# Example QA scorecard: evidence, score and coaching

Original Culvion Technologies example · 7 November 2026

ILLUSTRATIVE ONLY. The interaction, reviewer judgments and score are fictional. This is not a Culvion customer result or an industry benchmark.

## The fictional interaction

Context: a customer asks by email about a delayed order. The example playbook allows checking the order status and requesting a fulfilment review. Refund exceptions need client approval. An outbound update arrangement should be recorded for an unresolved case.

Customer: “The delivery date has passed. Can you tell me where the order is?”

Agent: “I’ve checked the order record. The latest delivery information is incomplete, so I cannot confirm a new date yet. I’ve asked the fulfilment team to check and will update you by email.”

Fictional case note: “Order status checked. Fulfilment review requested. Customer informed.”

Evidence available to the fictional reviewer: the request, agent reply, order-status check and escalation record. The escalation contains the reason, impact and work already done. The customer-update time is missing from both the reply and the case note.

## Scoring method

Each category has a maximum number of points. The reviewer awards points against the observed criteria. The five earned values are added directly; no further weighting is applied.

| Category | Maximum points | Earned points | Evidence and judgment |
| --- | ---: | ---: | --- |
| Accuracy | 30 | 24 | The agent uses the available status and avoids inventing a delivery date. The reply could distinguish the last recorded event more precisely. |
| Ownership | 25 | 20 | A fulfilment review is requested and an update channel is stated. The next update time is missing. |
| Communication | 20 | 18 | The answer is clear and candid. A brief acknowledgement of the delay would improve the response. |
| Case record | 15 | 12 | The note records the check and referral. It omits the customer-update arrangement and the detail needed to see that arrangement quickly. |
| Escalation | 10 | 10 | The separate escalation record identifies the reason, impact, work done and receiving role. |
| TOTAL | 100 | 84 | 24 + 20 + 18 + 12 + 10 = 84 |

Result: 84 / 100 = 84%. The score is a worked example, not a recommended pass threshold.

## Critical checks are separate

For this example, a critical issue would include an unauthorised refund, disclosure outside the agreed access rules or a fabricated delivery promise. None is observed in the fictional evidence. A high numeric score would not cancel a critical issue if one were present.

Define the critical criteria, review evidence and action with the client before using this method. This sample does not certify compliance or guarantee that every risk has been covered.

## Coaching that follows from the evidence

1. Agree the next update time or the route for obtaining it before closing the reply.
2. Put the update owner, channel and timing in the case note.
3. Acknowledge the customer's missed delivery expectation without guessing the outcome.
4. Keep the accurate status check and the complete escalation evidence.

## Example improved response structure

- Acknowledge the question or impact.
- State the information that was checked and what is still unknown.
- Explain the action taken and the receiving owner.
- State the actual agreed update channel and timing.

Fill that final step with a real arrangement. Do not replace a missing commitment with a made-up deadline.

## Calibration prompts

- Would two reviewers award the same points from the same evidence?
- Which criterion is ambiguous and needs an example?
- Is the gap an agent behaviour, a missing rule or an unavailable tool?
- What later evidence would show that the coaching helped?

The weekly-report sample uses a synthetic aggregate average of 84%. That aggregate is a separate illustrative data set; it is not inferred from this single interaction.