Scoring

For example, a person takes a purchase-readiness test. One person answered "yes" to all the key questions, another - to none of them

Without scoring you get two equally "completed" responses. With scoring - two different numbers that automatically determine who is a hot lead and who needs nurturing. Scoring is the conversion of qualitative answers into a quantitative total suitable for segmentation and automation.

Definition

Scoring - a mechanism for assigning numeric values to a respondent's answers and then summing them into a final score. Each answer option is given a weight - positive, negative or zero. The final score is used to classify the respondent, show a personalized result, segment the audience or pass the value to a CRM as a lead characteristic.

Where scoring is used

Scoring works in any scenario where you need not just to collect answers, but to automatically classify the respondent by the result.

Lead qualification. A marketing quiz "Let's find a solution for your business" - every answer about company size, budget and current problems gets a weight. The final score determines the segment: hot lead (hand off to sales today), warm (add to nurturing), cold (newsletter only). Without scoring the quiz just collects data. With scoring it qualifies the audience automatically.

Educational testing. A test with correct and incorrect answers. Each correct answer is +1 point (or a different weight for questions of different difficulty). The result is a completion percentage, access to the next level, a certificate. A classic of e-learning and corporate training.

Diagnostic surveys. "How effective is your project management system?" - a series of questions, each assessing a separate aspect. The final score shows which quadrant the company is in and recommends specific actions. Personalizing the result increases engagement and conversion into the next step.

HR assessment and screening. Tests of alignment with company values, basic professional knowledge during hiring, soft-skills assessment. Scoring automatically ranks candidates before manual review.

NPS and composite metrics. Quiz formats for calculating indexes, where the final score is not just a sum but a formula: the share of promoters minus the share of detractors.

How scoring works technically

The basic mechanic is simple: each answer option is assigned a number. The user goes through the questionnaire - the system sums up the points for their answers. The total is compared with threshold values that determine the result or action.

A few key elements:

Question weights. Not all questions are equal. A question about budget may weigh 3 times more than a question about timelines. This is implemented by multiplying the answer's points by the question's importance coefficient. The result: a "yes" answer to a priority question gives 15 points, to a secondary one - 5.

Negative points. For a "wrong" answer or a sign of low fit you can subtract points rather than simply not adding them. This lets you build more differentiated models: a candidate with one critical mismatch will get a low total even if everything else is fine.

Threshold values. Ranges that translate the final score into a category: 0-30 - low, 31-70 - medium, 71-100 - high. The number of thresholds and their boundaries is a decision for the specific task. For binary classification (pass/fail) a single threshold is enough.

Personalized result. Each score range corresponds to its own screen with text, a recommendation, an image or a link. A person with a high score sees "You're ready to implement - contact us". A person with a low one - "Start with these materials".

Example: scoring in a quiz for choosing a plan

A SaaS company launches a quiz "Which plan suits you?" - 6 questions. Each answer has a weight:

• Team size: up to 5 people = 0 points, 5-20 = 10, more than 20 = 20
• Number of surveys per month: up to 5 = 0, 5-20 = 10, more than 20 = 20
• Whether CRM integration is needed: no = 0, yes = 15
• Whether a white label is needed: no = 0, yes = 20
• Budget: up to $30/mo = 0, $30-150 = 10, higher = 20
• Urgency: "just browsing options" = 0, "ready to try now" = 15

Maximum - 100 points. Thresholds: 0-30 → recommend the free plan, 31-60 → starter plan, 61-100 → a call with a manager. 400 people took the quiz: 120 landed in "manager", conversion into a demo call - 34%. Without scoring all 400 would have received the same offer - conversion would have been significantly lower.

Scoring vs simple summation

Basic counting of correct answers is a special case of scoring where all weights equal 1. Full-fledged scoring differs in that:

• Questions have different weights depending on importance
• One answer can simultaneously add points on one scale and subtract on another (multidimensional scoring)
• The result can be not a single number but several - across different dimensions (for example, technical skills vs communication skills)

Multidimensional scoring is used in psychological tests and profiling surveys: a person gets a profile of several indicators rather than a single score. This is harder to set up but gives a richer picture.

Common mistakes when setting up scoring

Arbitrary weights without justification. Assigning weights "by eye" without analyzing the real contribution of each question to the total is a frequent mistake. The result: one question dominates all the others, and the final score is effectively determined by it alone. Before launch it's worth checking: if you change the answer only to the most heavily weighted question - how much does the total change? If it overrides all the rest - the weights need to be reconsidered.

Too many threshold categories. Five or six segments look detailed, but blur the meaning. If the difference between segments 3 and 4 does not imply different actions - they don't need to be separated. The optimal number of categories is as many as the actually different actions you plan to take based on the result.

Not testing the scoring model before launch. Running the extreme scenarios manually (all answers minimal, all maximal, mixed) is a mandatory step. Sometimes it turns out that with realistic answers everyone lands in one segment - and there is no differentiation at all.

Ignoring skipped answers. If a question is optional and the respondent skipped it - its score is usually 0. This needs to be accounted for in the threshold values: a person who skipped 2 questions automatically loses the maximum possible score not because of "bad" answers, but because of skipping.

Scoring in SurveyNinja

In SurveyNinja scoring is configured natively through correct-answer and score settings: each answer option is assigned a numeric value, and the system automatically sums the total. The result is available in a variable that can be passed via piping into the final screen - to show the person their score or personalized text.

For segmentation by result, logic jumps based on the accumulated score are used: when a threshold is reached the respondent lands on the corresponding result screen. Score data is exported together with the responses via export - for subsequent analytics or transfer to a CRM via webhook.

Scoring turns a set of answers into a classifiable result. The key decisions: question weights, threshold values, the number of segments and a personalized response for each range. A well-tuned scoring model automates what would otherwise require manually reviewing every questionnaire.

Frequently asked questions

How does scoring differ from a regular test with correct answers?

A test with correct answers is a special case of scoring with equal weights: +1 for a correct answer, 0 for an incorrect one. Full-fledged scoring lets you assign different weights to different questions and options, use negative points and create several scales at once. This gives flexibility for tasks where there are no unambiguously "correct" answers - for example, in qualification quizzes or diagnostic surveys.

How do you choose weights for questions?

Start from the business logic: how much does each factor really affect the final classification? If budget is twice as important as urgency - the weight of the budget question should be twice as high. After assigning weights, be sure to run several test scenarios: the highest, the lowest and a typical answer. If the distribution across segments is unexpected - revisit the weights.

Can you show the respondent their final score?

Yes. The final score can be passed via piping into the text of the final screen: "Your result: 74 out of 100". This increases engagement - people like to learn their "number". Whether to show the score or only the category ("high level") depends on the task: in lead qualification the score is usually not shown, in educational tests - it is.

How do you pass the final score to a CRM?

Via webhook or API: when the survey is completed the system sends the data, including the final score, to the specified endpoint. In the CRM the score is saved as a lead or contact field. This makes it possible to automatically prioritize handling: those with a high scoring value are worked first.

What to do if most respondents land in one segment?

This is a signal that the threshold values or weights are set incorrectly. Check the distribution of final scores across all responses: if 80% are in the 60-80 range - you need to shift the thresholds or revisit the weights to get a more even distribution across segments. An alternative is to reduce the number of segments to the number that is actually distinguishable in the data.