Sopact Academy · Synthetic practice workbook

Turn member data into a useful answer

Choose the example that fits your work. Each path takes you from source records to a checked answer, a useful report and a small pilot.

Practice data only. These records do not describe a real customer. Do not paste private member information into this workbook. It works locally: answers stay in the page until you save them; refreshing closes unsaved work.

Your question: Which members attended the 2026 conference, completed a certification in 2026, and explicitly described an unresolved concern about value for dues?

Your question: What can we report from Q2 member-organization returns, where do we need clarification, and what useful summary will each organization receive?

About 20–25 minutes per step is a study estimate. Complete one step at a time. You need a spreadsheet or text viewer for the CSV files; no platform purchase is required.

Coordinating a network of member organizations? Choose the network alternative within each step, using network-returns.csv and network-feedback.csv. Its question is: what can we report from our Q2 member returns, and what useful summary will each organization receive? You do not need to complete both examples.

Open pack guide

Save your answers before refreshing or closing this page.

Step 1 of 5

Choose the question and keep what works

Start with the question above. Decide what a human will do with the answer. List the smallest set of records needed; do not begin by replacing the member system.

Separate a reported concern from a prediction of non-renewal.

Name a human decision or conversation. Avoid claiming that the data tells you why someone will leave.

Use 01-members, 03-conference-attendance, 04-certifications and 05-renewal-feedback. For each: record needed | source owner | identifier | date | permitted use | gap.

Agree the difference between unique participants within one organization and attendance occasions. Map who supplies N01/N02/N03 returns, who reviews them, the Q2 period and what useful local summary each organization should receive.

Question | intended decision | unique participant definition | attendance occasion definition | Q2 sources | owner | expected useful return. Use network-returns.csv; treat N03's missing participant count as unknown, not zero.

Ready to continue when: your question has explicit conditions, an intended use and a source for each condition.

Step 2 of 5

Keep records connected to the right member

Use 02-identity-map to connect source IDs to the member roster. Keep the original row IDs beside every joined record. Check attendance for repeated rows. Keep anonymous R5 separate.

Conference attendees | certification completions | named feedback. Note the source rows you used and any duplicate rows.

Explain how you avoided double-counting and distinguish a missing record from evidence that something never happened.

Can you count its theme? Can you identify its respondent? Does it prove a sixth distinct member responded?

Map N01/N02/N03 as stable organization IDs, with separate representative and reviewer roles. Explain how a new representative can take over without erasing the organization's prior returns. The source provides no person-level participant IDs, so do not deduplicate people across organizations.

For each organization: stable ID | Q2 return source | current representative role | reviewer role | permissions to review. Show what stays when the representative changes and how a correction retains earlier evidence.

Ready to continue when: you can trace every joined row and explain why the anonymous response remains unlinked.

Ready to continue when: organization identity, representative roles and dated returns remain distinct, and a representative change will not erase the record.

Step 3 of 5

Review comments without losing their meaning

Read 06-comment-codebook before labeling the six comments. A response can have several themes. Record sentiment for each theme; do not allow a positive point to erase a specific concern. Use uncertain whenever the wording does not justify a judgment.

For each R1–R6: theme(s) | sentiment by theme | explicit unresolved value concern yes/no/uncertain | exact words | review note.

Explain what a single overall score would miss.

Name the qualifying fictional member IDs and cite attendance, certification and comment rows. Explain what this answer cannot predict.

Review NF1–NF3 for access and scheduling themes. Keep each organization summary separate from participant-level testimony. Mark the uncertainty in NF3 and write the next clarification question. Do not turn three organization summaries into a count of affected people.

Source ID | reported issue | exact excerpt | what is uncertain | next question | owner. Distinguish an organization's summary from a participant's own words.

Ready to continue when: your findings preserve mixed feedback, distinguish response counts from member counts, and cite the exact evidence.

Ready to continue when: each finding points to an organization summary, preserves uncertainty and names the next clarification or action.

Step 4 of 5

Make the report useful to its reader

Continue with the same member records. Turn the P101/P102 finding into a short brief for a colleague authorized to review member experience. Keep R2's mixed meaning, show the source records and explain the anonymous and missing feedback. You do not need a new dataset to finish this step.

Question and period | P101/P102 findings and source IDs | mixed feedback | anonymous and missing records | what the answer cannot establish. Keep R5 unlinked.

Name the authorized audience, the person who will check interpretation and the next decision or conversation. Explain what detail would need to change for a broader member-facing report.

Optional extension: create a benchmark from a separate dataset

Use 07-chapter-benchmark only if a peer benchmark fits your work. This separate aggregate represents a larger fictional population; it cannot be reconstructed from the six-member extract. It is not required to complete the core member report. The exercise releases chapter results only with at least 10 usable responses. That illustrative rule alone does not guarantee privacy or reliable comparison.

Coverage = usable / eligible. Value rate = reports value / usable. Record each denominator. Which chapter result must remain suppressed?

Use summed response counts, not an unweighted mean of percentages. State the included eligible population, usable responses and reports of value.

Include period, cohort, numerator, denominator, coverage and a limitation. Do not publish excluded C totals beside A+B totals if they could reveal the suppressed result.

If considering a paid report, write two questions you would ask intended readers to test usefulness and willingness to pay. Do not assume buyers exist.

Draft two views: a useful summary for N03 showing its 12 attendance occasions and missing participant count, and a network summary showing 70 reported participant counts across two of three organizations. Explain why the latter is not deduplicated network reach. If using the separate A/B/C benchmark, treat chapters as reporting groups that may be member organizations only when definitions and populations align.

What was submitted | what is known | what needs clarification | next action | who can help. Give N03 something useful before all fields are perfect.

Report participant counts across the two contributing organizations, attendance occasions across all three, and organization coverage. Do not describe 70 as deduplicated network reach. State what may be shared and who approves the report.

Ready to continue when: a reader can tell who is represented, who is excluded, what the number means and how it might be used.

Step 5 of 5

Test the workflow before expanding it

Use the same source files and requirements with the tools you already have. A small export-based pilot can reveal what needs configuration, integration or another platform. Assign a person to each unresolved issue.

Source owner | reviewer | membership lead | member reader. Who can see identities and comments, change mappings, correct a record, approve a report and receive grouped results?

1 Duplicate A6. 2 Mixed R2. 3 Anonymous R5. 4 Incorrect/changed identity mapping. 5 Denied person-level access for an unauthorized reader. 6 A traceable P101/P102 brief for the authorized owner. Optional benchmark extension: A+B report without a disclosure of C. For each: expected | actual | evidence | owner | next action.

Name unresolved gaps, what must change, who owns it and when you will review again. Record actual test results, not just intended behavior.

Test a member-organization submission, unclear definition, representative change, correction and returned report. Record who receives the local summary and whether it is useful. The pilot is unfinished until members get something useful back.

Submission | unclear definition | representative change | correction | returned local summary. For each: expected result | actual result | evidence | owner | next action.

Record the local summary recipient and date, what was useful, unresolved gaps and the next review point. A successful collection without a useful return is not a completed pilot.

You are finished when: the answer is traceable, the report has clear limits, and each material gap has an owner before wider use.