Bulk Enrichment Windows: Verify How Ongoing List Changes Affect a Batch
TL;DR: Test a bulk enrichment tool by recording the submitted record IDs, documenting authorized list changes, and reconciling results against the provider’s stated membership rule. For RevOps buyers, a completed batch or credit charge does not prove which people were enriched. Verify snapshot behavior, late additions, removals, retries, and destination updates separately.
What does a bulk enrichment batch actually include?
The input set is the group of records the enrichment operation is expected to process. It may be an explicit set of IDs, a file, selected rows, or a reference to a list. Those inputs are not interchangeable. A list can change while the operation is running, and a record’s fields can change even when its membership stays the same.
Ask the provider when membership is resolved for the exact feature you will use. Is it captured when you submit, when processing begins, or later as work is dispatched? Does the job use values copied from the submitted rows or read current record values during processing? Do not infer either answer from a progress bar.
This guide provides a proposed evaluation method, not a report of a completed enrichment experiment. It focuses on membership and reconciliation. Provider selection, credit pricing, and general data coverage remain separate buying questions.
| Boundary | Question for the provider | Evidence to retain |
|---|---|---|
| Membership | Which records belong to this specific operation? | Submitted IDs or an equivalent provider-supported input manifest |
| Input values | When are lookup fields read? | Documented behavior and the permitted input version |
| Execution | Which records were attempted and which reached a terminal state? | Per-record results associated with the operation |
| Destination | Which fields and systems may be updated? | Mapping and write-policy configuration |
| Billing | What event creates a charge? | Applicable billing rule and operation-linked usage record |
Why is a saved list not proof of a frozen batch?
A list is a way to organize records. The enrichment job has its own execution behavior. Even if list membership is manually maintained, the job may read that membership at a different point from the one you expect. Conversely, an evolving audience could be converted into an explicit input set before enrichment. Verify the operation rather than reasoning from the label.
In Unify, list building separates controlled list membership from audiences whose results follow saved conditions. Creating a list from criteria collects the records matching at that point; later additions and removals happen separately. This describes the list, not an in-flight enrichment snapshot guarantee.
Use the saved-search versus frozen-list review to settle how the source audience evolves. Then document what the specific enrichment action submits. The source view and the batch input may coincide initially without remaining identical throughout processing.
A count alone cannot resolve the question. The same total can represent a different group of people after additions and removals. Preserve record identity and the selected operation, not just the number displayed in the list header.
What should you capture before submission?
Record the source system, list or file reference, record type, chosen enrichment fields, and authorized destination. Capture the input IDs through a provider-supported export or request record. Preserve the original input separately from the working list so later edits cannot silently redefine what you intended to submit.
Document the selection method. Explicitly selected rows, all matching rows, a filtered view, and an entire list may produce different inputs. Ask the operator to show the selected scope immediately before submission. Retain the relevant filters and selection state with the input manifest.
Unify list search and filters narrow the records shown without changing membership. List CSV downloads also use the current search and filters. Clear them when you intend to export the whole list, or record them when the filtered subset is intentional.
If you collect the manifest through the Unify Data API, follow cursor pagination with the same filters until no next cursor remains. That retrieves the available pages according to the API contract; it does not by itself establish a transactionally frozen snapshot. Ask how concurrent changes affect your extraction before using it as authoritative evidence.
- Input identity: Record IDs and the system that issued them
- Lookup values: The fields the enrichment request is allowed to use
- Selection scope: Filters, explicit selections, and list or file version
- Operation reference: The batch or request identifier returned by the provider
- Allowed changes: The test mutations and destinations approved by the administrator
- Expected behavior: The provider’s stated membership and value-reading rules
How do you agree on expected behavior without inventing a result?
Write the expected behavior from a current product contract, documentation, or a precise vendor answer for the selected workflow. Keep a record of the feature, configuration, and scope to which that answer applies. A statement about a different import route should not become the expected result for your batch tool.
If the provider cannot establish whether membership is frozen or live, mark it unresolved. You can still conduct an authorized observation, but you should not call the outcome a contractual guarantee. Decide whether the workflow needs a stronger commitment or a narrower input method before adoption.
Ask separately about records added before submission, after submission but before processing, and during active processing. Ask what happens when an original member is removed. The point is not to prescribe a universal implementation; it is to identify behavior your team can operate and reconcile.
Also ask whether cancellation stops work not yet dispatched, interrupts work already running, or only prevents later stages. Until that behavior is confirmed, removing a row from a list should not be treated as a reliable way to cancel its enrichment.
How should the authorized evaluation be isolated?
Use an approved test workspace or a bounded workflow that will not accidentally trigger outreach or uncontrolled production writes. The administrator should confirm how downstream actions are held. Use permitted records and only the changes needed to examine membership behavior.
Keep the initial run focused. Avoid changing filters, lookup fields, membership, mappings, and retry settings all at once. When several things change together, the result may be impossible to attribute to a specific cause. Document one controlled change at a time where the provider’s test setup permits it.
For research performed through list building with chat, review the results and the saved list separately. A research response, a saved record, and an enrichment operation are different artifacts. Keep the exact submitted collection visible even when the surrounding conversation continues.
If source criteria come from shared segments, document the definition used for the evaluation. Changes to a shared definition affect its uses. Avoid casually editing a production segment merely to create a batch test condition; use an approved isolated setup instead.
How should additions and removals be observed?
Record each authorized membership change with the relevant record ID, operation reference, and observed order of events. Keep the provider’s processing state at the time of the change. Do not invent a timestamp or assume that an interface refresh reveals the exact moment of backend processing.
For an added record, determine whether it appears in the original operation’s result set, a later operation, or neither. For a removed record, determine whether work was skipped, canceled, completed, or still unresolved. Use the provider’s actual state names and definitions rather than translating every unfamiliar status into failure.
In Unify, removing a record from a list removes membership while leaving the company or person record in the workspace. It is not the same as deleting the record or canceling an already submitted action. Keep those separate operations out of your inference about batch behavior.
Do not use a missing row in the current view as the only proof of exclusion. Filtering, removal, permissions, and display timing can affect what you see. Reconcile against the operation’s results and the original ID manifest, then investigate any remaining discrepancy.
What if the record changes while membership stays the same?
Treat identity and field version as separate dimensions. The same record ID can acquire a corrected email, company domain, or other lookup value. Ask which version the enrichment request uses and how the result identifies the input it actually processed.
Unify CSV imports match companies by domain and people by email. Mapped values can update an existing matching record. A valid people record requires an email supplied in the file or found through enrichment; providing lookup information does not guarantee a match. Preserve the file and mapping decisions when analyzing the resulting records.
If you use a CSV upload, include its header row and respect the documented file limit of 20 MB. Separate rows accepted into the operation from rows rejected or skipped during import. An import rejection is not evidence that an accepted batch member was later dropped during enrichment.
The company CSV API tutorial uses domain matching and a fill-empty upsert property to avoid replacing existing values. That specific request configuration does not describe every UI upload or enrichment operation. Record the actual update policy your workflow uses.
If a CRM record is converted or replaced during the run, preserve the cross-system identity mapping. Use the separate Salesforce lead-conversion continuity review to investigate record continuity. Do not count a changed destination identifier as a new person without resolving that identity question.
How do you reconcile the completed operation?
Start with the original input manifest and join the available results by stable record identity and operation reference. Retain newly added records separately until their membership is explained. Preserve missing result rows as unresolved rather than silently excluding them from the denominator.
For every submitted record, identify its final processing state or the reason no final state is available. Distinguish a completed lookup with no match from a failed request, a skipped input, and work that has not finished. Keep the provider’s original result alongside any internal reporting category.
| Original input ID | Authorized list change | Expected inclusion and basis | Observed inclusion | Result state | Unresolved item |
|---|---|---|---|---|---|
Populate the observed columns only after a permitted evaluation. Store any output record ID, lookup-field version, and destination update reference in the linked working record. The blank worksheet defines what to collect; it does not claim that any provider has passed the test.
Compare sets before interpreting rates. Determine whether every expected input has an explained outcome, whether any unexpected input was processed, and whether a retry created another attempt for an already accounted-for record. Report unresolved identities separately from ordinary no-match results.
Then summarize the counts using the reconciled input set. A coverage percentage based only on returned rows hides missing inputs. A completed-job label can describe the operation while individual results still require interpretation. Keep the membership decision visible in the report.
How should retries and credit events be handled?
Retain the original operation and each retry reference. Ask whether the retry targets unresolved records, all submitted records, or the list as it exists at retry time. Those alternatives can produce different input sets even when the interface uses the same retry label.
Separate attempts from unique records. A second request for the same person should not automatically count as another person enriched. Keep the link between attempts, final outcomes, and any destination writes so your team can explain the result without inflating volume.
For Unify CSV enrichment, the review screen shows the credit cost and eligible rows, and successful enrichment uses credits. That specific behavior does not establish every provider’s no-match, cancellation, or retry billing rule. Verify the applicable event separately from the membership test.
A usage charge is supporting operational evidence, not an identity manifest. Reconcile billing only after you understand which operation and records it covers. Do not infer that every current list member received enrichment because a total credit amount changed.
How do you verify destination writes separately?
An enrichment result and a destination update are distinct outcomes. A useful value may be returned but not written, or a write policy may intentionally preserve an existing field. Retain both states so the evaluation does not misclassify controlled non-overwrite behavior as missing enrichment.
Unify HubSpot field mappings configure reads and writes independently. Inspect the selected field behavior before the test. The choice to fill an empty field differs from replacing an existing one, and the intended behavior should be part of the acceptance rule.
Check HubSpot sync rules for the event that actually triggers a write and whether it may create a record. These rules and mappings affect the destination result. Unify does not merge existing HubSpot duplicates, so a successful sync should not be treated as proof that duplicate CRM records were consolidated.
Keep company and contact enrichment focused on the data needed for the downstream task. Verify that the result belongs to the intended entity, is suitable for use, and reached the allowed destination. Membership reconciliation is necessary, but it does not establish every aspect of data quality.
What should make you accept, adapt, or stop the workflow?
Accept the configured workflow only when its membership rule is sufficiently clear, observed results can be reconciled, and destination behavior matches your approved policy. Keep the decision scoped to the feature and configuration tested. Do not generalize a successful file upload to every continuous enrichment path.
Adapt the process when a supported explicit input set or controlled export can remove ambiguity. Stop and escalate when unexpected records are processed, submitted records disappear without explanation, or the team cannot determine which inputs a retry will use. Resolve those gaps before increasing the operational scope.
Sign up for Unify to organize the records, research, and review steps that support an accountable enrichment workflow.
Frequently asked questions
Does a manually maintained list guarantee a frozen enrichment batch?
No. List membership and batch execution are different behaviors. Verify when the selected enrichment action resolves membership and reads lookup values.
Does removing a record from a Unify list cancel its enrichment?
Do not assume that. Removing list membership leaves the workspace record in place. Cancellation of an already submitted action requires separate confirmed behavior.
Is a completed batch label enough to establish which records were enriched?
No. Reconcile the submitted IDs with per-record results and operation references. Keep missing outcomes, unexpected records, and unresolved identities visible.
How should late additions be counted?
Determine whether they belong to the original operation, a later operation, or neither under the provider’s stated rule. Do not add them to the original denominator solely because they now appear in the list.
Can an enrichment credit event prove that every list member was processed?
No. Verify billing events separately and connect them to the actual operation and records. A credit total does not establish membership or successful destination writes.
What should I do if snapshot behavior is undocumented?
Ask for a precise answer for the selected feature and configuration. Record the gap, conduct only an authorized bounded evaluation, and avoid treating an observation as a universal guarantee.
Glossary
- Input manifest: The retained identities and scope intended for a specific operation
- Snapshot membership: A batch input set fixed at a defined point established by the provider
- Live membership: Membership evaluated against a changing source according to the operation’s rules
- Terminal state: A provider-defined outcome indicating that an attempt has finished
- Reconciliation: Accounting for expected and observed records using identities and operation results
Sources
- List building
- Saved Search or Frozen Prospect List? Test How the Audience Changes
- Lists
- Unify Data API
- Build lists with chat
- Segments
- Import data
- Import data from a CSV
- Enrichment Integrations After Salesforce Lead Conversion: Test Record Continuity
- Field mappings
- How bidirectional syncs work
- Your next 100 leads are a prompt away

