Score a survey: weights, scales and AI analysis
How Polzy turns answers into a score: the Scoring tab, each question's weight, reverse scoring and point values, and the AI reading on top — with a worked example.
On this page
Every response to a survey gets a score from 0 to 100, worked out from its answers. You decide which questions count, what each answer is worth and how they add up — on the builder's Scoring tab for the whole survey, and on each question's own Scoring tab. On top of the numbers, Polzy's AI can read every response for themes, sentiment and a summary.
Numeric scoring is on every plan. AI insights per response and What the AI looks for need AI scoring, on Solo and up; Themes for AI analysis also needs Themes (Solo). Only Admins change a survey's scoring.
How a score is worked out
- Each answer becomes a figure from 0 to 100. A rating and the scales built on it, an NPS, a slider, a matrix, a number between its ends, a choice or a ranking with point values and a consent box can have one. Written text, emails, phone numbers, websites, dates, files, addresses and contact details never do. A Yes or no has one of its own: Yes 100, No 0.
- The survey combines them. Average, weighted average or total — your choice. Questions that were skipped, or have no figure, are left out rather than counted as 0.
- The AI reads on top. With AI analysis on, Polzy writes a summary, finds themes and sentiment and scores each theme. It never gives a written answer a number of its own.
A response with no figure at all gets no score rather than 0: "—" in the Responses list, and Not scored in its panel. Results, analytics and the preview show scores out of 10 — a 74 reads 7.4; a single response's panel and the Responses list show it out of 100.
The survey's Scoring tab
Open the survey in the builder and choose Scoring in the strip above the outline. The tab opens with How scoring works, a reminder of the layers, then:
| Setting | What it does |
|---|---|
| Score from the preview | Answer questions in the live preview beside the builder, and this card shows the score those answers would get, and each question's figure ("—" where an answer has none). Nothing is saved. |
| Numeric scoring | On to start with. Turn it off for a survey you only want to read: responses then get no score, and AI analysis can still run. |
| How to combine question scores | Average (to start with): the mean of every scored answer — comparable across surveys of different lengths. Weighted average: the mean, with each answer counted as many times as its question's weight. Total: every scored answer added up, capped at the maximum. |
| Maximum score | Only with Total: the ceiling for a total, 100 to start with. |
| AI insights per response | A summary, the themes raised and the sentiment, as each response arrives. Uses AI credits. |
| Themes for AI analysis | Shown once AI insights are on: choose which of the workspace's themes this survey reports on. The chosen themes also become the Theme chips on each question. |
Weights only change a Weighted average. An Average or a Total counts every answer once.
A question's Scoring tab
Open a question and choose its Scoring tab. Headers, images and statements have none. On a question that can never have a figure — a hidden field, a meeting, a date or a time, a file, an address, contact details, an email, a phone number or a website — Include in scoring is off and greyed out, and the line under it says why, such as "A date or a time has no score".
| Setting | What it does |
|---|---|
| Include in scoring | On for every new question: "Counts towards the survey score and its theme". Off, the rest of the tab is hidden. Greyed out, with the reason, where the question can never be scored. |
| Weight | How much this question counts compared with the others, from 0.1 to 5 in steps of 0.1; 1 to start with. Used by Weighted average. A weight other than 1 shows on the closed question, such as "Weight 2". |
| Reverse scoring | For questions where a low answer is the good one — "How often do you feel rushed?". The figure becomes 100 minus itself. |
Then the settings for the question's type:
| Question type | Without changes | What you can set |
|---|---|---|
| Rating, Opinion scale, CSAT, CES | The point as a share of the top of the scale: 4 of 5 is 80, 7 of 10 is 70. | Lowest score and Highest score: the scale is spread between them, and the figure is the result as a share of the highest. With 0 and 10 on a 5-point scale: 1 = 0, 2 = 25, 3 = 50, 4 = 75, 5 = 100. |
| NPS | The answer out of 10: 9 is 90, 6 is 60. | What each band is worth: Detractors (0–6), Passives (7–8) and Promoters (9–10), 0, 5 and 10 to start with. The figure is the band's value as a share of the highest band: 0, 50 and 100. |
| Single select, Dropdown, Multi select, Picture selection | Not scored until you give the options values. | What each option is worth, from 0 to 100. An option you leave at 0 counts 0. With several options chosen, the figure is their average. The respondents' own "Other" has a line of its own. |
| Yes or no | Yes 100, No 0. | Nothing to set: Reverse scoring makes No the 100. The tab says "Yes scores 100 and No 0. Reverse scoring to score No higher." |
| Ranking | Not scored until you give the options values. | What each option is worth, How the order is scored and What each position is worth. The figure is the options' values, each counted by where it was placed: with Position based, the first of five places is worth 5 until you change it; with Weighted scoring, each place is worth one over its rank. |
| Consent | Not scored. | What each answer is worth: Ticked 100 and Not ticked 0 to start with. The question is scored once you change any setting on its tab. |
| Short text, Long text | No figure. | What the AI looks for: what the AI reading should look for in answers to this question. It guides the reading; a written answer is not given a score. While Include in scoring is on, the question shows "AI scored" in the list. The mark shows only where the workspace has AI scoring; a number, a website or a masked field never wears it, since the AI does not read them. |
| Number | Between its Lowest number (0) and Highest number (100): 7 between 0 and 10 is 70. Not scored without both. | Nothing more: set the two ends on Content. |
| Slider | Its place between the ends: the lowest 0, the highest 100. | Nothing more: set the ends on Content. |
| Matrix | Each statement by its answer's place: the first answer 0, the last 100, the others evenly between; the figure is the average of the statements answered. | What each answer is worth, from 0 to 100, starting at those places. With several answers per statement, a statement counts as their average. |
| Email, Phone, Website, Date, Time, Date & time, File upload, Address, Contact info | No figure. | Nothing: Include in scoring is off and greyed out, with the reason. |
For a Rating or an NPS, the steppers show their starting values (0 and 10; 0, 5 and 10) before anything is saved, but the question is scored the plain way — share of the scale — until you change one of them.
A worked example
A five-question survey. Four answers have a figure; the text answer has none:
| Question | Its settings | Answer | Figure |
|---|---|---|---|
| "How clear are your goals?" — Rating, 5 points | Weight 1 | 4 | 4 ÷ 5 = 80 |
| "How likely are you to recommend us?" — NPS | Weight 2 | 9 | 9 ÷ 10 = 90 |
| "How often do you get feedback?" — Single select | Weekly 100, Monthly 50, Rarely 0; weight 1 | Monthly | 50 |
| "How often do you feel rushed?" — Rating, 5 points | Reverse scoring; weight 1 | 2 | 2 ÷ 5 = 40, reversed: 100 − 40 = 60 |
| "What would make your week easier?" — Text | What the AI looks for | "Fewer meetings" | none |
- Average: (80 + 90 + 50 + 60) ÷ 4 = 70 — 7.0 out of 10.
- Weighted average: (80 × 1 + 90 × 2 + 50 × 1 + 60 × 1) ÷ (1 + 2 + 1 + 1) = 370 ÷ 5 = 74 — 7.4 out of 10.
- Total: 80 + 90 + 50 + 60 = 280, capped at the Maximum score of 100 = 100. With a maximum of 400, it is 280.
Scores are rounded to whole numbers. Answer the same way in the preview and Score from the preview shows these figures.
Themes and the AI reading
A theme's score comes from the questions filed under it with the Theme chips:
- Questions with a figure — a rating, an NPS, a slider, a matrix, a choice with values — are averaged into the theme straight away, with no AI. A theme measured this way is scored even when AI is off or out of credits.
- Written answers filed under a theme are read by the AI, which scores the theme from what people wrote. This needs AI insights per response.
- Themes on the survey but on no question are candidates: the AI looks for them in the answers and scores one only when it finds evidence.
An answer to an AI follow-up counts for the theme of the question it followed up. A paused theme keeps its scores and takes nothing new. More in themes and AI insights.
When the organization runs out of AI credits, responses are still collected and scored by the numbers; their AI reading waits until credits come back, and Organization settings → Usage says how many are waiting.
Good to know
- Changes apply to new responses. A response is scored when it arrives; editing a weight later does not change scores already given.
- Turning off Include in scoring. The question counts towards neither the score nor its theme — in the preview and in every new response. Until November 2026 a Rating or an NPS still counted at its plain value with it off.
- Theme chips need themes. A question can only be filed under a theme the survey has, chosen under Themes for AI analysis.
Troubleshooting
- A response shows "—" or Not scored. No answer had a figure: only written answers, or choice questions whose options have no values. Give the options values, or add a rating.
- A choice question shows "—" in the preview. None of its options has a value yet.
- The total is always 100. You chose Total with the Maximum score at 100; raise the maximum.
- AI insights per response is greyed out, with an upgrade mark. The plan does not include AI scoring (Solo and up), or the workspace has switched it off.
- Create themes first to use AI analysis. The workspace has no themes yet. Add them under Surveys → Themes.
Related articles
Signed-in members can ask Polzy or message the team from Support. How fast we reply depends on your plan.