Handle unknown values without treating them as false
Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.
Hypothetical scenario: a request reaches a condition asking whether internal approval has been granted. The approval field is empty. The workflow takes the otherwise branch, marks the request rejected, and closes it. Nobody rejected the request; the workflow converted an unanswered question into a negative answer.
To handle missing data in workflow conditions, evaluate each prerequisite as true, false, or unknown before choosing an action. True means acceptable evidence supports the proposition. False means acceptable evidence disproves it. Unknown means the available information supports neither conclusion reliably.
Give unknown an explicit destination: wait for information, request review, or end without taking the proposed action. An unresolved prerequisite can prevent action now without proving that the request should be rejected.
1. Define the proposition before evaluating the field
A condition such as approval exists can conceal several questions. Does an approval record exist? Does it concern this request? Is the decision affirmative? Does it cover the current revision? These questions are not interchangeable.
Write each prerequisite as a proposition that can be supported or disproved. For example: the designated approver has approved the current revision of this request. Then specify acceptable evidence:
- True: an applicable, authoritative decision explicitly records approval.
- False: an applicable, authoritative decision establishes that approval was not granted.
- Unknown: no usable decision is available, the evidence cannot be read, or relevant decisions conflict.
An explicit decline can establish false for this proposition, but it is not the only possible basis for a false result under a defined evidence contract. Keep the actual reason rather than translating every negative state into declined.
Do not let storage format define business meaning. A missing field, an empty string, an explicit negative answer, and a failed lookup are different observations, even if an engine evaluates some of them identically. Map observations into states before branching.
2. Establish an evidence and applicability contract
For each prerequisite, document its proposition, acceptable evidence, applicability rule, evidence owner, and recovery route. Each evaluated result should retain the state, reason, evidence reference, request revision, evaluation time, and recovery owner.
The reason makes uncertainty actionable. Unknown because no decision has arrived needs different handling from unknown because a lookup failed. Asking the approver again will not necessarily fix a technical failure.
Classify unresolved contradictions as unknown and retain conflicting as the reason. Do not choose the most recently imported value unless an agreed authority rule makes it decisive. A later copy of an older decision is not necessarily a newer decision.
Separate not applicable from unknown. An explicit applicability rule may remove a prerequisite from the required set; a blank field cannot establish that the prerequisite is optional. If applicability itself is uncertain, withhold action while resolving it.
Define empty-set handling once in the specification. For this design, if no applicable prerequisites remain in either an all-required or any-route decision, withhold action and request configuration review. Only an expressly authorised no-prerequisite rule may override that default. This is a configuration hold, not an evidence-based true or false result.
3. Worked example: a request with mixed evidence
Fictional worked example: all identifiers, decisions, and operating rules below are hypothetical. A team is deciding whether to activate a service change. All four prerequisites must be confirmed. This is a recommended design, not a claim about a product control.
The filled artifact can be reused in a workflow specification by replacing the propositions and evidence while preserving the distinction between state, reason, and handling.
Request: R-214, revision B.
Proposed action: activate the service change.
Combination: all applicable prerequisites must be true.
Empty-set handling: withhold action; configuration review.
No-prerequisite exception: none authorised.
| Proposition | Applicability | Evidence | State | Reason | Handling or recovery owner |
|---|---|---|---|---|---|
| The current revision has owner approval. | Applies to revision B. | The designated owner approved revision B. | True | Confirmed by applicable approval. | Request coordinator retains the reference and rechecks before action. |
| Required capacity is available. | Applies to this activation. | The capacity owner explicitly declined allocation for revision B. | False | Allocation explicitly rejected. | Request coordinator ends this attempt without activation. Capacity owner supplies any superseding decision. |
| An implementation slot is confirmed. | Applies to this activation. | No slot decision is recorded. | Unknown | Missing evidence. | Scheduling owner supplies a decision if a live attempt needs it. |
| The receiving team accepted the handover. | Applies to revision B. | Applicable acceptance and decline records conflict; governing authority is unresolved. | Unknown | Conflicting evidence. | Handover reviewer establishes which decision governs if a live attempt needs it. |
Combined result: false.
Destination: end the current attempt without activation.
Decisive reason: capacity allocation was explicitly declined.
Unresolved context: the slot is missing and handover evidence conflicts.
Reopening trigger: a new authorised capacity decision or request revision.
On reopening: reevaluate every applicable prerequisite.
The combined result is false because one required prerequisite is disproved. The missing slot did not become a rejection, and the conflicting handover records did not become acceptance.
This distinction changes the work queue. There is no need to chase a slot merely to close an already blocked attempt. Preserve unresolved observations, but request more information only when a live decision depends on it.
If an authorised capacity approval later supersedes the rejection, the result becomes unknown while the other gaps remain. The request cannot jump directly to activation.
4. Combine states without losing uncertainty
After resolving applicability, use explicit combination rules for the remaining nonempty set.
For an all-required decision:
- True: every applicable prerequisite is true.
- False: at least one applicable prerequisite is false.
- Unknown: none is false, but at least one is unknown.
For an any-route decision:
- True: at least one applicable route is true.
- False: every applicable route is false.
- Unknown: none is true, and at least one is unknown.
These rules answer different business questions. If one accepted route is sufficient, an unresolved alternative need not block that route. Shared prerequisites must remain separate: an accepted route does not bypass an approval required for every route.
Within this three-state model, logical NOT maps true to false, false to true, and unknown to unknown. Negation does not create evidence.
Also distinguish logical negation from business labels. Not approved must not be relabelled explicitly declined without decline evidence. Even when approval is false, that alone does not prove a particular person issued a decline. A raw not-equal-to-approved comparison may also match blanks or unexpected values, so it is not a substitute for evidence mapping.
5. Give every result an operational destination
These prose diagrams continue the hypothetical example. They separate the evaluated state from what the team does next.
True: proceed to an authorisation check
All required prerequisites confirmed → true → recheck current evidence and action authority → perform the authorised activation → record execution outcome.
A true result is neither permission to act nor proof that execution succeeded. A changed request requires reevaluation.
False: end the current attempt
Applicable capacity rejection confirmed → false → record the decisive reason → end without activation → reopen only through the defined trigger.
Keep the conclusion narrow: capacity blocked this attempt. It does not establish that the requester is generally ineligible.
Unknown because information is missing: pause
Capacity confirmed; no prerequisite false; slot missing → unknown → assign scheduling owner → wait for a decision → reevaluate.
The pause needs an owner, an expected event, and a follow-up checkpoint. Repeatedly evaluating an unchanged blank is not progress.
Unknown because evidence conflicts: request review
No prerequisite false; handover records conflict → unknown → send both references to the reviewer → establish the governing decision → reevaluate.
The reviewer resolves authority rather than overwriting a field merely to make the workflow continue. Preserve the resolution and its basis.
Unknown with no viable recovery: end unresolved
Required evidence unavailable; agreed recovery exhausted → unknown → close as unresolved → no activation → retain a reopening route.
This is not rejection. The team stopped trying to resolve the attempt without establishing a negative prerequisite.
No applicable prerequisites: configuration review
Applicable set empty → check for an authorised no-prerequisite rule → if absent, withhold action and assign configuration review; if present, follow its documented destination and authorisation requirements.
6. Make waiting and review finite
Choose recovery according to cause. Missing information with a known owner generally belongs in waiting. Conflicting authoritative records need review. A failed lookup needs technical recovery rather than another question to the requester.
A hypothetical team might follow up on an unacknowledged internal request after two working days. That is an operating choice, not a universal deadline. Specify who checks, what event they expect, and what happens if it never arrives.
Elapsed time is not evidence of rejection. If an attempt ends after the chosen waiting period, record closed unresolved after waiting, not declined. If silence has an expressly authorised business meaning, document that rule separately rather than hiding it inside a blank-value default.
Avoid sending every temporary gap to a reviewer. Wait when a normal information event is expected; review when established rules cannot reconcile evidence; end without action when agreed recovery is exhausted. Each destination should make the next responsibility visible.
7. Reevaluate safely after updates
Recovery returns to evaluation, not directly to execution. A reply may resolve one prerequisite while another has changed. A revised request may invalidate earlier approvals.
- Match request and revision. Evidence for revision A must not silently settle revision B.
- Check the attempt. Late information must not revive a closed attempt unless its reopening rule permits it.
- Recheck applicability. Changed circumstances can alter which prerequisites govern.
- Prevent duplicate action. Repeated updates must not cause the same activation twice.
- Separate execution failure. A failed activation after a true result is an execution problem, not proof that a prerequisite was false.
Retain the earlier evaluation alongside the new one. Otherwise, a corrected field can erase why an attempt originally stopped, making recovery and review harder.
These are design requirements, not assurances that a specific product implements them. For an Alactic implementation, check the relevant workspace and current plan for availability; do not assume equivalent controls across CRM, Communications, ACID workflows, or web and Android interfaces.
8. Test states, destinations, and permission to act
Start with one condition where a blank currently enters the negative branch. Before enabling consequential actions, specify each test's expected state, reason, destination, owner, and action permission.
Cover explicit approval, explicit rejection, missing evidence, unreadable evidence, contradictions, an older revision, a failed lookup, and uncertain applicability. Test both combination types with an empty applicable set, checking the configuration hold and any expressly authorised exception.
For the filled artifact, replacing the capacity rejection with approval must produce unknown while the slot or handover remains unresolved. Confirming every prerequisite should reach the authorisation check. A late update after closure must obey the reopening rule, and a duplicate event must not repeat activation.
Test negation separately: unknown must remain unknown, and a false approval proposition must not acquire an explicit-decline reason without corresponding evidence.
In operation, inspect whether negative decisions have supporting negative evidence, unknown records have recovery owners, and waiting records reach their checkpoints. Keep unresolved closures distinct from rejections. These observable outcomes show whether the design preserves uncertainty instead of merely adding an unknown label while retaining the old negative branch.
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