deepsight cloud · Data quality check Check unusual responses before the analysis.
Spot empty or unintelligible entries and responses without substance such as “no comment”, and filter them out before the analysis. Filtered responses no longer count toward the base.
Flagged does not mean wrong
deepsight assigns unusual responses to a category; flagged responses are not included in the analysis. Whether a category still says something about your question depends on the question. Make your own assessment in the example.
| Category | Example response | Your assessment |
|---|---|---|
| No issue mentioned“all good”A valid answer to “What did you like?”. | “all good”A valid answer to “What did you like?”. | |
| Unintelligible input“asdfjkl”No recognizable content. | “asdfjkl”No recognizable content. | |
| Empty response–Question skipped. | –Question skipped. | |
| No comment“no comment”The person chose not to answer. | “no comment”The person chose not to answer. | |
| No key point“won’t change anything anyway”Frustration without a concrete issue – still revealing for some questions. | “won’t change anything anyway”Frustration without a concrete issue – still revealing for some questions. |
No substance for your question: 3 of 5 categories. If a flagged category does say something about your question, weigh up whether the check is the right fit. Illustrative example. The check is optional; the data quality overview shows how many responses are affected.
Questions about the data quality check
No. The check is an optional step before the analysis. Without it, every response counts toward the base. Analyses started directly from a survey run without the check.
Empty responses, unintelligible strings and responses without substance such as “no comment” or “all good”. The check does not detect duplicate responses; for responses in other languages there is translation.
They are not included in the analysis: topics, sentiment and metrics are then calculated from the remaining responses only. Check the data quality overview to see how many responses are affected.
Shares and metrics refer to the analyzed responses; anything the check filters out does not count toward the base. This also applies to the NPS: with the check switched on, it only covers respondents whose responses remain after the check.
In market research studies, this is how you decide up front whether responses without substance go into the base.
More features · Prepare data
See the feature with your own data
We will walk you through the analysis using an example from your field.