Sentiment trend analysis: reading month-on-month change

Why trends matter

Sentiment trend analysis is where review data earns its keep. Levels tell you where you are; trends tell you what is changing and give you time to act. But month-to-month numbers move for many reasons, and not all of them are real.

Comparing months fairly

• Same stores: new or closed stores change the mix.
• Same sources: adding app reviews changes the picture.
• Same categories: a taxonomy change can create an apparent trend.
• Enough mentions: a topic with ten mentions a month will swing.

Signal vs noise

• Look for changes that persist for two or more periods.
• Prefer changes visible across several stores, or clearly concentrated in a few.
• Check whether the change is larger than the usual month-to-month variation for that topic.
• Read the reviews: real changes have consistent explanations.

Seasonality

Festive seasons, weddings, holidays and sales change both volume and complaints. Compare with the same month last year as well as last month. A waiting complaint rising in Diwali week is expected; rising in a quiet month is not.

Explaining the change

• What moved: the topic and the size of the change.
• Where: which stores, regions, products or versions.
• When: the week it started.
• Why: what customers say, and what changed in operations.

An example

Billing negative share rises from 40% to 55% in March. The same month last year was flat. The rise is concentrated in 30 stores that switched billing systems in late February, and the reviews mention slow card payments. That is a real, explainable change, not noise.

Common mistakes

• Reacting to a single month.
• Ignoring changes in store or source mix.
• Comparing across a taxonomy change.
• Reporting changes without reviews.

Common questions

Checklist

• Same stores, sources and categories.
• Enough mentions.
• Year-on-year check.
• Explanation with reviews.

In short

Compare months fairly, separate persistent and concentrated changes from noise, check last year for seasonality, and explain every reported change with the reviews behind it.

Trends at small stores

For stores with few reviews, monthly topic sentiment swings too much to read. Use rolling three-month windows, compare with the same period last year, and read every review. Reserve month-on-month trend calls for stores and topics with enough mentions.

Charting trends

Plot each topic's negative share as a line by month, with a small marker for the number of mentions. Mark the dates of known changes, such as a new system or policy, directly on the chart. Readers then see the change and its likely cause together.

Annotate changes

Keep a simple log of operational changes: new systems, policies, staffing changes, store refits and launches, each with a date and the stores affected. When a sentiment trend moves, check the log first; most real changes have an operational explanation.

Doing this with Falcon

Falcon measures sentiment for every theme a review mentions rather than one score per review, keeps the exact words that support each label, and reports negative share by topic, store and month with small samples flagged and unclear texts kept visible. Every number opens the reviews behind it.