One review, several issues: counting multi-topic reviews

The problem

Multi topic reviews are the norm, not the exception. "Beautiful designs, patient staff, but billing took 25 minutes" mentions product, staff and billing. If you count it once, under one category, two of those three facts disappear. If you count it three times without saying so, totals stop adding up.

Two counts, two questions

• Mentions: how often is each topic raised? A review with three topics adds one mention to each. Use for topic volume and topic sentiment.
• Reviews: how many customers raised a topic at all? Each review counts once per topic. Use for prevalence: the share of all reviews that complain about billing.

Both are valid. Report which one each number is.

Sentiment per mention

Each mention gets its own sentiment. The example review is positive about product and staff, negative about billing. Averaging them into one review sentiment loses exactly what you need to act on.

Mixed sentiment on one topic

Sometimes a review says something good and bad about the same topic: "great food, small portions". Either split into two topics (food quality and portion size) or set a rule, such as counting the review once as negative for that topic's negative share while also recording the positive. Write the rule down and keep it.

A worked example

• 100 reviews; 60 mention staff, 40 mention billing, 30 mention product.
• Total mentions: 130, more than the number of reviews, which is expected.
• Billing negative mentions: 24 of 40, so negative share within billing is 60%.
• Prevalence of billing complaints: 24 of 100 reviews, or 24%.

Common mistakes

• Assigning each review to one category only.
• Summing mentions and calling them reviews.
• Averaging sentiment across topics.
• Changing the mixed-sentiment rule mid-year.

Common questions

Checklist

• Mentions labelled separately.
• Each number marked as mentions or reviews.
• Mixed-sentiment rule written down.
• No double counting within a topic.

In short

Label each topic in a review separately with its own sentiment, report mentions and review prevalence clearly as different measures, and apply one rule for mixed sentiment.

Reporting it clearly

In every chart or table, label whether numbers are mentions or reviews. A footnote such as "Counts are topic mentions; a review mentioning three topics is counted once in each" prevents the most common question leadership asks: why the topic counts add up to more than the total number of reviews.

When presenting a single topic, give both: "Billing came up in 40% of reviews; 60% of billing mentions were negative." The first tells you how big the topic is; the second how bad it is.

Checking a tool's approach

Ask any analysis tool one question: if a review praises staff and complains about billing, what happens? If the answer is one label or one sentiment for the review, its topic numbers will understate complaints inside positive reviews, which are often the most fixable ones.

Doing this with Falcon

Falcon builds the theme list from your own reviews, then labels every review against it with a multilingual model, scoring sentiment for each mention and keeping the exact quote behind each label. Texts too short to judge are marked unclear rather than guessed, and low-confidence labels are flagged for a person to check. Negative share by topic, location and month then shows what is rising where, with small samples flagged.