A digital ballot rarely fails on the day it opens. It fails quietly, weeks earlier, when a rule was misread, a membership extract was imported twice, or nobody checked what happens to a second preference when the first is exhausted. By the time voting goes live, those small oversights stop being administrative irritations and become governance problems — contested results, unhappy candidates, delayed AGM business, or a resolution that has to be re-run. For company secretaries, returning officers, HR teams, union officials and club committees, the discipline of testing the ballot before launch is the cheapest insurance available. It is also the part of the project most often squeezed when timelines slip.
Ignore this line.
Build Your Test Plan Around the Rules, Not Just the Software
The most valuable testing document is not a technical checklist. It is a plain-English translation of the rules that govern your ballot. Start with the constitution, the rules of association, the enterprise agreement, the meeting notice, the scheme document or the board resolution that authorises the vote, and work through what each provision actually requires the system to do. Who is eligible and as at what date? Is it strictly one member one vote, or are there weighted votes, branch entitlements, life members or proxy holdings? Is the ballot secret, or is attribution required? Does an abstention count in the denominator? How are ties resolved? Where preferential voting applies, how are preferences distributed and what happens to exhausted ballots? Each of these is a test case, not a footnote.
Turn those answers into a test matrix with three columns: what should happen, how you will prove it happened, and who signs it off. Prove it with evidence — a screenshot of the confirmation screen, a results report, a count reconciliation — rather than a verbal assurance that the setting was enabled. Where an independent returning officer is engaged, this matrix should be agreed before configuration begins so that the person conducting the ballot, not the person building it, owns the interpretation of the rules.
Voter data deserves its own testing stream. Extract the roll from the source of truth, not from a spreadsheet someone has been editing since the last election. Test the import for duplicates, members who have resigned or passed away, joint and household memberships, expired and pending applications, and records where an email address belongs to two people. Confirm the cut-off date is enforced, so someone who joined after the qualifying date cannot vote and someone who lapsed on it still can. If your organisation operates across Australia and the Asia-Pacific, check how the roll handles members in different jurisdictions, time zones and naming conventions.
Then test the edges of eligibility and authentication. Can a credential be issued twice? What happens when a delivery email bounces, or when a member’s mobile number has changed since last year? Is there a controlled re-issue process with a log, and does it require a second pair of eyes? For enterprise agreement ballots and other regulated votes, be explicit about the majority test your organisation must satisfy and the report that will demonstrate it. Confirm the audit trail captures who configured the ballot, who amended it, and when — because after a disputed result, the record of changes is as important as the result itself.
The Dress Rehearsal: Running a Full-Scale Mock Election End to End
Configuration testing proves settings; a dress rehearsal proves the experience. Run a complete mock ballot using real data, not synthetic names, on a subset of the roll that deliberately includes awkward cases: a member with no email address, a member with two records, a member who shares a device, a member using a screen reader, a member on an older phone with a slow connection, and a member who waits until the final ten minutes. Ask a handful of genuine users — a board member, a branch delegate, an HR coordinator, a club volunteer — to vote without coaching, and watch where they hesitate. Friction that your project team no longer notices is exactly the friction that drives support calls and abandoned ballots.
Test the journey rather than the page. The notice goes out, the credential arrives, the member authenticates, reads candidate statements or explanatory material, casts a vote, receives a confirmation, and later receives a reminder. That reminder deserves particular scrutiny: if it is scheduled in the wrong time zone, or fires after voting has closed, it damages trust in the whole process. Set the ballot to close at a specific moment and verify that it does — including during daylight saving transitions on the east coast and for members in Western Australia, Singapore, Hong Kong or New Zealand who may be voting while your office is closed.
Peak load is the second rehearsal. Most votes in a typical poll arrive in the final 48 hours, and live poll voting at an AGM compresses an entire election into a few minutes in a room full of scrutineers. Rehearse that pressure: simultaneous logins, a live poll opened and closed by the chair, proxy lodgement alongside the ballot, and a scrutineer asking for a turnout figure mid-meeting. Organisations conducting online elections across several entities or branches should also rehearse switching between ballots and confirming that access permissions follow the right people.
Finally, rehearse the count. Close the mock ballot, generate the results report, and reconcile it against a manual count of the test votes by a different person. Walk through the preferential distribution round by round, confirm exhausted ballots are treated as your rules require, and check any tie-breaking behaviour. Compare the certified report, the turnout dashboard and the audit log — three views that should tell the same story. Log every defect with an owner, a severity and a retest date, and do not accept “we’ll watch it on the day” as a resolution for anything touching eligibility, counting or the audit trail.
Launch-Week Checks, Incident Playbooks and the Paper Trail That Proves It
Once testing is complete, freeze the ballot. Apply change control to the voting configuration, the candidate statements, the question wording and the voter roll, and require the secretary or returning officer to sign off the final version of the ballot paper as it will appear to voters. Any change after that point should trigger a targeted retest of the affected area — a corrected candidate statement needs a proofread and a rendering check, while a change to eligibility needs the full roll re-imported and reconciled. Keep a dated record of who approved what, so that if a candidate queries the wording, the answer is documentary rather than anecdotal.
Hold a short go/no-go meeting in the week before launch with a named decision-maker. Work through a checklist covering credentials, roll reconciliation, reminder schedule, support coverage, scrutineer access, close time, reporting outputs and the escalation path. Confirm who is on call during the voting window, including outside standard business hours if your membership spans the Asia-Pacific, and brief the frontline team with a set of approved answers to the predictable questions: where is my email, can I vote on my phone, can I change my vote, what happens if I share an address, when will results be announced.
Write the incident playbook while everything is calm. Define, in advance, what the organisation will do if a ballot question is misstated, a candidate statement is wrong, a member’s eligibility is challenged, credentials fail to deliver at scale, or turnout looks implausibly high or low. Critically, agree the position on rectification after votes have been cast: restarting a ballot is rarely simple once participation has begun, and the acceptable approach should be decided by the returning officer and the board before launch, not improvised under pressure. Where a separate paper or proxy channel runs alongside the digital ballot, document how the two will be merged and how duplicates are prevented.
During the voting period, treat monitoring as part of testing. Review delivery rates, authentication failures and turnout by cohort each day, compare against your rehearsal baseline, and watch for segments of the roll that have not engaged — that is often the first sign of a credential or data problem rather than apathy. Keep a running reconciliation between ballots cast, confirmations issued and audit log entries. When voting closes, generate the certified result, retain the full audit report and the test evidence pack, record the outcome in the minutes, and store the material according to your records policy so that the organisation can answer any challenge months or years later, with evidence rather than memory.
