• How to Structure Stakeholder Data: Four Common Patterns

  Choose how organizations, people and events connect.

• How to Collect Feedback Offline and Keep Records Connected

  Plan field collection and reconcile submissions when connectivity returns.

• Prepare data governance before using AI

  Before using identified or restricted evidence, agree ownership, permitted uses and a small test plan.

• How to Map and Prioritize Stakeholders

  Map affected groups and evidence contributors, including voices missing from the current contact list.

The full guide is preserved here. Use this detail when it helps your plan, then continue with the exercise.

How do you run a survey across a member network?

Start with an agreed member list, a clear reporting purpose and a useful result to return to each member. Keep the organization separate from the person answering, define the reporting period and specify who can see each field. Pilot the collection and return process, track missing submissions, then share findings with their coverage and limitations. A completed form is one part of a continuing relationship.

This lesson is for associations, federations, partner networks and multi-location organizations collecting recurring updates. You will produce a member register, a submission tracker and a sample member report. Use this guide to complete the collection plan in the first Membership & networks lesson. The later lessons build the review, history, comparison and governance steps on that plan.

1. Define which organizations belong in this reporting round.
2. Agree the information members need back.
3. Separate member identity, respondent access and dated submissions.
4. Test the questions, supporting documents and access rules.
5. Monitor complete, partial and missing submissions.
6. Return approved findings and record the next action.

Design the member report before the questionnaire

Ask what a member should be able to do with the results. Possibilities include reviewing its own history, identifying a support need or comparing a defined measure with an appropriate peer group. Choose a concrete use rather than promising a dashboard full of every collected field.

Sketch a one-page report: the member's current information, what changed, what remains unconfirmed and one discussion question. Show it to a few representatives before finalizing collection. If the report would not help them, reconsider what you are asking them to supply.

Returning useful information may make participation more worthwhile, but it does not guarantee responses. Time, trust, accessibility, survey burden and the relevance of the request also matter. AAPOR's survey best practices recommend clear questions, pretesting and attention to collection methods. Use those basics alongside the member report, rather than expecting one incentive to fix every participation problem.

An operational update and a confidential opinion survey may need different arrangements. Do not imply that comments are anonymous while also promising a named account history containing every response. Explain the collection purpose and access model before someone answers.

Separate the member from its representative

A member organization can change its name or representative without becoming a new organization. Give it a stable identifier. Store each representative's identity, role and authorized access separately, with start and end dates.

Keep submissions dated and linked to the member. Preserve who submitted or approved each update, the period it describes and the form version. This lets a replacement representative see the organizational history they are authorized to access without inheriting the former representative's personal account.

A merger, split or membership lapse needs an explicit decision. Retain the history and record the relationship between old and new entities. Reusing an old ID for a different organization would make a trend misleading.

Can members use different surveys and still compare results?

Yes, when the comparison is designed around a small common core. Agree which questions the network must answer together, then identify the few shared fields needed for those questions. Keep local questions where they serve a member’s own work. One questionnaire should not be imposed simply to make a central spreadsheet easier to assemble.

Start with the continuing member or registration record. Reuse stable information where appropriate. If location, role or another profile field changes, record the effective date and choose a suitable confirmation cycle. Keep that profile history separate from the annual return, event survey or research observation.

Then use a versioned data dictionary to define and map the comparable fields: unit, population, period, allowed values, missing states and calculation. A local visits field can map to attendance occasions when their meanings match. It cannot become unique participants unless the underlying evidence supports removing repeated attendance.

Before aggregating, check member coverage and overlap. Keep incompatible measures separate and show where evidence is missing. The benefit of a connected Sopact workflow is preserving local evidence and history while giving shared definitions to the analysis; the dictionary cannot manufacture a missing observation.

Collect only what serves the reporting purpose

Separate relatively stable profile information from period-specific metrics and feedback. Let a representative confirm an existing profile where appropriate, rather than repeatedly typing the same address or service description.

For a quarterly partner update, the collection might contain a metric, a short explanation, a financial document and a social-audit attachment. Specify the reporting period and unit for each. A figure of “250” is not usable until the reader knows whether it means people, visits or another measure.

Use conditional questions where they genuinely remove irrelevant work. Provide appropriate missing or not-applicable responses. Test whether members interpret the questions as intended, including in the languages and formats they will use. Avoid requiring sensitive personal details merely because the form supports another field.

For uploaded evidence, ask what the reviewer needs from the document. Record the file's source and period. If AI extracts a value or flags an inconsistency, retain the cited source location and a review status. An extracted draft value should not silently become an approved network result.

Choose annual collection or continuing updates deliberately

Continuing updates can suit changing profiles; a defined survey window can suit a shared reporting period. You can combine them: maintain profile information throughout the year and collect a specific quarterly or annual submission separately.

Distinguish three dates: when an event happened, the period a response describes and when someone entered it. A September snapshot of the database may contain information last confirmed in March. The snapshot date alone does not make every record current or comparable.

Set a confirmation deadline and identify stale records. Preserve the issued snapshot and log later corrections. If a member revises a prior number, record whether the published report will be restated or the correction will appear in the next cycle.

Keep definitions and important measurement choices stable where comparison requires it. When they change, document the break or a justified mapping. Revisit changing survey questions without losing comparability before carrying a revised question into the same trend.

Decide who can see each kind of information

Write an access policy for the actual data, rather than giving everyone at the center unrestricted access. A representative may need their organization's operational returns but not another person's confidential comments, restricted case information or internal investigation notes.

Specify who may submit, approve, amend, export and analyze each category. Separate those permissions where the work requires it. Decide what happens when a representative leaves, a member withdraws or a temporary analyst's work ends.

Benchmarks also need a disclosure review. Removing names does not necessarily prevent identification: a distinctive location, size or result may reveal a member. Where information relates to people, the ICO's anonymisation guidance explains why indirect identification and information available elsewhere matter.

For a network report, consider whether groups are large and comparable enough to share, whether extreme values reveal an organization and whether combinations of filters expose a hidden result. Use broader groups, suppress unsafe comparisons or omit a benchmark when necessary. There is no universally safe group-size rule.

Test the same boundaries in downloads and AI answers, not just the dashboard. Use the AI data-access lesson for an allowed-and-denied test. A written instruction to the assistant is not a substitute for enforced access controls.

Choose a census or sample based on the question

If every member needs its own current record or support review, invite every eligible member. That is an attempted census of the defined network. It still may have missing responses, outdated contact details or measurement problems.

A designed sample can be appropriate when the objective is estimating a network-wide characteristic rather than updating each member's record. Define the population, selection method and intended inference. The choice is not determined by an arbitrary rule that networks above or below a certain size must use one approach.

Do not describe whoever happens to respond as a representative sample without supporting that claim. Equally, do not assume all nonrespondents have the least capacity. Investigate what you can learn about missingness from the member register and follow-up, without inventing their answers.

AAPOR's Standard Definitions distinguish response information from evidence about nonresponse error. A response rate helps describe collection; by itself it does not establish whether estimates are biased. Keep both the rate and the operational list of unresolved submissions.

Worked example: report completion and item coverage separately

The following network and numbers are fictional. There are 40 eligible member organizations in this round. Twenty-eight complete the agreed core questions, eight submit a partial return and four provide no response. A valid answer to one capacity question is available from 32 members.

Report: “Twenty of the 32 members answering the capacity question selected this support option. The question had usable answers from 32 of 40 eligible members.” Do not turn 62.5% of answers into a claim about all members.

These are transparent operational measures for this exercise, not an assertion that each is a standardized AAPOR response-rate formula. Define the eligibility and completion rules beside them. If you need a formal research outcome rate, use the appropriate method and disposition categories.

For follow-up, identify which partial submissions lack the specific information needed. A member may have answered the capacity question but not uploaded a document. Sending a generic “you have not responded” message would be inaccurate and unhelpful.

Build a useful member result without an unsafe ranking

Start with the member's own approved information: this period's result, the prior comparable result and any unresolved evidence. Add a comparison only when its population, measure and disclosure treatment are defensible.

For example, a fictional report for Member M-014 might state: “Your submitted count is 140 service visits this quarter, compared with 125 last quarter under the same definition. Your financial attachment is awaiting review. The network comparison is withheld because the comparable subgroup is too small under our release policy.”

That is still a useful report. It provides continuity and a next action without pretending every member must receive a percentile rank. Where a benchmark is appropriate, show its period, contributing members, unit and missing-data treatment.

Keep narrative feedback near the measure it helps interpret. Review comments for confidential details before sharing them. Do not use another member's distinctive story as an “anonymous” example without checking whether the audience can recognize it.

Pilot the whole cycle, including representative turnover

1. Use three fictional members: create two reporting periods and different submission statuses.
2. Submit mixed evidence: include a metric, narrative and document with a deliberately mismatched period.
3. Replace a representative: revoke the old account and give the new person only the approved organizational access.
4. Test isolation: try another member's URL, export and assistant query from the test account.
5. Check completeness: distinguish an absent submission from a missing answer or unreviewed attachment.
6. Return the report: ask the representative to interpret it and identify their next action.
7. Reproduce a snapshot: make a later correction and confirm the prior issued view remains explainable.

Document failures and their owners before expanding. A correct survey followed by an incorrect export is a failed workflow. Conversely, an approved manual export may be a reasonable first implementation; automation is not the only valid way to return results.

Where a connected workflow can help

You can begin with a controlled member register, a form, a submission table and individually reviewed reports. Keep evidence references and access decisions explicit. Assess the actual coordination effort rather than assuming that a fixed number of members makes manual work impossible.

As the process grows, evaluate whether your setup keeps new responses and uploads attached to the correct member and period, prepares analysis for review and returns an authorized view without recreating the reporting logic every cycle.

For Sopact, bring the pilot above to an evaluation. Ask to demonstrate the configured collection, document analysis, continuing record and member-specific reporting. Verify the access, export and integration arrangements your network needs. Your team still approves the definitions, resolves flagged inconsistencies and decides what can be shared.

Watch: why qualitative evidence needs a continuing record

Watch the video · 2 minutes 34 seconds. This companion introduces connected qualitative evidence; it is not a demonstration of every network permission described here. Browse more videos in the video library.

Frequently asked questions

How do we begin a member-network survey?

Define the member list, reporting purpose and result each member will receive. Separate the organization from its representative, set the period and access rules, then pilot collection through reporting. Track partial and missing returns as well as completed ones before interpreting network findings.

Will returning results improve participation?

It may make the request more useful, but it is not a guarantee. Test what members need, reduce unnecessary work and address access or trust problems. Monitor participation across rounds and ask nonrespondents about barriers rather than assuming one reason explains every missing response.

Should the survey remain open all year?

That can work for profile updates. Period-specific metrics may still need a defined collection window. Record when information was confirmed and the period it describes; a snapshot taken today does not make an old answer current. Preserve issued snapshots and explain later corrections.

Can we safely show members their network rank?

Only after checking comparability and disclosure risk. An unnamed rank or small subgroup can still reveal another organization or person. A member's own history can be useful without a benchmark. Withhold or broaden comparisons when the proposed release would expose restricted information.

Is sampling ever appropriate for a member network?

Yes, when a designed sample suits the research question and intended inference. If the task requires a current record for every member, invite every eligible member instead. Neither approach removes the need to examine missing responses, coverage and the quality of the measures.

What happens when a representative leaves?

Keep the organizational identifier and dated submissions. Revoke the former representative's access and authorize the replacement according to the policy. Do not transfer a personal login or automatically expose confidential material just because the new person represents the same organization.

Can AI prepare each member's report?

It can assist in a configured workflow using permitted evidence, clear definitions and source references. Check that it cannot retrieve another member's restricted information, and review extracted values and narrative claims. Human owners remain responsible for the release and unresolved inconsistencies.

Bring one tested cycle back to your course

Return to your membership module with the register, example submission, issued report and test log. Use them to test collection, context, continuity and access in your workflow plan.

Reviewed September 12, 2026. All member identifiers, counts and reports in the exercises are fictional.