Back to all articlesCampaign delivery operations

Set a Retry Budget Before Recovery Becomes Repeated Sending

Alactic Editorial10 min read

Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.

The morning shift retries a failed campaign message. The afternoon shift exports unresolved records and tries again. A recovery run later includes the same recipient because the original campaign still shows an incomplete outcome. Each action may look reasonable in isolation; together, they create repeated sending without a shared stopping rule.

Set campaign delivery retry limits before launch: define the initial attempt, the permitted recovery attempts, the maximum elapsed time, the message’s relevance deadline, and who may approve an exception. Apply those limits to the same recipient and campaign purpose across every shift and recovery run. When any stopping condition is reached, give the record a terminal disposition rather than leaving it available for another retry.

A retry budget is a ceiling, not a target. An available attempt does not mean another send is appropriate. The worksheet below turns that distinction into an operating rule a team can use at handoff.

1. Make the budget follow the recipient, not the recovery job

A job-level retry limit is insufficient if a new job starts a fresh counter. The useful unit is the recipient’s attempt history for a defined campaign purpose. That history must remain visible when someone creates a recovery list, changes shifts, or moves an unresolved case into manual handling.

Before setting a count, define what belongs to the same budget. Include the original send and any recovery activity intended to deliver the same substantive message. Renaming a campaign, changing an export filename, or making a minor wording adjustment should not create a new allowance.

Use a shared record containing:

  • Campaign purpose and version: what the message is trying to communicate and which recipient group it covers.
  • Recipient reference: the stable identifier used to reconcile attempts across lists.
  • Attempt history: initial submission, recovery submissions, known outcomes, and unresolved submissions.
  • Budget state: remaining allowance, deadline, holds, and any approved exception.
  • Owner and disposition: who can act next and why the record is open or closed.

This can be a controlled worksheet or an authorised operational record. The important property is that all operators consult and update the same history. A separate spreadsheet for each shift defeats the design even if each spreadsheet contains a retry-limit column.

2. Decide what consumes an attempt

Define an attempt as a submission for sending, not as opening a record, preparing a draft, or selecting a recipient. Keep preparation, template review, and actual delivery separate. Otherwise, operators cannot tell whether a record has merely been checked or has already been submitted.

A practical budget can reserve an attempt when submission begins and consume it when the submission is recorded. If the submission outcome is uncertain, retain that reservation and place the record on hold. Do not treat missing confirmation as proof that nothing was sent.

For each recovery candidate, distinguish these situations:

  • Known non-delivery with a resolved cause: potentially eligible for recovery if permission, time, and budget checks also pass.
  • Known non-delivery with an unchanged cause: not ready for another attempt merely because allowance remains.
  • Unknown submission or delivery outcome: reconciliation hold, not an automatic retry.
  • Delivered: close delivery recovery. Lack of a reply is not a delivery failure.
  • Opted out or otherwise ineligible: stop campaign sending regardless of the remaining allowance.

A draft rejected before submission need not consume a delivery attempt, but it should still leave an operational note. Conversely, do not erase a real submission from the count because it failed quickly. Keeping those distinctions makes the budget explainable during a handoff.

3. Filled hypothetical retry-budget worksheet

Fictional worked example: A business is sending a notice about a scheduled maintenance period for its customer portal. Every date, time, allowance, and role below is hypothetical. The values demonstrate a policy choice, not a recommended universal retry count or a claim about any platform’s limits.

Campaign-level worksheet

  • Campaign reference: PORTAL-MAINTENANCE-NOTICE-A.
  • Purpose: inform eligible customers about planned portal unavailability on 18 September, from 18:00 to 20:00 UTC.
  • Initial attempt: one submission per eligible recipient, scheduled for 17 September at 09:00 UTC.
  • Permitted recovery attempts: up to two additional submissions per recipient for the same notice, shared across automated and manual recovery activity.
  • Total ordinary budget: three submissions, including the initial attempt. Uncertain submissions retain their place in the budget until reconciled.
  • Maximum elapsed time: 30 hours from that recipient’s initial submission.
  • Relevance deadline: 18 September at 12:00 UTC. No recovery submission after this point, even if the elapsed-time allowance remains.
  • Effective cutoff: the earlier of the elapsed-time limit and the relevance deadline.
  • Recovery eligibility: a documented recoverable problem has been resolved; no unresolved earlier submission; recipient remains eligible; appropriate message and template requirements have been checked.
  • Spacing rule: recovery is scheduled after the cause is resolved and earlier outcomes are reconciled, not immediately because the previous run finished.
  • Exception authority: the campaign operations lead may approve one additional submission before the effective cutoff. The operator requesting it cannot approve it.
  • Exception evidence: recipient reference, complete attempt history, resolved cause, reason another attempt remains useful, proposed submission time, approver, and approval timestamp.
  • Non-overridable stops: opt-out, relevance deadline, unresolved prior submission, or unmet sending requirements.
  • Exhausted-case disposition: close as “Retry budget exhausted—notice not confirmed delivered.” Exclude from further recovery for this campaign purpose.
  • Uncertain-case disposition: retain a reconciliation hold; at cutoff, close as “Deadline reached—delivery outcome unknown” if still unresolved.
  • Owner: the campaign operations lead owns closure; the receiving shift acknowledges open holds.

Filled recipient record

  • Recipient reference: CUSTOMER-EXAMPLE-014.
  • Initial submission: 17 September, 09:00 UTC; recorded as failed after a campaign configuration issue.
  • Recovery attempt one: 17 September, 13:00 UTC; submitted after the configuration correction; recorded as not delivered.
  • Recovery attempt two: 18 September, 08:00 UTC; submitted after review of a documented temporary failure; recorded as not delivered.
  • Ordinary allowance remaining: none.
  • Exception request: submitted at 08:30 UTC; declined by the campaign operations lead because there is no new evidence that another submission would have a different outcome.
  • Final disposition: retry budget exhausted; closed at 08:40 UTC; excluded from campaign recovery lists.

The declined exception matters. Approval authority is not a mechanism for making every unresolved record green. It exists to assess a bounded, evidence-backed departure from the ordinary rule.

4. Use count and time as independent stopping conditions

An attempt limit answers how much repeated sending the team will permit. An elapsed-time limit answers how long a recipient can remain in recovery. A relevance deadline answers when the message stops being useful. They solve different problems.

Use the earliest applicable stop. A recipient added late should not receive a fresh recovery period extending beyond the campaign’s relevance deadline. A recipient with unused attempts should not remain eligible indefinitely. Equally, a record can exhaust its count well before its time allowance ends.

Choose these limits from the campaign’s purpose and the team’s ability to resolve failures. A notice tied to a specific event has a different useful life from information without a near-term expiry. Neither deserves a larger allowance simply because a recovery queue is easy to run.

The trade-off is explicit: tighter limits may leave more notices undelivered, while looser limits permit more repeated submissions and longer operational exposure. Record why the chosen budget is proportionate to the message’s usefulness. Do not present an internal choice as a provider requirement.

5. Put a decision gate before every recovery submission

A recovery list should contain candidates for review, not unconditional instructions to send. Before each submission, the authorised operator or authorised workflow should check the current shared record:

  1. Does this recipient still belong to the campaign’s eligible audience?
  2. Has any earlier attempt been delivered, or is its outcome still uncertain?
  3. Is there an ordinary attempt available or a valid, unused exception?
  4. Are both the elapsed-time limit and relevance deadline still open?
  5. Has the reason for the previous failure changed enough to justify recovery?
  6. Are permission and applicable messaging requirements satisfied now?

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.

A recipient reply may change the service-window context, but it should not silently reset the campaign retry budget. If the reply needs individual assistance, move it to the appropriate human handling path rather than treating it as permission to repeat the campaign notice.

Recheck eligibility immediately before submission, not only when the recovery list is assembled. A list can become stale while awaiting review.

6. Make exceptions narrow, visible, and consumable

An exception should identify a specific recipient, campaign purpose, additional allowance, expiry, and approving person. “Retry remaining failures” is not sufficiently bounded: it can be reused across runs and interpreted differently by the next shift.

Require the requester to explain what changed. A corrected configuration or reconciled submission history may support another review. An unresolved target or pressure to clear a dashboard does not establish that another attempt is useful.

Keep exception authority separate from routine execution where practical. After approval, record whether the exception was used, expired unused, or withdrawn. An approval must not remain available after it has funded a submission.

If several operators can act at once, designate one recovery owner or use a shared claim process so the same remaining allowance is not spent twice. This is a recommended control design, not an assertion that a particular product provides it. Where shared enforcement is unavailable, reduce concurrent handling rather than relying on operators to notice duplicate work afterward.

7. Close exhausted cases without hiding uncertainty

“Failed” alone is a poor final state. It does not tell the next operator whether another attempt is allowed, whether the message expired, or whether delivery is still unknown. Use terminal outcomes that explain both the result and the permitted next action:

  • Delivered: no further delivery recovery for this campaign.
  • Budget exhausted: no ordinary allowance remains; close and exclude from recovery.
  • Relevance or elapsed-time cutoff reached: close even if attempts remain.
  • Recipient ineligible or opted out: stop campaign sending and preserve the applicable exclusion.
  • Unresolved outcome at cutoff: close recovery with uncertainty recorded; continue reconciliation separately if needed.

Closing recovery does not require pretending the outcome is known. If late evidence later confirms delivery, update the outcome history without reopening permission to send.

If the team discovers that overlapping runs exceeded the budget, pause the affected recovery activity, reconcile all attempt histories, and close ineligible records before resuming. Do not respond by deleting duplicate rows or relabelling attempts. The failure to investigate is the loss of a shared boundary, not merely an untidy list.

8. Test the boundary at the next handoff

Before the next campaign launch, complete the worksheet with the campaign owner and the person responsible for the following shift. Ask the receiving operator to decide whether each open case can be retried using only the shared record. Any need to search private messages for approval or reconstruct an attempt count is a gap to fix before recovery begins.

For the next completed campaign, review observable outcomes: recipients submitted beyond their approved allowance, submissions after cutoff, uncertain records resubmitted before reconciliation, exceptions used without recorded approval, and exhausted records that re-entered a recovery list.

Also inspect unresolved cases closed without delivery. That reveals the consequence of the chosen budget and supports a deliberate change for a future campaign—not a retrospective expansion of the current one. The immediate goal is a boundary the next shift can follow: every submission has remaining authority, and every exhausted case has somewhere final to go.

Campaign operationsRetry budgetsDelivery recoveryWhatsApp

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