Contextualized reading
Interpret a score, average, or comment from the question asked, period, response count, and filters actually visible.
Help Centre · Customer experience
What can you actually conclude from your survey responses?
Start by checking the period, filters, number of responses, and comments actually displayed. You can then spot a trend or something worth reviewing without turning a few responses into a certainty.
Immediate answer
Start with what the results actually show
A survey result should be read in light of the covered period, selected scope, and visible filters. Response count provides important context. When a relevant denominator is genuinely available, the corresponding proportion may also help contextualize results. The question type and scale used influence interpretation. A mean, distribution, or change over time, when displayed, describes only part of the situation. A variation or comment does not prove a single cause. Interpretation produces an observation for the team to verify, not automatically a priority, recovery case, or prediction.
01Scope
Check what the view covers
To review satisfaction or feedback data cautiously, start by describing what the page actually displays before proposing a hypothesis.
Measurement and interpretation
Distinguish information displayed in the interface from the interpretation formulated by the team.
Describe before explaining
First describe what is observed before proposing a hypothesis, priority, or action.
A view remains partial
A view may contain missing data, response bias, or differences between customers who respond and those who do not.
02Indicators
Read available elements without making them say more than they do
The following elements describe possible types of reading. Their availability and presentation may vary according to the view used.
Score or average
Summarizes several responses into one value when such an indicator is displayed. This value may hide differences among individual responses.
Response count
Indicates the volume included in the view. It guarantees neither representativeness nor the absence of bias.
Proportion, when a denominator is available
When a relevant denominator is genuinely available, a proportion can help place the response count in context.
Distribution or change
When a distribution or change over time is displayed, it can help identify differences or movement. It does not prove their cause.
Comments
Comments add qualitative context, but an individual comment does not automatically represent all respondents.
Grouping or summary
When an automated grouping or summary is displayed, verify it against original responses and available context.
03Compare
Compare consistent views
A useful comparison seeks to limit methodological differences between views. It does not turn a variation into proof of causation.
- Same reference: compare the same question or indicator.
- Comparable scale: check that the scale and choices allow a reasonable comparison.
- Consistent periods: review durations and contexts that can reasonably be compared.
- Visible filters: keep comparable selections between views.
- Known changes: record modifications made to the questionnaire or process.
- Response count: account for volume and avoid strong conclusions when it is limited.
Remember: a variation between two periods does not prove a single cause, operational improvement, or future financial result.
04Contextualize
Contextualize comments and any groupings
Text responses may explain a perception, but they require cautious reading before qualification.
Individual experience
A comment describes an experience or perception in a given context. It does not automatically represent all respondents.
Displayed grouping
When a theme or automated summary is displayed, use it as a reading aid and verify it against available comments.
Avoid generalization
A word, theme, or small number of comments is not enough to demonstrate a cause or general trend.
Transition to a signal
A comment or grouping does not automatically become a diagnosis, priority, or case. Qualification is a separate step.
Method
Formulate an observation to verify
Use this six-step method to maintain a clear separation between what is displayed, what is assumed, and what will be decided.
Describe the observation
State factually what is visible in the view without immediately adding an explanation.
Specify the context
Recall the period, visible filters, and number of responses included.
Name the element
Identify the question, indicator, visualization, or group of comments concerned.
Record the limits
Note missing data, low volume, or methodological differences that may weaken the interpretation.
Define the verification
Indicate which human or operational step will confirm, qualify, or refute the observation.
Separate levels
Distinguish the observed fact, explanatory hypothesis, possible signal, and final decision.
Observed fact
What is actually displayed in the period and scope being studied.
Hypothesis
A possible explanation that still requires verification.
Possible decision
An action determined separately after reviewing context and limits.
05Resources
Continue reviewing surveys
Review the guide corresponding to configuration, sending, qualification, or support.
Surveys and customer experience
Learn about documented features for preparing surveys, collecting responses, and reviewing available results.
View the page →How do you configure a useful survey?
Learn how to prepare and review questionnaire sections, response formats, and language variants.
Read the article →How do you choose survey timing and channel?
Learn which elements to review to understand the sending options available in your journey.
Read the article →How do you turn a response into a signal to review?
Learn how to review a response in context before treating it as a signal to qualify.
Read the article →Support
Contact supportDocument unexpected behaviour
If you observe unexpected behaviour in the results display, specify the page and journey used, visible filters at the time, the element, indicator, or visualization concerned, observed result, exact message, approximate date and time, and a sanitized screenshot if useful. Never send a password, authentication code, key, token, technical secret, or unnecessary personal information.