Quick answer
To compare Chatbot handoff results, keep input, format and settings fixed and change one variable. Handoff should carry context, not restart the conversation.
Why it matters
Changing several variables at once makes results unattributable.
Pass intent, order and attempts at handoff.
Step by step
- Define QA goals and observable behaviors before building a giant scorecard.
- Stratify samples by shift, channel and risk so they represent real problems.
- Calibrate scoring boundaries before ranking or performance decisions.
- Attribute issues to knowledge, process, system or people before choosing fixes.
Common mistakes
- Scoring without calibration makes managers incomparable.
- Sampling only easy tickets biases the result.
- Using QA as punishment makes agents hide context.
- Looking only at CSAT while ignoring resolution and rework.
Checklist
- Scorecard has behavioral anchors.
- Sampling has stratification rules.
- Calibration records disputed samples.
- Issues have owners and review dates.
Key facts
- Pass intent, order and attempts at handoff.
- Define triggers: sentiment, amount, permissions, failure count.
- The bot should not pretend to handle out-of-scope requests.
FAQ
How do I compare two Chatbot handoff results?
To compare Chatbot handoff results, keep input, format and settings fixed and change one variable. Handoff should carry context, not restart the conversation.
How long does it take?
Usually 1-3 minutes; for intermittent issues, run 30 seconds and retest elsewhere.
Is data uploaded?
Runs locally and does not upload video/audio; extensions may still read the page.
Source: Intercom handoff