# Workshop feedback report

Use the completed fictional example to check a report, then fill in the blank card for your own survey. The companion CSV and question schema define the example; this is not a platform export or an automatic importer. No real people or responses appear here.

## Completed example

**Question version:** workshop-feedback-v1. **Cutoff:** November 9, 2026, 17:00 UTC. **Scope:** twelve invitation records, I01–I12. Each invitation ID identifies one distinct fictional participant here. Real repeat invitations or responses need a declared respondent-identity rule before reporting unique people. Delivery and recipient eligibility are unknown. **Inclusion:** use supported captured answers from submitted and partial sessions. A final Submit event does not establish which optional questions were answered.

| Submission evidence | IDs | Count |
| --- | --- | ---: |
| Final Submit captured | I01, I02, I03, I04, I05, I06, I08, I09 | 8 |
| Partial progress | I07 | 1 |
| No captured response | I10, I11, I12 | 3 |

These counts total twelve. Do not label 8/12 a standardized response rate: successful invitation delivery and recipient eligibility were not established.

### Q1: Did you attend the workshop?

Yes: I01–I07 (7). No: I08–I09 (2). Unknown: I10–I12 (3). I07's captured answer counts even though its session is partial.

### Q2: How useful was the workshop for your next task?

Only attendees are eligible. The rating scale is 1 Not useful, 2 Slightly useful, 3 Moderately useful, 4 Very useful, 5 Extremely useful.

| State | IDs and values | Count |
| --- | --- | ---: |
| Answered | I01=5, I02=4, I03=4, I04=3, I05=2 | 5 |
| Explicitly skipped | I06 | 1 |
| Not reached | I07 | 1 |
| Not asked by branch | I08, I09 | 2 |
| Unknown | I10, I11, I12 | 3 |

**Result:** mean usefulness among five answering attendees = (5+4+4+3+2)/5 = **3.6 out of 5**. Rating distribution: 1→0 people, 2→1, 3→1, 4→2, 5→1. Seven known attendees comprise five answers, one explicit skip and one not reached. The other five invitations comprise two known non-attendees and three unknowns. Never fill their absent ratings with zero.

### Q3: What would you like offered next?

Offered to any respondent, including non-attendees. Choose any positive options or explicitly choose none.

| Answer or missing evidence | IDs | Count / denominator |
| --- | --- | --- |
| Guided practice | I01, I02, I05 | 3/7 answering people |
| More worked examples | I01, I03, I08 | 3/7 answering people |
| Session recording | I03, I05, I09 | 3/7 answering people |
| Explicit none | I04 | 1/7 answering people |
| Explicitly skipped | I06 | 1; excluded from answer denominator |
| Not reached | I07 | 1; excluded from answer denominator |
| Unknown | I10, I11, I12 | 3; excluded from answer denominator |

**Answer denominator:** I01, I02, I03, I04, I05, I08, I09 = seven people. Six chose at least one positive option; one chose none. The positive options contain nine distinct person–option pairs. Each option is 3/7 = about 42.9%. Their exact combined fraction is 9/7, so summing those percentages exceeds 100%. Do not add the option rows to count people. An exact repeated pair still counts once; retain a note that duplication was found.

**Decision:** the three positive topics tie in observed selections. Offer a choice; do not call any topic the preference of all invitees. If comments were collected separately, review them for follow-up. This example contains no comments.

## Translate source states before calculating

| State | Evidence required | Treatment |
| --- | --- | --- |
| Answered | Supported value or positive choices | Include in that question's denominator |
| Explicit none | Deliberate none choice and no positive choices | Include in Q3 denominator; no positive pair |
| Skipped | Question shown and explicitly bypassed | Report separately |
| Not reached | Captured progress ends before eligible question | Report separately |
| Not asked | A known branch excludes Q2 | Report separately |
| Unknown | Insufficient evidence for a narrower state | Preserve as unknown |

The fixture supplies states explicitly. A blank value, a generic skipped label or no final submission does not prove one of them. If your export cannot distinguish them, document that limitation instead of inventing exposure.

**Resolve before reporting:** duplicate invitation rows; unknown options; none combined with a positive choice; values under skipped/unknown states; a rating outside 1–5; a rating from someone marked did-not-attend. Keep the original data and the correction evidence. Do not silently choose the convenient value.

## Blank report card

Survey/event: ______. Question version: ______. Reporting cutoff/timezone: ______. Source file or captured revision: ______. Reviewer: ______.

Respondent identity / repeat-response rule: ______.

Invitation records: ______. Delivery evidence: ______ / unknown. Recipient eligibility evidence: ______ / unknown. Submitted: ______. Partial: ______. No captured response: ______. Inclusion rule for captured partial answers: ______.

Copy this table for each question. Add a separate state if your source needs it; do not force unknowns into a known category.

| Field | Fill in |
| --- | --- |
| Exact question and choices | ______ |
| Who is eligible; how that is known | ______ |
| Answered IDs and values | ______ |
| Explicit-none IDs, if allowed | ______ |
| Skipped IDs and evidence | ______ |
| Not-reached IDs and evidence | ______ |
| Not-asked IDs and branch evidence | ______ |
| Unknown IDs and missing evidence | ______ |
| Denominator definition and value | ______ |
| Numerator / value sum and contributing IDs | ______ |
| Distinct person–option rule, if multiple choice | ______ |
| Result with units and scope | ______ |
| Contradictions or duplicate evidence | ______ |
| What this supports changing | ______ |
| What it does not establish | ______ |
| Next check, responsible person and due date | ______ |

Check that all record IDs appear in the appropriate question-state inventory. Positive checkbox option rows may overlap; state partitions must not. Keep zero answers as “no observed answer” with no mean or percentage, not 0% satisfaction.

## Brief the app you need

Use [the Overskill brief builder](https://www.overskill.com/learn#brief-builder). Describe the question version, explicit states, report calculations and who may see individual answers. Edit the input fields and review the prompt, then choose **Start build**; **Copy brief** is secondary. The link imports no files. Supply this fictional setup separately, and test the calculations and access rules before using real responses. Account access and credit requirements apply when building.