The "what did you like" and "what did you dislike" questions decide whether a concept moves forward. Survey Coder Pro codes them with attribute nets so you can compare across concepts, cells and waves.
Concept test timelines are short, and the comparison across cells is what the client wants to see.
A monadic test with several concepts needs the same codebook in every cell, or the comparison doesn't hold.
Responses mix taste, price, pack and brand in one sentence, and they need to be separated without losing any of them.
The client wants the likes and dislikes table by concept a few days after fieldwork closes.
Brands and products have proper names and regional nicknames that an outside coder doesn't always recognize.
The AI reads all the responses, proposes a codebook with attribute nets (taste, texture, price, pack) and codes every response with all the codes that apply. Your team adjusts the codebook and reviews only the doubtful answers.
One codebook for every cell in a test, so concepts are compared by the same rule.
Attribute nets, with positive and negative codes kept apart for likes and dislikes.
With the .sav, a tabulation book with concepts in the banner and significance testing between cells.
Codes come out as columns per respondent in the .sav or Excel file you deliver.
Applications for consumer goods brands
Analyze purchase behavior, point-of-sale decisions and choice drivers.
Process hundreds of mystery shopper reports identifying service patterns.
Rapid feedback from concept tests, packaging and positioning.
Analysis of reviews, cart abandonment and online shopping experience.
Member feedback on benefits, redemption and satisfaction.
In-store experience, display, availability and service.
Concept test for a ready-to-drink oat beverage. After exposure, respondents are asked what they disliked; here is part of the coded dataset.
Open-ended question
What did you like least about this product?
| Response | Codes |
|---|---|
| “Way too sweet for an oat drink, I expected something more natural.” | Too sweetDoesn't seem natural |
| “The price. For that I'd rather buy my usual bag of oats.” | High price |
| “Nice-looking bottle, but it doesn't say how much sugar is in it.” | Missing nutrition infoAttractive pack (positive) |
| “It felt kind of gritty at the end, didn't love that texture.” | Gritty texture |
| “Nothing really, I liked everything.” | Nothing / liked everything |
| “It's a brand I've never heard of, not sure I'd trust it.” | Unknown brand |
| “Tiny for what it costs, it's gone in two sips.” | High priceSmall size |
| Net / code | % of respondents who answered |
|---|---|
| NET Taste and texture | 38% |
| Too sweet | 24% |
| Gritty texture | 12% |
| Doesn't seem natural | 7% |
| NET Price and size | 31% |
| High price | 26% |
| Small size | 9% |
| NET Pack and information | 15% |
| Missing nutrition info | 11% |
| Attractive pack (positive) | 5% |
| NET Brand | 8% |
| Unknown brand | 8% |
| NET No dislikes | 18% |
| Nothing / liked everything | 18% |
Multi-code question: one response can carry several codes, so percentages add up to more than 100%. A net counts each respondent once, even if they have several codes from that group.
Example built for this page: this is not client data and the percentages are illustrative.
They usually sit in two questions with their own codebooks, or in one codebook with positive and negative codes kept apart. Both work; what matters is keeping the same approach in every cell and wave.
All cells are coded with a single codebook. With the .sav, the tabulation book puts the concepts in the banner and marks significant differences.
The AI proposes codes from what respondents write, brands included. You can adjust the codebook to group variants of the same name.
Yes, as long as the information is open text: purchase reasons, brand switching reasons or comments from visit reports are coded the same way as a survey question.
Upload a dataset with an open-ended question and see the codebook the AI proposes. The free trial includes 250 responses.