Analyzing qualitative data in Excel for business feedback

Analyzing qualitative data in Excel starts with people reading feedback and assigning consistent codes, followed by spreadsheet counts of those codes. Keep the original comments beside the labels so every summary can be checked against what customers said. A useful business report compares themes across customer segments and names the next question or action.

Start with the response export and a codebook

Export the open-ended question responses from your customer survey or feedback register. Keep a stable response ID, submission date, customer segment, question asked, original text, reviewer, and review status. Decide whether the reporting unit is a response or a customer: several comments from one account can otherwise dominate a customer-level conclusion.

Use one question and one collection period for the first comparison. A renewal interview and a delivery complaint form invite different answers. Do not combine them into an unlabeled theme percentage. Preserve segment labels from the customer record rather than guessing a segment from the tone of the comment.

Read the full set before finalizing codes. The CDC's guide to qualitative analysis describes using a coding scheme to organize text, then reviewing coded material for patterns. For business feedback, translate that method into a short codebook people can apply consistently.

These are descriptive codes. A broader finding, such as uncertainty during handoffs, needs an explanation supported by the underlying comments. Frequency alone does not supply that explanation.

Build a small, auditable worksheet

The following fictional US distributor receives feedback from enterprise and small-business customers. Paste the table into A1:F7. A flag of 1 means the reviewer found the theme; 0 means the reviewer read the response and did not find it. An unread response stays blank until reviewed.

Do not use a word search as the final coder. A customer could write that a shipment was not late, or describe a problem without using your chosen word. A reviewer must interpret the sentence in context.

For scored survey questions collected alongside these comments, use the separate method for analyzing Likert-scale data in Excel. Keep those response scales distinct from the binary flags used here.

Count themes within each segment

Put Enterprise in H2 and Small business in H3. Use I for response count, J for delivery count, K for billing count, L for onboarding count, and M for delivery share. Enter these formulas in row 2 and fill down:

=COUNTIF($B$2:$B$7,H2)
=COUNTIFS($B$2:$B$7,H2,$D$2:$D$7,1)
=COUNTIFS($B$2:$B$7,H2,$E$2:$E$7,1)
=COUNTIFS($B$2:$B$7,H2,$F$2:$F$7,1)
=IF(I2=0,NA(),J2/I2)

Microsoft documents COUNTIF for counting a matching condition and COUNTIFS for counting rows that meet multiple conditions. The second function lets the workbook require both the segment and the approved theme flag.

Percentages are rounded to one decimal place. Enterprise delivery share is 2 / 3 = 66.7%; small-business delivery share is 1 / 3 = 33.3%. The difference is 33.3 percentage points using unrounded inputs. This tiny example identifies comments to inspect; it does not establish how all customers feel.

Prove the counts before interpreting them

The segment denominators must sum to the number of eligible responses: 3 + 3 = 6. Theme counts must match their source columns: delivery is 1 + 1 + 0 + 1 + 0 + 0 = 3. Billing also totals 3, and onboarding totals 2.

There are 3 + 3 + 2 = 8 theme mentions across 6 responses. Dividing mentions by responses gives 133.3%, which is valid for overlapping themes. It would be wrong to label that figure a share of distinct respondents.

Check response uniqueness in G2 and fill down:

=COUNTIF($A$2:$A$7,A2)=1

Every row should return TRUE. Then verify the code cells contain only approved numeric flags:

=AND(COUNT(D2:F2)=3,COUNTIF(D2:F2,0)+COUNTIF(D2:F2,1)=3)

A failed check requires review before publication. For a larger workbook, follow the separate guidance on analyzing large data sets in Excel without changing the response grain or denominator.

Turn the comparison into an operational question

Customer success owns the code definitions and the interpretation. Operations can investigate the shipment comments; finance can inspect the invoice examples. Keep a short action register containing the theme, supporting response IDs, accountable owner, and follow-up date.

Read examples that challenge the apparent pattern as well as those that support it. If a segment's customers were asked a different question, investigate the collection method before claiming a service difference. When the segments need redesigning, use customer segmentation from business records rather than changing labels to make a chart look cleaner.

Where it goes wrong

• Counting comments as customers. Repeat submissions need an explicit customer-level rule before making customer-level claims.
• Treating unread blanks as zero. An unreviewed comment is missing coding, not evidence that a theme is absent.
• Changing definitions midstream. A revised codebook requires reconsidering affected earlier responses.
• Forcing totals to sum across overlapping themes. One response can contribute to several theme counts.
• Equating frequency with impact. A rarely mentioned contractual problem may deserve attention before a common minor inconvenience.

Covirage is not affiliated with Microsoft.

Compare your coded feedback in Covirage

Bring a workbook of manually coded feedback with response IDs, segments, and approved theme flags. Covirage maps the columns, and its tools compute and check the counts and comparisons; the external AI model explains the results with every figure traced to its rows. Create a free Covirage account and bring a sample of your own data.

See Excel analysis in Covirage for the file-based workflow.