Severity levels
Severity levels show how much attention an issue needs. They help you prioritize what to fix first when several problems appear in the same report card.
Inquio uses four severity levels: Low, Medium, High, and Critical.
What they measure
- how often the issue appears in the analyzed conversations
- how serious the issue is in context
The percentage shows how widespread the issue is across the dataset. Semantic seriousness can raise the severity level when the issue has a high business, safety, compliance, or customer-experience impact.
For example, an issue that appears in a small number of conversations may still be marked as High or Critical if it involves sensitive information, unsafe bot behavior, reputational risk, or a serious failure in an important user journey.
Severity ranks attention. It does not only count volume.
The four levels
| Severity | Share of conversations | What is means |
|---|---|---|
| Low | Less than 3% | The issue appears rarely or has limited impact. Review it after higher-severity issues. |
| Medium | 3–7.9% | The issue appears often enough to show a pattern. It should be reviewed and planned for improvement. |
| High | 8–14.9% | The issue affects a significant share of conversations or has a serious semantic impact. Prioritize it. |
| Critical | 15% or more | The issue is widespread, business-critical, safety-critical, or likely to seriously damage user trust. Treat it as a top priority. |
These percentage bands are the baseline. Inquio can level up severity when the meaning of the issue is serious enough.
Low
Low severity means the issue appears in less than 3% of conversations and does not usually suggest a broad failure.
Low issues are still worth reviewing, especially if they affect an important topic, but they normally come after Medium, High, and Critical items.
- an edge case
- a rare phrasing problem
- a small gap in one topic
- a conversation pattern that is not yet widespread
Do not ignore Low issues completely. They can become more important if the same pattern grows in later report cards.
Medium
Medium severity means the issue appears in 3% to 7.9% of conversations.
At this level, the issue is no longer isolated. It appears often enough to suggest a repeatable pattern in the bot’s behavior, content, or conversation flow.
- a missing answer in a recurring topic
- a dialog step that often confuses users
- an escalation path that does not work smoothly
- a product or service area where the bot is only partly prepared
Medium issues should be reviewed and planned for improvement, especially when they affect a high-volume topic.
High
High severity means the issue appears in 8% to 14.9% of conversations, or that the issue has been leveled up because of its semantic seriousness. A High issue may not be the most frequent issue in the report card, but it is important enough to prioritize.
- repeated incorrect answers in an important topic
- users frequently repeating themselves or getting stuck
- unresolved conversations in a key business process
- a serious customer-experience problem
- a safety or compliance concern that appears in multiple conversations
High issues should usually be addressed before ordinary Medium and Low issues, even if another issue has a similar conversation count.
Critical
Critical severity means the issue appears in 15% or more of conversations, or that the issue is serious enough to require immediate attention regardless of volume.
- a widespread failure in a core user journey
- a repeated inability to answer an important request
- a major containment problem where users are kept in the bot without resolution
- unsafe or non-compliant bot behavior
- exposure of sensitive information
- a pattern that can damage user trust or company reputation
Critical issues should be handled first. Open the affected conversations, confirm the pattern, and follow the recommended steps before moving on to lower-severity issues.
When seriousness raises the level
The percentage of affected conversations is not the only factor.
Inquio can raise the severity level when the issue is semantically serious. This means the content of the issue makes it more important than the raw volume alone would suggest.
- sensitive personal or financial information
- unsafe, discriminatory, or inappropriate bot behavior
- legal, compliance, or reputational risk
- incorrect answers about money, contracts, deadlines, or user rights
- churn risk from important customers
- failed handling of vulnerable or emotionally sensitive situations
- missed business opportunities with clear sales value
For example, a typo in 10% of conversations may be High because it is frequent. A data exposure issue in 2% of conversations may also be High or Critical because the impact is much more serious.
How to use severity
- Open Issues and Recommendations.
- Start with Critical issues.
- Within the same severity level, compare impact and the number of affected conversations.
- Read several affected conversations before making a change.
- Follow the recommended steps for the issue.
- Move to High, then Medium, then Low.
- Re-upload a fresh dataset later and compare whether the issue severity changed.
Severity helps you decide what to look at first. The affected conversations show you what actually happened.
Severity and Impact
Severity and impact are related, but they are not the same.
Severity tells you how serious the issue is and how urgently it should be reviewed.
Impact tells you how much the overall Inquio Score™ is expected to change if you follow the recommendation for that issue.
A Critical issue may have a high business or safety risk even if the predicted score impact is not the largest. A Medium issue may have a large score impact if it affects a heavily weighted dimension such as Resolution Accuracy.
- use severity to understand risk and attention level
- use impact to estimate how much the score may improve
- use affected conversations to understand the real pattern
Where to find it
- Issues and Recommendations — severity summary boxes, issue list, affected conversations, and recommended steps.
- Bot Report Card — identified issues and priority recommendations.
- Defender — incident severity for safety and compliance cases.
- Sales Radar — severity for Leads and Churn risks.