Stop Two Operators From Retrying the Same Campaign Failures
Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.
Two operators open the same failed-send list. Each filters for unresolved records, prepares a retry, and assumes the other is working elsewhere. Both can follow reasonable individual procedures and still authorise overlapping sends.
Prevent overlapping campaign retry operations by making ownership visible before preparation begins: define the exact failure group, record one accountable operator, require an explicit handoff, and recheck the proposed recipients immediately before another send is authorised. A claim reserves recovery work; it does not authorise a send.
The operating procedure below uses a shared recovery-control log. It is a recommended human coordination design, not a claim that a particular product supplies these controls. Its most important rule is simple: when ownership or an earlier attempt is uncertain, pause the next send rather than infer that nobody has started.
1. Define the recovery unit before anyone claims it
A label such as “yesterday’s failures” is too vague to own. Different operators may use different filters, refresh at different times, or include records already selected for another attempt.
Give each recovery group a stable reference and preserve the exact membership at the time of creation. Record the source campaign, original attempt references, failure category, snapshot time, and recipient-record references. Use internal references rather than exposing phone numbers in a widely shared log.
Keep three sets distinct:
- Source group: the frozen set of failed records being investigated.
- Proposed retry set: the subset the operator currently wants to send again.
- Authorised retry set: the exact subset approved for a particular attempt.
Do not let a refreshed filter silently change an authorised set. If investigation adds recipients, removes exclusions, or changes the message, create a new revision and seek approval again.
Groups also need to be mutually understandable. Splitting work by failure reason is convenient, but one recipient may have several failed records. Splitting by campaign can likewise hide a recipient already being recovered elsewhere. Compare recipient membership and the underlying message purpose, not just group names. The question is whether two operators are preparing the same intended recovery, not whether their labels match.
2. Make claiming a visible, acknowledged action
Use one authoritative log, not separate spreadsheets attached to private messages. A conversation can alert colleagues, but ownership should be determined from the shared record.
For a manual workflow, appoint a recovery coordinator for the working period. Operators request a group; the coordinator checks for an existing claim or overlapping scope, records the owner, and acknowledges the claim. The operator starts preparation only after that acknowledgement. This avoids treating two nearly simultaneous “I’m taking this” messages as valid assignments.
Give each claim an owner, claim time, next update time, and current state. A useful state sequence is available → claimed → awaiting approval → authorised → attempt recorded → reconciled. Add hold for uncertainty and released for work explicitly returned to the queue.
The next update time is a checkpoint, not automatic permission for someone else to take over. If an operator misses it, the coordinator investigates. Work may already have been submitted even when the log has not been updated.
One person can coordinate and approve in a small team, but record those decisions separately. “Assigned to Priya” must never be interpreted as “Priya may send whatever subset she finds.”
3. Use a filled recovery-control log
Fictional worked example: All names, times, counts, identifiers, and outcomes below are hypothetical. The team is recovering a fictional appointment-reminder campaign. The coordinator maintains one shared log; recipient membership is held in a restricted working file referenced by the log.
The example uses a group register plus an event history. The register shows current ownership; the history preserves what happened before a handoff. A short structured record is often easier to maintain than a very wide spreadsheet.
RECOVERY GROUP REGISTER — HYPOTHETICAL
Working date: 12 May; all times use the team's declared local zone
Coordinator: Lena
Group: RG-041
Source campaign: REMINDER-MAY12
Source attempts: ORIG-501, ORIG-502
Frozen membership: SCOPE-041-v1 — 24 recipient records
Failure category: submission failure reported by campaign tooling
Current owner: Priya
Current state: reconciled
Next update: none; recovery attempt reconciled
Group: RG-042
Source campaign: REMINDER-MAY12
Source attempt: ORIG-503
Frozen membership: SCOPE-042-v1 — 11 recipient records
Failure category: status unresolved after submission interruption
Current owner: Noor
Current state: claimed; send approval absent
Next update: 11:30
EVENT HISTORY
09:05 | RG-041 | Lena acknowledged Priya's claim.
09:25 | RG-041 | Priya proposed RETRY-041-v1: 20 records.
| | Excluded: 2 opt-outs; 2 already in another attempt.
09:30 | RG-041 | Lena approved AP-041 for RETRY-041-v1 only.
| | Approved message revision: MSG-R3; one attempt only.
09:33 | RG-041 | Priya completed final gate against current log,
| | exclusions, and available attempt-status evidence.
09:34 | RG-041 | Submitted attempt reference ATT-701; 20 records.
| | AP-041 consumed; no further send authorised.
10:10 | RG-041 | ATT-701 reconciled: 18 delivered, 2 failed.
| | Remaining 2 placed on hold; no repeat approved.
09:10 | RG-042 | Lena acknowledged Mateo's claim.
09:40 | RG-042 | Mateo stopped work unexpectedly.
| | Preparation file: PREP-042-v1; no approval recorded.
| | No completed attempt reference in log.
09:45 | RG-042 | Lena marked HOLD: possible unrecorded submission.
10:05 | RG-042 | Available send history reviewed with Mateo.
| | No submission found; check evidence: CHECK-042-A.
10:10 | RG-042 | Mateo released claim; Lena acknowledged release.
| | Handoff: verify unresolved statuses before proposing.
10:15 | RG-042 | Noor accepted claim; Lena acknowledged transfer.
| | Inherited preparation only, not send permission.The crucial entry is the hold at 09:45. An empty attempt-reference field did not establish that nothing had been sent. The coordinator resolved that uncertainty before transferring the claim.
Likewise, the two remaining failures in RG-041 did not inherit approval for another retry. Approval covered one defined action, and the completed attempt reference consumed it.
4. Separate preparation from action approval
Preparation answers whether a retry is a sensible next step. Approval answers whether this operator may initiate this particular action now. Keeping them separate prevents an old approval from becoming an open-ended instruction to keep retrying.
For each proposed action, record:
- The recovery group, current owner, and exact recipient-set revision.
- The message or template revision and reason for retrying.
- The exclusions applied, including opt-outs and recipients already covered by another relevant attempt.
- The approval reference, approver, and any conditions requiring a fresh review.
- The attempt reference once submitted, or an explicit note that submission is uncertain.
Do not collapse campaign preparation, template review, and delivery into one status. For WhatsApp, recipient permission matters and opt-outs must be honoured. Approved templates are used to initiate Platform conversations and outside the 24-hour customer-service window, which opens or resets with a user message. Template approval does not guarantee delivery. Check the current WhatsApp Business Messaging Policy for precise requirements.
These messaging requirements are separate from the proposed ownership procedure. A correctly claimed group can still be unsuitable for another send. Neither a prior campaign approval nor a recovery assignment removes the need to check current eligibility.
5. Put the final check next to the send decision
A morning review cannot protect an afternoon send if someone else acts in between. The final gate belongs immediately before authorisation and execution of the next attempt, with no unreviewed changes between them.
Use a short read-back between operator and coordinator:
- Confirm ownership: Is this still the acknowledged owner of this group and revision?
- Check active work: Does another claimed, approved, queued, or submitted attempt cover any of the proposed recipients for the same intended recovery?
- Refresh exclusions: Have opt-outs, successful outcomes, or other disqualifying changes appeared since preparation?
- Resolve uncertainty: Is there an earlier submission whose outcome or existence is not yet known? If so, hold the affected records.
- Match approval: Does the approval name this exact recipient set and message revision, and is it still unused?
- Record the result: After submission, capture the attempt reference promptly. If confirmation is missing, record uncertainty rather than trying again.
For manual coordination, the coordinator should handle final authorisations sequentially and mark an approved action as committed before moving to the next overlapping request. That gives colleagues a visible warning while the attempt reference is being captured. It is a procedural safeguard, not a technical guarantee against simultaneous actions.
If the group changes or the operator is interrupted after the gate, repeat the check. A remembered “all clear” is not evidence that the shared state remains unchanged.
6. Release abandoned work without erasing its history
A clean handoff names the outgoing owner, incoming owner, remaining scope, completed actions, and unresolved questions. Preserve the old events; change the current owner only after the transfer is acknowledged.
There are three useful handoff cases:
- Prepared but not approved: transfer the working material and explicitly state that it carries no send permission.
- Approved but not submitted: cancel the unused approval and have the incoming operator obtain a fresh one after checking current conditions.
- Submitted or possibly submitted: hand over reconciliation work first. Do not authorise another send until the uncertainty is resolved for the affected records.
If the outgoing operator cannot be reached, the coordinator can reassign investigation without granting retry permission. That distinction keeps work moving without turning an abandoned claim into a duplicate attempt.
A coordinator should also be named for shift changes. Otherwise operators may follow the procedure correctly but wait on an approver who has left, creating pressure to bypass it.
7. Handle collisions and accept the trade-offs
If two active claims overlap, stop the overlapping portion of both proposed sends. Keep non-overlapping work separate only if its membership can be established clearly. Choose one owner for the contested records, invalidate conflicting approvals, and document the decision.
If an overlap is discovered after submission, do not add a compensating retry. Reconcile the existing attempts, remove affected records from pending recovery work, and record what is known versus uncertain. Review where visibility failed: unclear scope, an unrecorded action, stale approval, or an unacknowledged transfer.
This procedure introduces deliberate friction. A single coordinator can become a bottleneck; small groups create more logging; large groups tie up work that another operator could investigate. Start with boundaries the team can explain and audit. Split a group only through a recorded change that gives each new group a distinct membership and owner.
A shared log also depends on timely human updates. If the team repeatedly cannot maintain it, narrow concurrent recovery work while reviewing the operating process. Do not treat the log as a software lock or assume that a missing entry proves no action occurred.
8. Pilot the control on the next recovery session
Before the next session, name the coordinator, create the group register, and agree that no operator starts preparation without an acknowledged claim. Then require an exact-set approval and completed attempt reference for every retry.
Review the session using observable evidence:
- Were any recipients present in overlapping active recovery groups?
- Did every submitted retry have an owner, final-check entry, and matching approval?
- Were abandoned claims explicitly released or transferred?
- How many uncertain submissions remained unresolved at handover?
- Where did operators wait for coordination, and could clearer group boundaries reduce that delay?
These checks do not prove that duplicates are impossible. They show whether the team can reconstruct who owned the work, what was authorised, and which attempt actually followed. If any submitted retry lacks that chain, repair the missing control before expanding concurrent recovery work.
Put the next conversation to work.
Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.
Explore Communications