How to analyse customer satisfaction survey data

The process in brief

This guide covers how to analyze customer satisfaction survey data (or analyse, in British spelling) from a raw export to a report someone can act on. The steps are the same whether you have 300 responses or 30,000: clean, score, cut, explain, prioritise, report.

1. Clean the responses.
2. Calculate scores consistently.
3. Cut by store, touchpoint and segment.
4. Code the open comments.
5. Find the drivers.
6. Report and assign actions.

Step 1: clean the data

• Remove test responses and staff entries.
• Remove duplicates, such as the same customer submitting twice.
• Decide how to treat partial responses and apply it consistently.
• Check that store IDs, dates and channels are filled in and valid.
• Look for impossible patterns, such as hundreds of perfect scores from one device in an hour.

Step 2: calculate scores

Calculate CSAT as the share of satisfied answers (commonly 4 and 5 on a 5-point scale) or as the average, using the method you used before. Show the number of responses beside every score. Track the share of dissatisfied answers as well; it often moves earlier.

Step 3: cut the data

• By store or branch, with counts.
• By touchpoint or channel: visit, delivery, support.
• By segment: new or repeat customer, value band, region.
• By time: day of week and hour if you have them.

Flag any cell with too few responses to trust, rather than ranking it.

Step 4: code the open comments

Build a short list of themes from a sample of comments, such as staff, wait time, product availability, cleanliness, price, returns. Tag each comment with themes and sentiment. Keep the list stable so themes can be compared month to month.

Step 5: find the drivers

Compare theme frequency among dissatisfied and satisfied respondents. A theme that is common in low scores and rare in high scores is a driver of dissatisfaction. If you asked diagnostic questions, such as wait time bands, compare scores across those answers too: if customers who waited over ten minutes score far lower, wait time matters.

Step 6: prioritise

• Volume: how many customers mention it.
• Severity: how low their scores are.
• Trend: is it growing.
• Spread: one store or many.
• Fixability: is there a clear owner and action.

A worked example

A 60-store retailer's monthly CSAT falls from 81% to 77% on 4,000 responses. Cutting by store shows the drop is concentrated in 12 stores. Coded comments from those stores mention billing waits far more often than before. The diagnostic question confirms it: customers reporting waits over ten minutes rose sharply. The 12 stores share a new staffing roster introduced that month.

Step 7: report

• Headline score with counts and trend.
• What changed and where.
• Top drivers with quotes.
• Recommended actions with owners and dates.
• Method note: question, scale, calculation, period.

Cross-check with other sources

Survey data comes from invited customers who chose to answer. Complaint logs and public reviews come from different customers. If a driver appears in both, act with more confidence; if it appears only in one, find out why.

Common mistakes

• Calculating before cleaning.
• Ranking stores with tiny samples.
• Changing the theme list every month.
• Reporting the score without the drivers.
• Treating correlation in a small cut as proof.

Common questions

Checklist

• Data cleaned.
• Scores with counts and fixed method.
• Cuts with small samples flagged.
• Comments coded with stable themes.
• Drivers compared across high and low scores.
• Actions assigned.
• Cross-checked with reviews or complaints.

In short

Analyse customer satisfaction survey data by cleaning it, calculating scores with counts, cutting by store and segment, coding comments into stable themes, finding drivers, prioritising and reporting actions with owners.

Doing this with Falcon

Falcon does not run surveys or feedback forms. It analyses the unprompted feedback customers already leave in Google Maps reviews of your locations and Google Play reviews of your app: each review is classified into themes, sub-themes and specific issues from your own taxonomy, with the sentiment of each mention and the exact quote behind it. Falcon does the theme coding for reviews with a taxonomy built from your own reviews, so the same theme names can be used to compare survey drivers with review drivers by store and month.