Resolve Custom Fields That Look Equivalent but Mean Different Things
Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.
In a hypothetical migration, a sales team uses Customer Type to distinguish direct buyers from resellers. A support team uses the same label to distinguish standard from premium service. Combining those fields would not remove duplication; it would erase a distinction.
All illustrative scenarios, field values, and mapping rules in this article are hypothetical. They demonstrate a recommended review method, not Alactic customer results or verified product capabilities.
Resolve custom field mapping conflicts by comparing actual usage, not labels. Combine fields only when they describe the same property of the same kind of record. Separate different questions, translate equivalent representations, and preserve uncertain legacy values without making them authoritative. The objective is to preserve existing meaning during migration, not redesign segmentation or reporting while records are moving.
1. Establish what each field actually means
A field label is a starting point, not a definition. Write a meaning statement for each source field:
This field describes [property] of [record or relationship], at [time or event], for [operational purpose], and is maintained by [team or process].
Identical values can conceal different meanings. In a hypothetical example, High describes commercial pursuit urgency in one system and support-case urgency in another. Conversely, differently named fields may be equivalent if their definitions, update rules, and uses align.
Collect evidence beyond field configuration:
- Recorded values: inspect populated records, blanks, unusual values, and disagreements between apparent equivalents.
- Write paths: identify people, forms, imports, and processes that populate or overwrite the field.
- Read paths: identify reports, filters, routing rules, exports, and integrations that consume it.
- Timing: establish whether the value describes current state, an event, or a historical snapshot.
- Ownership: ask responsible teams to explain concrete records rather than merely approve a definition.
Keep each field's source system, internal identifier, and visible label in the mapping register. Otherwise, renamed labels can make an old field look new or make distinct fields indistinguishable during review.
2. Choose among four mapping decisions
Use explicit outcomes rather than treating every conflict as naming cleanup:
- Combine: the fields represent the same property, record scope, timing, and value meaning. Any conflicting populated values still need an approved resolution rule.
- Separate: the fields answer different questions, even when their labels or values overlap. Give them distinct destinations.
- Translate: the property is equivalent but its representation differs. Document a conversion covering every observed value.
- Preserve legacy: the meaning is uncertain, the information is historical, or no safe operational destination exists. Retain the raw value and provenance without treating it as a current business fact.
These decisions can work together. Translate understood values while holding ambiguous ones for review; do not quietly extend the translation to cover them.
Scrutinise many-to-one conversions. Mapping several source categories to one broader destination may be intentional, but it loses detail. Make that loss an explicit approval decision and preserve source detail wherever it remains necessary.
3. Use a filled semantic decision table
The following fictional worked example compares Sales CRM and Service CRM. This filled artifact can be reused as a mapping-register structure: replace the hypothetical entries with verified definitions, evidence references, and named approvers.
| Field | Conflicting meanings | Decision | Destination constraint | Evidence required | Acceptance check |
|---|---|---|---|---|---|
| Customer type | Sales: buying relationship. Service: service entitlement. | Separate into Buying Relationship and Service Tier. | Two independent destinations; no concatenation into one category. | Assignment records, entitlement changes, consuming rules, and approval from both owners. | A premium reseller retains both facts. |
| Region | Sales: sales territory. Service: installation area, sometimes abbreviated. | Separate territory from installation region; translate verified installation codes. | Installation region belongs to the location, not automatically the account. | Codebooks, location relationships, boundary history, and disagreeing records. | Different selling and installation regions survive on the correct records. |
| Priority | Sales: pursuit urgency. Service: case response order; Queue Rank encodes that same order. | Separate sales priority; translate and combine equivalent service fields. | Accepted options must fit; no silent default for unknown values. | Queue definitions, write processes, change examples, and approved conflict rule. | Equivalent service inputs agree; sales urgency never changes case order. |
| Account value | Sales: annual potential. Service: active contract value. Legacy import: undefined amount. | Separate defined measures; preserve ambiguous legacy figures. | Check currency, precision, period, and calculated-field restrictions. | Calculation definitions, dates, currency evidence, supporting records, and legacy provenance. | Potential is not contracted value; uncertain figures stay out of operational totals. |
Apply the table to actual record transformations
Customer type: Sales Reseller becomes Buying Relationship = Reseller; service Premium becomes Service Tier = Premium. If only one suitable destination exists, hold the unresolved dimension in a controlled legacy reference until a separate destination is approved. Do not create a combined category that mixes relationship and entitlement.
Region: Sales NE remains Sales Territory = NE. Installation NE becomes Installation Region = Northeast only when its codebook confirms equivalence. For an account with multiple installation locations, map each location separately rather than choosing one account-wide region.
Priority: Queue Rank 1, 2, and 3 translate to High, Normal, and Low respectively. If a case has Queue Rank = 1 but Service Priority = Normal, hold it for review under this hypothetical rule. Selecting the highest value would invent precedence that the evidence has not established.
Account value: an annual potential of 120000 and an active contract value of 40000 remain separate measures, with their verified currencies and periods. A legacy 85000 with no confirmed currency or period remains a raw legacy value. It is not added to either measure.
Before approval, attach evidence references, the responsible owner's decision, and the transformation version to each row. A proposed rule is not an approved rule simply because it appears in a completed table.
4. Check destination constraints before approving
A semantically sound mapping can still fail if the destination cannot represent it. Check field type, allowed options, length, precision, currency handling, required status, record scope, and whether the field is editable or calculated. Identify processes that react when imported values change.
A required field is a common pressure point. If the source is blank or uncertain, supplying a convenient default creates information the source never contained. Use an explicitly approved unknown state if supported; otherwise hold affected records for resolution.
Distinguish blank, unknown, not applicable, and explicit negative values wherever the source distinguishes them. Collapsing them into an empty destination may change report membership or workflow behaviour.
Also review who can view and change imported values. A broadly visible notes field is not automatically an appropriate legacy store for restricted data.
If Alactic is the destination, check the relevant workspace and current plan for field and import availability. These recommendations do not establish that a particular control exists, or that CRM capabilities are available in Communications or the Android app.
5. Resolve disagreements without inventing certainty
When equivalent fields disagree on the same record, investigate before choosing a winner. One may be stale, one may describe a different date, or their apparent equivalence may be wrong.
A record's latest modification timestamp does not necessarily identify its latest field value: another property may have caused the update. Likewise, the source with the most populated records is not necessarily authoritative.
Approve precedence only when evidence supports it. A rule might use a designated system of record, a reliable field-level change event, or manual owner review. State the fallback when that evidence is missing.
For unresolved values, retain the source record identifier, field identifier, raw value, extraction reference, and exception reason in a controlled register. Assign an owner and a disposition: approve a mapping, preserve as legacy, or continue holding the record.
Keep future redesign separate. If teams disagree about what priority should mean going forward, open a distinct decision. Migration should not settle that debate by rewriting historical meaning.
6. Test meaning, not just successful imports
An import can accept every row while changing the meaning of many records. Build tests from the decision table, including ordinary records and known exceptions.
- Different dimensions: confirm that buying relationship and service tier both survive.
- Equivalent representations: confirm approved code translations and exception handling for unrecognised codes.
- Conflicts: confirm that disagreements follow the documented review or precedence rule.
- Missing values: confirm blanks do not acquire unapproved operational defaults.
- Record scope: confirm location and case values do not overwrite account properties.
- Downstream use: compare report and queue membership, not only totals.
For a hypothetical priority test batch of 12 cases, suppose eight have agreeing service values, two contain only recognised queue ranks, one has conflicting values, and one contains an unknown code. The expected disposition is ten mapped cases and two held exceptions. Those totals are test expectations, not evidence of migration success.
Inspect the ten mapped cases individually: recognised ranks must have the approved meanings, and sales priority must remain separate. Then verify that both exceptions stayed out of operational priority assignment. Twelve accepted import rows would fail this test if the two exceptions silently became Normal.
Reconcile source values across mapped, preserved, and unresolved outcomes. Counts alone cannot detect swapped categories. Obtain owner approval tied to the mapping version and before-and-after records showing the relevant distinction.
7. Recover from a mapping that changed meaning
If testing or live review exposes a semantic error, stop further application of the affected mapping where authorised and practical. Identify the migration run, transformation version, source fields, destination fields, and affected record identifiers.
Compare destination values with the preserved source extract and available pre-import values. Check for legitimate edits made after migration before restoring anything: a blanket overwrite may repair the mapping while destroying newer work.
Prepare a targeted correction or rollback, test it against affected cases, and obtain approval before execution. Inspect downstream consequences separately. Correcting a field does not necessarily reverse queue changes, tasks, exports, or other actions triggered by the wrong value.
Repeat value reconciliation and business-use checks after correction. Keep unresolved records visible in the exception register rather than treating a completed corrective import as proof of recovery.
8. Approve one collision set before expanding
Start with one repeated label feeding an active report or workflow. Complete its decision-table row, attach evidence, document transformation and conflict rules, identify the approver, and test its acceptance check.
Expand only when every observed source value has an approved destination, a documented legacy-preservation route, or an explicitly held exception. No unexplained defaults or unapproved overwrites should remain in the test set.
The useful milestone is not fewer custom fields. It is an auditable explanation of which original fact each migrated value still represents.
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