Glossary · Reviews & analysis
What is review severity?
Severity measures how bad a review’s problem is, independently of how the reviewer felt. SellerSort scores it on five levels as a probability-weighted position — so a review can sit between levels — and counts anything at 4 or above as urgent.
What this comes down to
- Five ordered levels, from nothing wrong to harm, lost money or a safety issue.
- The score is probability-weighted, so it can fall between levels — read it as a ranking.
- Reviews at severity 4 or above are counted as urgent.
- A loud complaint is not always a severe one, and a calm one can be.
Why severity is separate from sentiment
A furious review about slow delivery and a calm review saying the handle snapped off after two weeks are both negative. Only one describes a product failing. Sentiment captures the first dimension; severity captures the second, which is usually the one that decides what to fix first.
How the score works
Each review is placed on five ordered levels. Instead of picking one level, the model gives a probability for each and the score is the weighted position, so it can fall between two levels. Use it to rank and to compare against the urgent line; a 3.4 and a 3.5 are the same review for practical purposes. The summary counts reviews at or above 4 as urgent and reports the mean severity per topic, so “shipping” and “durability” can be compared on how bad, not only how often.
The practical method is in prioritise negative reviews by severity.
The five levels
| Level | What the review describes |
|---|---|
| 1 | Positive or neutral feedback. Nothing is wrong and nothing needs fixing. |
| 2 | A minor gripe stated in passing. The reviewer is broadly satisfied and would buy again. |
| 3 | A real problem that annoyed the reviewer but did not stop them using what they bought. |
| 4 | A significant problem: the item or app did not do what they needed, or they wasted real time or money on it. |
| 5 | A severe problem. The reviewer was harmed, lost money they have not recovered, is warning other buyers off, or reports a safety or data issue. |
Each level describes a situation, not a feeling, so two people reading the same review should land in the same place. The question the model is asked says so directly: judge the problem itself, not how strongly it is worded.
Questions
What does severity 4 mean?
It is the threshold SellerSort uses for urgent: a problem serious enough that it should be looked at before the rest. The Review Analyzer counts these separately.
Is a one-star review always high severity?
No. The star rating is how the customer felt; severity is what happened to them. A calm three-star review about a product that stopped working can score higher than an angry one-star review about a late parcel.
Read next
- Review sentiment analysisClassifying each customer review as positive, neutral, negative or mixed, so the tone of thousands of reviews can be counted rather than read.
- Actionable reviewA review that needs a response from the seller — an unresolved problem, a direct question, or something owed to the customer.
- How to prioritise negative reviews by severityNot every one-star review is urgent. Rank complaints by severity and how often they recur, so the fix list starts with what actually hurts.