# How We Track Review Sentiment Over Time

> A look at how tldr.money turns public Google reviews into a month-by-month sentiment timeline on brand pages, and how to read it.
- **Published**: 2026-08-16
- **Modified**: 2026-08-16
- **Category**: health-insurance
- **URL**: https://www.tldr.money/articles/health-insurance/how-we-track-review-sentiment

---

A single star rating hides the story. Before you trust a rating on the [health insurance](/health-insurance) hub, it helps to know whether the experience behind it is improving or slipping. That is what the new **review sentiment timeline** on our brand pages is built to show.

## What the timeline shows

Every brand page now carries a sentiment section with four things:

- an overall sentiment read, from very positive to very negative,
- the share of reviews that are positive, neutral, and critical,
- a month-by-month line of average sentiment, with review volume underneath,
- a trend badge that compares recent months against earlier ones.

The point is direction, not just a number. A brand sitting at four stars for years reads very differently from one that slid from four to two over six months, even though a plain average might look similar.

## How we build it

We run a small pipeline that does three jobs in order.

1. **Collect reviews.** We gather publicly posted Google reviews for the brand, keeping the review text and, crucially, the timestamp. The timestamp is what makes a timeline possible.
2. **Score sentiment.** Each review is scored on a scale from -1 (very negative) to +1 (very positive) using a lexicon-based model. It accounts for negation ("not helpful"), intensifiers ("very slow"), and emphasis, so it does not just count happy words.
3. **Aggregate by month.** Reviews are bucketed into months. For each month we compute the average sentiment, the average star rating, the review count, and the positive/neutral/critical split. Then we compare the earliest months to the latest to label the trend.

The result is saved as structured data that the site reads at build time, so the widget stays fast and does not call out to anything while you browse.

## How to read it

- **Watch the slope, not one dot.** A single dip can be noise. A steady decline across several months is a signal worth taking seriously.
- **Check volume too.** A great score built on three reviews is weaker evidence than a solid score built on hundreds. The bars under the line show how many reviews each month rests on.
- **Read the highlights.** We surface a few of the most positive and most critical reviews so the number has context. Recurring complaints about claim delays matter more than one angry outlier.

## What it is not

Sentiment scores are directional, not precise, and they are not an endorsement. Automated scoring can miss sarcasm and context, and public reviews skew toward people who felt strongly either way. Treat the timeline as one input. For claim reliability, still check the settlement ratio and the specific plan wording, and compare at least two brands before you buy. For premium planning, our [tax planning](/tax-planning) guide covers how to structure annual payments.

We would rather show you the trend and be honest about its limits than hand you a single score and pretend it settles the question.
---
- [More health-insurance articles](https://www.tldr.money/articles/category/health-insurance)
- [All articles](https://www.tldr.money/articles)