Voice of the customer
What unhappy Starbucks customers are really saying
Hundreds of public reviews, coded into themes, to find which problems drive one-star ratings.
Original project · 2025Data AnalysisBusiness StrategyStakeholder Presentation
The question
Which themes separate unhappy customers from happy ones, and where should a recovery plan start?
The original project
In my Starbucks project I led a team analysis of negative customer feedback on canned drinks and presented a structured product-recovery strategy. This page applies the same voice-of-customer method to a public dataset of Starbucks reviews.
At a glance
Reviews read850
Average rating1.9out of 5
1-star share64%
Most-mentioned theme in 1–2★42.2%Staff & service
The analysis
How customers rate their visit
705 reviews with a star rating
What unhappy customers talk about
Share of 1–2★ reviews that mention each theme
- Staff & service42.2%
- App & payments24%
- Price & value22.6%
- Wait time20.7%
- Drink & food quality11.3%
- Order accuracy10.2%
- Cleanliness6%
Themes by star rating
Share of reviews in each rating group that mention the theme
1–2★3★4–5★
Staff & service42.2%39.4%41%
Wait time20.7%18.2%11.5%
Order accuracy10.2%12.1%3.3%
App & payments24%18.2%11.5%
Price & value22.6%30.3%13.9%
Drink & food quality11.3%27.3%15.6%
Cleanliness6%0%1.6%
Average rating by year
Years with at least 15 rated reviews
Most-reviewed states
| CA | 139 | 1.8 | 61.9% |
| FL | 41 | 1.8 | 65.9% |
| TX | 35 | 2.1 | 60% |
| WA | 34 | 1.7 | 67.7% |
| NY | 31 | 1.6 | 74.2% |
| GA | 23 | 1.5 | 78.3% |
| NC | 23 | 1.6 | 65.2% |
| AZ | 22 | 1.9 | 68.2% |
| IL | 21 | 1.8 | 66.7% |
| NJ | 18 | 1.7 | 77.8% |
What the data shows
- The average rating is 1.9 out of 5, and 64% of rated reviews give a single star.
- Staff are the story either way: they come up in 42.2% of 1–2★ reviews and 41% of 4–5★ reviews.
- App & payments is the most distinctly negative theme — mentioned in 24% of low ratings vs 11.5% of high ones.
- The average rating was 1.2 in 2011 and 1.7 in 2023.
Recommendations
- Invest in service recovery training — staff are what customers remember, good or bad.
- Fix the app & payments issues first: they are the clearest driver of low ratings.
- Close the loop: reply to low reviews and track the theme mix monthly.
How the analysis works
- Parsed each review's rating, date, state and text; dropped placeholder "No Review Text" entries from text analysis.
- Built a keyword lexicon for seven themes and counted reviews that mention each theme at least once.
- Compared theme mentions across 1–2★, 3★ and 4–5★ reviews.
- Tracked the average rating by year.
Caveats
- Review sites over-represent unhappy customers, so the absolute ratings are biased low.
- Keyword themes are a first pass; a production version would use a trained classifier.
Methods
- Text coding & lexicons
- Theme frequency
- Segment comparison
- Trend analysis