Power BI survey dashboard: responses and comparisons

A Power BI survey dashboard should separate how many customers responded from what those customers said. Compare groups using eligible invitations for participation and valid answers for each question's score, then keep the counts visible beside every percentage.

Microsoft describes a Power BI dashboard as a single page of highlights. For a survey review, that page should answer a narrow business question, such as which customer group needs a service follow-up, while retaining enough detail to challenge the conclusion.

Covirage is not affiliated with Microsoft.

Define the survey population before calculating

Start with an aggregate export containing group, survey window, eligible invitations, completed surveys, valid answers to the selected question, and favorable answers. A separate question identifier keeps answers from different questions out of the same denominator. Use aggregate customer groups rather than names or free-text responses for this review.

Define an eligible invitation as a unique customer contact selected for this survey after the agreed exclusions. Document how duplicates, undeliverable invitations, and ineligible contacts are handled. Define a completed survey consistently, and do not silently count partial completions as finished surveys in one group but exclude them in another.

For this example, the question uses a five-point satisfaction scale. Ratings of four or five count as favorable. A valid answer is any recorded rating on that scale; a blank is missing, rather than a low rating. This is the scoring rule for the example, not a claim that every survey should use it.

Response rate = Completed surveys / Eligible invitations

Favorable percentage = Favorable answers / Valid answers to the question

Missing question answers = Completed surveys − Valid question answers

These measures have different denominators because they answer different questions. Keep the definitions visible wherever someone might compare the results.

A worked comparison across customer groups

A US business services company reviews a September 2026 customer survey. It invited customers in mutually exclusive midmarket, enterprise, and partner groups. The export contains the following counts; percentages are calculated from unrounded counts and displayed to two decimal places.

Midmarket response rate is 200 / 500 = 40.00%. Its favorable percentage is 133 / 190 = 70.00%. Dividing favorable answers by completed surveys would answer a different question and incorrectly treat the missing satisfaction answers as unfavorable.

Enterprise reports both higher participation and higher favorable sentiment than the other groups in these returned surveys. That finding supports a follow-up question about differences in service experience. It does not establish that the nonresponding customers share the respondents' views or that customer group caused the difference.

In a spreadsheet matching the table, response rate in F2 is =IF(B2=0,"",C2/B2). Favorable percentage in G2 is =IF(D2=0,"",E2/D2). The portfolio formulas are:

=SUM(C2:C4)/SUM(B2:B4)
=SUM(E2:E4)/SUM(D2:D4)

Those return 48.00% and 71.78%. The second figure weights each valid answer equally; it does not give each customer group equal weight.

Check missing answers and reconcile the counts

The question has missing answers in every group:

The first proof is 450 valid + 30 missing = 480 completed surveys. Question coverage is 450 / 480 = 93.75%. The second proof is that 323 favorable + 127 other valid answers = 450 valid answers. The 127 follows from 450 − 323; it includes every valid rating outside the favorable definition.

For every group, favorable answers must be no greater than valid answers, valid answers no greater than completions, and completions no greater than invitations under this example's completion rules. Tie invitation and completion totals back to the survey export, and confirm that each customer belongs to only one comparison group.

These checks are easiest when the exact counts remain visible. Microsoft's guidance describes Power BI tables as useful for detailed comparisons and exact values, rather than relying entirely on charts.

Change one input: more favorable midmarket answers

Keep invitations, completions, and valid answers unchanged. Change only the midmarket favorable count from 133 to 152. This is a sensitivity calculation showing how the score moves when the answer mix changes; it does not claim that a service action produced those answers.

The proof is 152 − 133 = 19 additional favorable answers. Total favorable answers therefore become 323 + 19 = 342, and 342 / 450 = 76.00%. The overall increase is 4.22 percentage points, rounded from the exact difference between 76.00% and the original unrounded result.

Participation remains 480 / 1,000 = 48.00%. Missing answers remain 30. The page should show a change in sentiment, with no corresponding change in either participation or question coverage.

Arrange the review around decisions

Use a participation summary, a group score comparison, and the count reconciliation. Add the question wording, favorable definition, survey dates, and exclusions in a visible note. Have the customer insights owner maintain those definitions and the relevant service manager own the follow-up action.

The wider Power BI dashboard examples show how other business summaries fit together. An executive dashboard can carry the approved conclusion, while the Excel dashboard guide covers a spreadsheet approach to displaying the same verified measures.

Where it goes wrong

• Changing question wording. A score change may reflect a different question. Flag wording changes before drawing a trend line.
• Counting missing answers as zero. Keep missing responses separate so question participation is visible.
• Overlapping groups. Duplicate customers inflate the total and distort the apparent weight of a segment.
• Reading every gap as meaningful. Small or selectively responding groups need caution. Counts and participation should accompany the score.
• Mixing weighted and unweighted results. A weighted estimate answers a different population question. Label it and preserve the weighting method.

Do this on your own file in Covirage

Bring an aggregate survey export or Google Sheet containing group counts and the question definitions. Covirage maps the columns; its tools compute and check response rates, favorable percentages, and missing-answer totals, while the external AI model explains the results with every figure traced to its rows. Create a free Covirage account, which takes about a minute, then set up your business and bring a sample of your own data.

See self-service analytics for the supporting business use case.