What should NPS survey software do?
NPS survey software collects recommendation ratings, calculates Net Promoter Score and helps teams interpret and act on feedback. A useful buying evaluation covers delivery, sampling, comment analysis, customer context, follow-up and governance—not only the score chart.
The right scope depends on your workflow. A team running occasional anonymous research needs different controls from a customer success team following named accounts across onboarding and renewal. Decide what action the program should support before choosing how much customer history to connect.
For a growing organization, the practical question is whether its team can maintain the collection and analysis as products, segments and reporting needs change. Use the requirements and pilot below to compare that ongoing effort.
Check the score before interpreting it
NPS is the percentage of promoters minus the percentage of detractors. Ratings of 9–10 are promoters, 7–8 passives and 0–6 detractors. Passives stay in the valid-response denominator even though their percentage is not subtracted. See Bain’s NPS calculation guidance.
In a fictional wave with 100 valid responses, 50 promoters, 30 passives and 20 detractors produce an NPS of 30. Report the score with the valid-response count and collection period. NPS is usually presented as a score, not “30% satisfied.” Recommendation and satisfaction are different questions.
Test how the software handles blank ratings, invalid values, duplicate submissions and exclusions. A missing comment does not make a valid rating disappear. Conversely, a comment without a valid rating should not be silently assigned to a score group.
NPS software requirements and demo tests
Scroll horizontally to see all columns →
| Area | Requirement | Demo test |
|---|---|---|
| Collection | Appropriate delivery, timing and contact controls. | Show how a customer avoids receiving overlapping requests. |
| Calculation | Inspectable counts, exclusions and denominator. | Reproduce the score from a small known dataset. |
| Comment analysis | Useful themes with source responses available for review. | Review mixed, contradictory and blank comments. |
| Segmentation | Relevant product, account, role or journey context. | Apply a segment and verify its membership and counts. |
| History | Clear wave and identity handling where linking is appropriate. | Separate matched respondents from the full wave. |
| Follow-up | An owner, status and appropriate response process. | Take one issue from review to a recorded action. |
| Governance | Defined access and controlled changes. | Restrict a record and change a theme definition. |
| Portability | Usable exports of the information you need. | Export ratings, comments and relevant context together. |
Confirm what is included in the proposed plan, what requires configuration and what depends on another tool. A feature label such as “AI insights” does not explain how results are checked, corrected or maintained.
Keep collection focused on the decision
A recommendation question followed by an open question about the reason is a useful starting point. For example: “What is the main reason for your rating?” Do not imply that exactly two questions are always sufficient. A short additional question may be necessary to understand the relevant product, interaction or permission to follow up.
Use information you already hold when it is appropriate and reliable, rather than asking the customer to enter it again. Check the source and freshness of that context. An outdated account segment can make a carefully written survey difficult to interpret.
Separate a relationship survey from feedback collected immediately after a specific interaction. Changes in timing, channel or audience can affect the responses. Record those conditions and avoid treating every recommendation rating as directly comparable.
For wording and survey design, continue to NPS survey questions. Keep this software evaluation focused on how the selected design will be operated and reviewed.
Evaluate comment analysis beyond a word cloud
A word cloud can show frequently used words, but it does not reliably distinguish “support solved my problem” from “support never replied.” Test whether the analysis captures topics, sentiment and relevant context, and whether a reviewer can open the underlying response.
Do not assume competing products stop at word clouds. Text and theme analysis are available in the market. The useful distinction is how well the proposed setup handles your comments, how your team governs the definitions and how the findings support action.
Include comments with several topics, unfamiliar terminology, mixed sentiment and little detail. A response can praise a product while criticizing onboarding. Check whether the analysis preserves that distinction instead of forcing the whole comment into one positive or negative label.
Review a sample against a clearly defined coding guide. Record corrections and how they affect later analysis. If the software ranks themes, ask what the rank represents: response frequency, number of customers, severity or another rule. A frequently mentioned topic is not automatically a statistically established driver of the score.
Worked example: score coverage and comment coverage differ
Suppose 500 customers are invited and 100 provide valid ratings: 50 promoters, 30 passives and 20 detractors. The response rate is 20% if the invitation count is the defined denominator. The score is 30. That score describes the respondents; it does not establish what every nonrespondent thinks.
Sixty respondents leave usable comments: 25 promoters, 20 passives and 15 detractors. Eight of the detractor comments mention onboarding. That is 8/15, or about 53.3%, of detractor comments, and 8/20, or 40%, of all detractor respondents. Neither is “53.3% of customers have an onboarding problem.”
Show the denominator for each theme. If comments can have several themes, their percentages may add to more than 100%. Keep the rating analysis and comment coverage connected without pretending the two datasets have identical coverage.
Compare waves and customer relationships carefully
Two quarterly scores may come from different respondents. A change from 30 to 35 is a five-point movement in the wave-level score, not proof that individual customers became more positive. Review changes in the audience, response coverage and collection conditions.
Where named follow-up is appropriate, a stable respondent reference can support a matched view. Keep that analysis separate from the full-wave results and show how many customers appear in both. Matching alone does not establish that a follow-up action caused the change.
For B2B programs, distinguish people from accounts. Several contacts at one account may have different roles and experiences. Decide whether the primary result is response-weighted or uses a defined account-level method; do not silently let accounts with more respondents carry more influence.
Anonymous feedback can still support useful group-level trends. Do not promise anonymity while building a hidden personal history. Choose the collection arrangement around the purpose and explain it accurately to respondents.
Make routine changes without losing comparability
Define the shared fields needed for analysis: rating, date, survey purpose and the few relevant segments. Teams serving different products or locations can retain local follow-up questions. A shared data dictionary should explain which fields are comparable and how each segment is assigned.
Record changes to wording, segment rules, theme definitions and exclusions. A higher theme count after expanding its definition may reflect the classification change rather than a worsening customer experience. Preserve enough history to explain the difference.
Test permissions on responses, account context and supporting notes. A report for a broad audience may need summarized findings, while an authorized follow-up owner needs a specific contact. AI-generated summaries should follow the access boundaries of the sources they use.
Turn a finding into a responsible follow-up
Choose an owner and a practical action for each issue that warrants attention. A named customer complaint may need an individual response; a recurring product issue may need a team investigation. The process should distinguish acknowledgement, investigation and resolution.
Do not use a low score as an automatic churn prediction or the sole basis for escalating an account. Read the context. A short negative comment may describe an isolated incident, while a promoter may still report a serious problem. For score-group interpretation, see NPS detractors.
Record what changed and how the team will check whether it helped. Later feedback and operational evidence can support that review. Avoid reporting that the issue was fixed merely because a follow-up message was sent.
Compare total cost, including the work your team does
Request pricing for the intended audience, response volume, channels, users, integrations and analysis functions. Include setup, historical imports, staff training and ongoing administration. Verify whether routine changes can be made by the feedback team or require specialist support.
A lower subscription can still require substantial manual review and reconciliation. A broader platform can also add implementation work your team does not need. Compare the same complete cycle: prepare the audience, collect responses, review the analysis, follow up and produce the next report.
Test exports and continuity before purchase. You should understand how the team will retain usable ratings, comments and context if a contract or workflow changes.
A practical NPS software pilot
- Use an authorized or de-identified wave containing all three score groups and missing comments.
- Calculate the expected NPS and theme denominators independently.
- Import the relevant context and resolve a deliberate duplicate or mismatch.
- Review theme coding on mixed and contradictory comments.
- Compare a second wave, separating matched and unmatched respondents.
- Take one issue through ownership, follow-up and review.
- Have your team update a definition and prepare the next cycle.
Document accuracy, unresolved requirements and staff effort. A successful pilot should leave the team able to explain both the score and its limitations, not just show a persuasive dashboard.
Where Sopact fits in the decision
Sopact’s focus is self-managed collection, analysis and governance for recurring evidence. For an NPS program, the useful test is whether your team can keep ratings, explanations and appropriate customer context connected, review findings and maintain the next cycle with clear definitions and ownership.
Verify the proposed configuration against the pilot rather than assuming any platform automatically supplies every control. This approach is most relevant when the problem extends beyond sending a survey to maintaining usable feedback across teams, sources and time.
For the wider workflow, explore customer feedback management. If satisfaction with an interaction is the decision you need to support, compare the requirements in CSAT survey software.
Frequently asked questions
Do passives count in the NPS denominator?
Yes. Valid ratings from passives remain in the denominator. NPS subtracts the percentage of detractors from the percentage of promoters; it does not remove passives before calculating those percentages.
Can NPS software analyze comments automatically?
Many products offer text analysis. Evaluate the results on your own authorized sample, including mixed sentiment and multiple topics, and check how a reviewer can inspect sources and correct classifications.
Should we identify every NPS respondent?
Not necessarily. Named relationship follow-up and anonymous research serve different purposes. Choose and explain the appropriate arrangement, and do not imply anonymity if responses can be linked back to a person.
Does an improving score prove the follow-up worked?
No. Audience changes, timing and other influences may affect the score. Review comparable evidence and distinguish descriptive change from a causal claim about the action taken.
What is the most useful buying test?
Run a complete cycle with known counts, difficult comments, customer context, a repeated wave and a follow-up action. Verify the calculations, source evidence and routine work your team will need to perform.
Choose your next step
King Center
Related practice: making participant feedback useful for program review. This story illustrates feedback use, not an NPS result.
Read the customer story →