Coaching Summary
The agent shows developing technical skills but requires improvement in protocol adherence and documentation completeness, reflected in a 2.7 overall score this week.
A consistent gap in collecting critical device information and verifying baseline connectivity before proceeding with advanced troubleshooting steps.
Risk Flags
Critical dimension below threshold
Accuracy score of 4.0 but protocol score of 2.0 indicates significant protocol deviations in the single call handled
ExampleAgent failed to collect product model number and skipped WAN verification during Velop node troubleshooting
Correct behavior: Follow KB sequences for Velop troubleshooting: collect model number, verify WAN status, check firmware, document findings
Impact: Risk of misdiagnosis and repeated outages for customer with nine-node network
Week-over-Week Progress
Scorecard
| Dimension | Week Average | Calls Reviewed |
|---|---|---|
| Overall | 2.7 | 1 |
| Technical Accuracy | 4.0 | 1 |
| Protocol | 2.0 | 1 |
| Communication | 2.0 | 1 |
V2 Rubric (Shadow Grading)
V2 overall: 79.73% across 1 v2-scored calls this week
| Category | Week Average |
|---|---|
| Resolution | 4.06 |
| Technical Accuracy | 4.06 |
| Communication | 3.75 |
| Customer Ownership | 4.29 |
| Escalation Judgment | 5.0 |
| Customer Experience | 2.5 |
- Appropriate Escalation: 1
Score Diagnostics
Based on 1 calls reviewed this week.
Technical Findings
improvement
Did not collect product model number, which is critical for Velop troubleshooting [KB: velop_child_node_setup.md, velop_wifi_connectivity.md].
call f2ecc6dc-6be1-11f1-95d6-42010a62006f
improvement
Failed to verify parent node WAN/internet status before proceeding with child-node steps [KB: universal_isp_modem_diagnostics.md, velop_wifi_connectivity.md].
call f2ecc6dc-6be1-11f1-95d6-42010a62006f
improvement
Did not confirm firmware version or check for known issues with Velop models [KB: universal_eol_firmware.md].
call f2ecc6dc-6be1-11f1-95d6-42010a62006f
improvement
No case documentation or HappyFox ticket created despite clear need for escalation [KB: universal_escalation_guide.md].
call f2ecc6dc-6be1-11f1-95d6-42010a62006f
strength
Technical steps aligned with KB guidance for Velop node recovery (power cycle, sequential reconnection).
call f2ecc6dc-6be1-11f1-95d6-42010a62006f
Call Handling Findings
- Missing product model collectionAlways collect and verify the exact product model number before initiating any Velop troubleshooting steps, cross-referencing with model-specific KB articles.
- Incomplete WAN verificationAlways verify parent node WAN status (e.g., via wired connection test) before troubleshooting child-node connectivity, following universal_isp_modem_diagnostics.md procedures.
Calibration Notes
- Call duration 4035 seconds with overall score 2.7 and protocol score 2.0; high time investment but low protocol adherence suggests possible grader inconsistency or complex case
Callback Chains
No callback chains detected.
Documentation Mismatches
No documentation mismatches found.
Suggested Coaching Conversation
1
Clark, I noticed you effectively guided the customer through Velop node power cycling and sequential reconnection, which aligns well with our KB procedures — great technical adherence.
2
You mentioned not collecting the product model number during troubleshooting. How do you think this omission affected the diagnostic process?
3
When verifying internet connectivity, you focused on the modem but didn't confirm the parent node's WAN status. Can we walk through the KB steps for checking parent node connectivity?
4
The call ended with a callback scheduled but no HappyFox documentation. What processes do you use to ensure troubleshooting details are captured for L2 engineers?
5
You maintained a polite tone despite customer frustration — this is crucial for customer experience. How do you balance technical accuracy with empathetic communication under pressure?
Coaching Best Practices
Five Principles for Effective Coaching Conversations
- SBI feedback model — Describe the Situation, the Behavior you observed, and its Impact. Avoid labels (“you always…”) — describe the specific instance.
- Ask more than you tell — Start with “What did you notice on that call?” before offering your interpretation. Agents who self-diagnose retain more.
- Recognition before correction — Open with a genuine strength observation. Agents in a defensive posture can’t hear coaching.
- Psychological safety first — Frame mistakes as data, not character. “That call had a tricky wrinkle — let’s pull it apart” lands better than “you made an error here.”
- Close with a specific commitment — End every coaching conversation with one thing the agent will try on their next call, and a follow-up check-in within 48 hours.
Deep Evidence
| Case | Date | Duration | Overall | Accuracy | Protocol | Comms | V2 | Product | Issue Type | Outcome |
|---|---|---|---|---|---|---|---|---|---|---|
| f2ecc6dc-6be1-11f1-95d6-42010a62006fINBOUND | 2026-06-19 | — | 2.7 | 4 | 2 | 2 | 79.7% Developing | CONNECTIVITY | Partial restoration of node connectivity; callback scheduled for further monitoring. |