What people ask about
Conversations by intent
Order Status is the single largest intent (133 of 500), so it is the first place to invest in self-service.
Resolution rate
Share of conversations resolved by intent
“Other” resolves least often (68%). Order Status, the biggest bucket, also sits below average at 75%.
Satisfaction
Average satisfaction by intent
Scores cluster near 3 out of 5 for every intent, with “Other” lowest. There is no intent where the assistant clearly delights.
Model performance
Precision, recall, F1 and misclassification by version
Across six versions F1 rose from 0.67 to 0.87 and misclassification fell from 30% to 12%.
About the data. Values come from the four full Plotly dashboards on the old site (dashboard_*.html), which were generated from the underlying sample conversation log; resolution and satisfaction are averages per intent from that log. The old site also had smaller hand-typed versions of these charts with different numbers; those were not used because they did not match the generated dashboards.
Rebuilt from:
- dashboard_intent_distribution.html
- dashboard_resolution_rate.html
- dashboard_satisfaction.html
- dashboard_model_performance.html
- interactive_cases_refined/*_case.html (earlier hand-typed copies, superseded)
More reports
Recruitment Funnel Analytics
Where candidates drop out of the hiring funnel, which stage takes the longest, and how time to hire trends as the process tightens.
Open reportPredicting Turnover: AI-Driven Attrition Risk
A predictive attrition model that flags employees at risk of leaving, so HR can have the stay conversation early.
Open reportReducing 90-Day Attrition
Predictive and prescriptive analytics to decide where a limited onboarding budget does the most good for early retention.
Open report