Status and Enum Dictionary
This page captures the most important status, enum, and flag fields that appear repeatedly across BriteCore data.
The goal is to reduce the chance of misreading the same concept under different names across CSV, API, and SQL/report layers.
Why this matters
The same business state can appear in several ways:
- as a raw CSV field like
policyStatusorstatus - as an API field like
status,policy_status, orrevision_state - as a SQL/report field such as
claim_status,policy_status, ordeleted
These fields often have different names across layers, but the underlying semantics are very similar.
Core status families
Policy and revision status
Common concepts:
- active / inactive
- in force / canceled / expired / rewritten
- renewal status
- revision state and term state
Typical field names:
- CSV:
policyStatus,renewalStatus,cancelDate,newOrRenewalTerm - API:
status,policy_status,renewal_status,cancellationdata - SQL/report:
policy_status,revision_state,term_effective_date,term_expiration_date
Interpretation rule:
policy_statusdescribes the policy lifecycle staterevision_statedescribes the revision or term workflow statecancelDateand cancellation reason are usually historical/operational facts, not the same as a simple current-state status flag
Claim status
Typical field names:
- CSV:
status - API:
status - SQL/report:
claim_status,claim_active_flag
Interpretation rule:
claim_statusis usually the operational lifecycle stateclaim_active_flagis a boolean-ish signal used for current-versus-historical filtering- status text may include historical or conversion states, so do not assume a narrow modern enum without checking sample data
Contact status
Typical field names:
- CSV:
active,terminated,terminationReason,flag - API:
is_agent,terminated,termination_reason,system_tags - SQL/report:
deleted,agent_terminated,termination_reason,roles
Interpretation rule:
deletedis often a soft-deletion or inactive flag, not always the same asterminatedterminatedandagent_terminatedoften reflect a business status, not a system-delete staterolesmay carry a large amount of semantic status/context beyond a simple boolean flag
Common boolean and flag patterns
These fields recur often and are usually easy to misread:
deleted/is_deleted/deleted_flagactive/is_active/active_flagterminated/is_terminatedrewrittenhistoricalsuperuser_flagterms_conditions_acceptededelivery_enabledfollow_agency_quoting_restriction
Rule of thumb
deletedusually means a soft-delete or inactive record in report logicactiveusually means currently active in business logicterminatedusually means business termination, not necessarily database deletion- when a table is historical or reporting-oriented, the flag may mean “currently visible in this grain” rather than “this record has been permanently removed”
Role and tags vocabulary
Role fields
Common examples:
rolescontact rolesNamed InsuredMortgageeClaimantAgentAgency
Interpretation rule:
- role data is often denormalized into text in the report layer
- the raw relationship tables may hold the actual role assignments separately
- the report view may store these as comma-delimited strings rather than normalized rows
System tags
Common examples:
systemTagspolicySystemTagsclaimSystemTagsquestionsSystemTags
Interpretation rule:
- these often behave like a free-form metadata bag
- they are useful for operational filtering but not always clean enough for strict report logic
- do not treat them as canonical master data unless the schema or API docs explicitly say so
Date and lifecycle semantics
Date fields that matter most
dateAdded/date_addeddateUpdated/date_updatedeffectiveDate/term_effective_dateexpirationDate/term_expiration_datecancelDatedateReportedlossDatecommitDate/commitDateTime
Interpretation rule:
dateAddedanddateUpdatedare operational metadataeffectiveDateandexpirationDateare term or coverage lifecycle datescancelDateandterminationReasonare often historical lifecycle indicatorslossDateanddateReportedare claim event dates, not necessarily policy-term dates
Common status confusion patterns
Policy and revision confusion
A common mistake is to interpret the latest policy row as the same thing as the current revision or the current term.
The correct mental model is:
- policy number identifies the policy
- revision captures the term or change iteration
- a policy can have many revisions over time
Claim confusion
A claim can have many lifecycle and status changes, and the claim table may carry both operational status and historical content. Do not assume a single status field tells the full story.
Contact confusion
A contact may be active in one context and inactive or terminated in another. A single contact record can carry both role and status information applied across multiple policy or claim relationships.
Recommended reporting rule
When you are writing business logic, use this sequence:
- identify the business state you want
- determine whether it is a master-object status or a relationship status
- determine whether it is historical or currently active
- validate against both the raw CSV and the report view before assuming semantics
Short reference list
| Concept | Typical naming patterns | Meaning |
|---|---|---|
| Active | active, is_active, active_flag |
currently active / enabled |
| Deleted | deleted, is_deleted |
soft-delete or inactive status |
| Terminated | terminated, agent_terminated |
business or user termination status |
| Cancelled | cancelDate, cancellation_reason |
cancellation lifecycle event |
| Revision state | revision_state, revisionDate |
term/change lifecycle |
| Policy status | policy_status, policyStatus |
policy lifecycle state |
| Claim status | claim_status, status |
claim operational state |
| Role | roles, role_name |
relationship / assigned business role |
| Tag | systemTags, claimSystemTags |
metadata or categorization label |
| Historical | historical |
record retained for prior-state context |
Good operational rule
If a field name looks like a status or tag, always check:
- Is it master-record status or relationship status?
- Is it current-state or historical-state?
- Is it a boolean flag, a text enum, or a denormalized text field?
That question will usually resolve most of the ambiguity quickly.