
What a survey result looks like
And your own human data is layered in: the range your real customers have given on the same type of question, or, where you have asked the same question before, the human result beside the synthetic one, cell by cell.
"Coffee maker replacement triggers" · sample
100 respondents · 2 questions · Complete
Respondent distribution
● Core Enthusiast: 24 ● Practical Owner: 22 ● Premium Seeker: 18 ● Value Shopper: 20 ● Newcomer: 16
Male: 52 Female: 48
Questions
Q1: Which of these would most make you replace your current coffee maker this year? (single-select)
100 responses · click to drill downCalibration state: Measured · topic area: Purchase Intent · a human benchmark for this topic area is attached; results are read against it.
Evidence: same question type the range your real customers gave on this type of question, asked as a single-select purchase trigger, across 214 previous studies and 31,600 respondents
What is being measured: the leading answer's share of respondents
Your human data · 214 measurements
This survey
Based on 214 human measurements · evidence strong. A result outside the full human range would be held for review rather than reported.
Q2: If it were priced at $X, how likely would you be to purchase? (Gabor-Granger price ladder)
100 responses · click to drill downCalibration state: Inferred · topic area: Price Sensitivity · inferred from a related topic area, not measured for this question. No human pricing benchmark is attached to this question; the prices below come from the synthetic panel alone.
$49
78% likely
$69
61% likely
$89
38% likely
$109
19% likely
Generic small-appliance category. Segment names, figures and findings are illustrative, produced under sample conditions, not from a client portal.
"Please RANK your TOP 5 finishes for the new carafe, from MOST to LEAST favorite" · sample
100 respondents · 1 question · Complete · stimulus images attached
Closest comparable human question
Please RANK your TOP 5 finishes for the carafe, from your MOST favorite to LEAST favorite.
Product / Finish Preference · Carafe Concept Test, 2024 · n=812 · q17/q18 · single_select_pct
Closest comparable human question (same scale family), shown as a directional reference, not a calibrated benchmark for this exact stimulus. Human scale points mapped to the survey scale by sentiment.
| Item | #1 | #2 | #3 | % in Top 3 |
|---|---|---|---|---|
| Matte black | 17 (17%) H 16.3% Δ+0.7 | 14 (14%) H 10.1% Δ+3.9 | 10 (10%) H 10.3% Δ-0.3 | 41% H 39.7% Δ+1.3 |
| Brushed steel | 14 (14%) H 16.4% Δ-2.4 | 11 (11%) H 14.6% Δ-3.6 | 13 (13%) H 12.7% Δ+0.3 | 38% H 46.7% Δ-8.7 |
| Cream | 15 (15%) H 15.5% Δ-0.5 | 13 (13%) H 9.8% Δ+3.2 | 19 (19%) H 11.9% Δ+7.1 | 47% H 40.2% Δ+6.8 |
| Forest green | 11 (11%) H 5.8% Δ+5.2 | 12 (12%) H 10.9% Δ+1.1 | 11 (11%) H 9.4% Δ+1.6 | 34% H 29.1% Δ+4.9 |
| Navy | 14 (14%) H 9.2% Δ+4.8 | 5 (5%) H 10.1% Δ-5.1 | 8 (8%) H 8.3% Δ-0.3 | 27% H 30.6% Δ-3.6 |
| Terracotta | 9 (9%) H 5.7% Δ+3.3 | 9 (9%) H 9.2% Δ-0.2 | 5 (5%) H 10.5% Δ-5.5 | 23% H 28.4% Δ-5.4 |
Each cell: the synthetic count and share on top; underneath, H is the human result from your own study and Δ is the gap in points. Green within 5 points, amber beyond. Where no human question exists, the row shows the synthetic result alone and the calibration state says so. Two ways the human layer appears: when you have asked the same type of question many times, the portal shows the range your real customers gave and places the synthetic result on it (first sample above); when you have asked the same question before, the comparison is cell by cell (this sample). The more you have asked real people over the years, the more of your synthetic results come back with your own human data beside them.
Generic small-appliance category. Item names, figures and the human reference are illustrative, produced under sample conditions, not from a client portal.
Focus groups
Name the study, set the number of groups and panelists per group, choose a discussion mode (conversational or round-robin), write the discussion guide or load a saved sampling plan, and run. The output has two tabs, Transcript and Report, with PDF and CSV export.
Coffee maker replacement (sample) · Groups: 3 · Mode: round-robin
Group 1 (3 participants)
MODERATOR: Welcome everyone. I'm going to go around the room on each question so everyone gets a chance to share. Dana, your turn.
Dana T. [Pragmatic] The carafe. I was making a pot for one person and pouring half of it out. The new one does a cup and I stopped feeling guilty.
Marcus W. [Confident] Opposite for me. If it can't make ten cups when my in-laws show up, it's a toy.
Priya N. [Considered] I don't care about either until it goes on sale. Then I care whether it fits under the cabinet.
MODERATOR: Marcus, how much does capacity matter versus the price?
Marcus W. [Confident] If it holds ten cups I'll pay forty more. If it holds six I won't pay anything.
Report · Topline
Of 9 participants, 5 named single-cup flexibility as the replacement trigger, 3 named guest capacity, 1 named counter footprint. Price surfaced as a gate rather than a driver: nobody replaces until a sale, then fit and capacity decide.
Overall sentiment Pragmatic 33% · Considered 22% · Confident 22% · Warm 11% · Other 12%
One-on-one
Pick a panelist from any segment and interview them in depth about their category behavior. Show a stimulus in the conversation and follow up as many times as you like.
Renee K.
44 · two kids · suburban Midwest · Practical Owner · heavy user
I want to show you a coffee maker we're considering. What's your honest first reaction?
Stimulus shown in-conversation
RENEE K. First thing I look at is whether the carafe is glass or thermal, because the glass ones burn the last cup by nine o'clock. I like that the water tank comes off the side; the one I have you fill from the top and it drips on the counter every time. What I can't tell from a photo is how loud it is, and that matters more than you'd think when a kid is asleep down the hall.
What would make you buy this over the one you have now?
RENEE K. Thermal carafe and a tank I can fill at the sink. I'd pay twenty more for those two things. The phone app I'd never use.
Names, segments and findings are illustrative, produced under sample conditions, not from a client portal.
Under the hood · Calibration Report Card
Every client's portal has a report card: how closely the synthetic panel's answer matches real customer data for each of 19 marketing topic areas, for each segment, measured in percentage points. It updates as new waves of real customer research come in.
Pass rate: 72 / 76 cells ≤ 5pp
Rows fully passing: all 5 segment gaps ≤ 5pp
How to read this page
Each cell is the gap between the synthetic panel and real customer data for that segment, in percentage points. Green is within ±5pp; amber is wider and flagged; rows that are not yet calibrated say so. Calibration is ongoing, and the card updates as new waves of real customer research come in.
Across all clients, on calibration fit: about 3.5pp average error, 92% of 95 cells within 5pp. On blind holdout, which is the prediction number, the median miss is about 6.8 points with 42% of cells within 5. For scale: when a real survey is split randomly in half, the two halves agree within 5 points on about 85% of answers (58 consumer studies, 41,340 answer pairs). No synthetic method can beat a human study against itself, and Pervasive Insights reports both numbers on the same report card. Explore the Constellation view →
| Category | Core Enthusiast | Practical Owner | Premium Seeker | Value Shopper | Newcomer |
|---|---|---|---|---|---|
| 1Functional Perception | 2.2pp | 1.5pp | 1.4pp | 1.8pp | 2.3pp |
| 2Emotional Perception | 3.0pp | 2.0pp | 2.3pp | 1.4pp | 1.5pp |
| 3Value Perception | 2.1pp | 3.2pp | 2.2pp | 1.9pp | 2.4pp |
| 4Brand Preference | 4.2pp | 2.2pp | 2.0pp | 3.4pp | 2.9pp |
| 5Purchase Intent | 1.4pp | 5.8pp | 1.9pp | 3.7pp | 4.3pp |
| 6Repurchase Intent | 2.3pp | 1.6pp | 2.7pp | 2.4pp | 1.4pp |
| 7Consideration Set | 1.4pp | 1.4pp | 1.4pp | 1.4pp | 1.4pp |
| … and 12 more marketing topic areas, including Usage Frequency, Price Sensitivity, Feature Importance, Channel Preference, Awareness & Ad Recall and Ad / Concept Response. Rows not yet calibrated say so. | |||||
Partial, illustrative view with generic segment names. Every client sees the full card for all 19 marketing topic areas and their own segments.
Under the hood · Constellation
The Constellation is the map behind the report card. Each point is one of the 19 marketing topic areas the portal calibrates. The closer a point sits to the center, the more of your own human data anchors it; the color says how far calibration has gone.
Every new study you field with us moves points inward. The live view opens from the portal sidebar under Under the hood.
Your research
history
Illustrative constellation, produced under sample conditions. A live portal shows the client's own topic areas and states.
What a week with it looks like
Forty concepts, budget to field eight. The panel screens all forty on Monday, kills the bottom of the pile with the evidence attached, and the eight fielding slots go to survivors instead of favorites. Real respondents crown the winner.
Twelve ideas for a seasonal line. Drop them all, get the segment splits and the obvious kills by lunch, iterate the survivors ten times if you like, then field the best three. The fielded winner goes back into the research stack for next season.
The tracker fields in March; the question arrives in July. Ask the panel before the meeting ends. Where the report card shows a segment cell in amber, a five-question pulse fields exactly the flagged items instead of the whole wave.
Everything in the suite is built on studies fielded by City Research Solutions. Explore the research services behind it.
A 30-minute walkthrough of a working portal on generic sample data. Bring the question your team argued about last week.
Request a walkthrough