Enrichment Providers Disagree: Check the Identity Before Choosing an Email or Phone
When enrichment providers disagree, establish whether their results describe the same person at the same employer before choosing an email or phone number. Preserve each returned identity, source and observation date. Review the identifiers sent in the original lookup, then verify the person-to-company relationship with suitable evidence. Keep unresolved channels out of outreach instead of selecting whichever result is newest or most complete.
A deliverable email or callable number can still belong to the wrong person. The core task is identity resolution: connecting the intended contact, employer and communication channel without combining facts from different people. This guide proposes a review process; it does not claim a measured accuracy improvement or rank data providers.
What kind of disagreement are you looking at?
First, separate an identity conflict from an ordinary field difference. Different punctuation, title wording or phone formatting may concern the same person. Different profile identifiers, employers or names may indicate different people. Do not treat all disagreements as a single “pick the best value” problem.
A job change introduces another distinction. Two results can refer to the same person while describing different employment periods. The historical employer may be correct for an earlier date without being suitable for current outreach. Preserve that history instead of declaring every older value erroneous.
There can also be a company-association issue. A parent organization, subsidiary, customer workspace and consulting employer can all appear near the same person in business data. A shared brand or related domain does not by itself prove which entity employs the contact or owns the relationship you want to discuss.
| Observed difference | Question to resolve | Premature shortcut to avoid |
|---|---|---|
| Different person or profile identifiers | Did both lookups return the intended human? | Combining one result’s title with another’s phone |
| Same person, different employers | Which employment relationship is current and relevant? | Keeping the newest lookup without checking its underlying evidence |
| Different company domains | Are these aliases, related entities or distinct organizations? | Merging because the company names resemble one another |
| Different emails or numbers | Which channel belongs to the resolved identity? | Selecting the channel solely because a validator accepts it |
| Different formatting only | Does normalization preserve the same underlying value? | Treating harmless display differences as a new identity |
Which identifiers produced the match?
Start with the request, not just the response. Record the original name, domain, profile URL, email or internal identifier used for the lookup. A provider asked to search by a broad name-and-company combination has different evidence from one given a specific person identifier. A bad input can propagate through several enrichment services.
Check field mapping in imports and integrations. Make sure a company domain was not supplied as a person identifier, an old employer was not treated as current, and a display name was not substituted for a stable record key. Review the original input before assuming the disagreement is entirely downstream.
Preserve raw values alongside normalized ones. Normalization can help compare capitalization, spacing or phone formatting, but it should not erase meaningful distinctions in names or domains. If your process changes a value before lookup, retain the transformation so another reviewer can reproduce the request.
Also record the unit of work. A list row, CRM contact, company record and enrichment response are separate artifacts. Give the review a stable reference back to the originating record. Otherwise, a correction can be applied to a similarly named contact while leaving the actual disputed record untouched.
How should you preserve conflicting outputs?
Store the competing responses before updating the accepted contact fields. Retain the provider, lookup date, returned name, employer, profile identifier and channel. Preserve the source’s own update date when available, but distinguish it from the time your system fetched the response.
Use an access-controlled review location appropriate for contact data. The goal is a traceable decision, not unnecessary duplication across spreadsheets. Keep only the evidence needed for the review and follow your organization’s retention practices.
A proposed review record should explain the conflict in plain language. “Employer differs between sources” is more useful than a generic quality flag because it identifies the unresolved relationship. Add the owner and next step so the row can move toward resolution rather than circulate between teams.
| Field | Value to enter |
|---|---|
| Originating record and intended account | ________ |
| Exact lookup inputs and transformations | ________ |
| Returned identities, channels and providers | ________ |
| Lookup dates and source dates | ________ |
| Conflict classification | ________ |
| Supporting and contradicting evidence | ________ |
| Accepted identity and employer, if resolved | ________ |
| Channel validation and outstanding questions | ________ |
| Reviewer, decision date and permitted next action | ________ |
What evidence can resolve the person and employer?
Look for evidence that directly addresses the disputed relationship. A recent first-party business interaction, a current employer page naming the person, or a directly maintained professional profile may help. Read what each source establishes rather than relying on its mere existence.
Match more than a name when the name is ambiguous. Review the role, employer, profile identifier and relevant dates together. An official company page can establish an employment association, while a prior correspondence record can help establish a channel relationship. Neither should be stretched beyond its actual content.
Evaluate whether apparently independent sources share the same upstream material. Agreement between several providers is weaker corroboration if every result repeats one stale record. Provider count is not an identity vote. Seek evidence that addresses the uncertainty rather than more copies of the same assertion.
Recency is useful but not decisive. A lookup performed today may return information collected much earlier. An older, direct exchange may be more specific than a recent broad match, yet still require confirmation of current employment. Document those distinctions instead of reducing them to a single unexplained confidence score.
When sources conflict, state the narrowest supported conclusion. You may establish that the person is the same while leaving the current employer unresolved. That partial result is useful. It does not justify selecting an employer-specific work email until the remaining relationship is supported.
When can you choose an email or phone number?
Choose a channel only after establishing that it belongs to the resolved contact in the relevant context. Email validation concerns properties of an address and its mail infrastructure; it does not automatically prove that the mailbox belongs to the person your campaign intends to reach.
Apply the same reasoning to phones. Number formatting, dialing availability and a successful connection are different from person ownership. A shared office number can be legitimate company contact information without being a direct line to a named employee. Keep the channel type visible so the caller understands what is known.
Check whether a channel is associated with a former employer, an alias, a shared function or a personal account. Select the channel appropriate for the business relationship and your organization’s outreach rules. Do not promote an uncertain channel to “primary” merely because every other field is complete.
Retain rejected candidates with a concise reason where your record policy allows it. “Different person,” “historical employer” and “ownership unresolved” should not collapse into the same empty field. Those reasons help prevent the next enrichment run from reintroducing the exact uncertainty that a reviewer already identified.
What should happen when identity remains unresolved?
Leave the accepted value unknown. A visible blank with a review reason is more accurate than a fabricated synthesis of incompatible records. Continue unrelated account research if useful, but do not release the disputed person into automated outreach.
Use decision states that describe evidence rather than provider reputation. A proposed set is “resolved,” “same person, employment unresolved,” “different person,” and “insufficient evidence.” Your systems may use other labels; consistency and visibility matter more than the vocabulary.
Define a stopping point for research. Assign the unresolved question, the next suitable evidence source and a review date. If no appropriate source is available, preserve the hold instead of repeatedly querying vendors until one response happens to match the desired answer.
Do not use a campaign send as the identity test. A reply, bounce or wrong-person response may expose a problem after outreach, but deliberately releasing an unresolved match shifts data-cleaning work onto the recipient and can contaminate your reporting.
How should reviewed records move back into the workflow?
Separate the review decision from the writeback. A reviewer can resolve identity correctly while an integration updates the wrong record or only one of several systems. Specify the destination record, fields to change and fields to preserve before applying the correction.
Inspect the saved record afterward. Verify that the accepted employer and channel are associated with the correct person, that historical context remains available where required, and that the review status is understandable to the next rep. Check pending drafts that may already contain the disputed employer or name.
Then review downstream audience membership, task assignment and sequence eligibility. Correcting a CRM field does not necessarily remove an already prepared message from another system. Give someone ownership of that final operational check.
For recurring conflicts, fix the upstream cause. Review lookup inputs, import mappings, merge rules and source refresh behavior. Keep a record of the change so the team can distinguish a repaired workflow from a one-time manual correction. This article focuses on identity; the companion waterfall implementation guide covers source ordering and cascade design.
Where does Unify fit?
Unify’s B2B Company & Contact Data product brings prospecting and enrichment into one workflow. Multiple sources can provide more candidate information, but their presence does not replace checking whether the returned identity matches the intended contact.
The Build lists with chat documentation, checked October 13, 2026, describes refining result tables, researching rows and examining reasoning and references when supporting evidence is available in an enriched cell. Use that review path to investigate a disputed match before selecting records for outreach. It does not establish that every cell has evidence or that conflicts are automatically resolved.
Bring the conflict record into your operating process and ask focused questions about the person, employer and channel. Keep the final identity decision explicit. To put research and contact preparation together, try Unify with a reviewed account list.
How should you measure the review process?
Count reviewed conflicts separately from total enriched rows. Record how many were resolved, remained unknown or were identified as different people. Those categories describe the review workload; they should not be presented as accuracy unless you have a defensible reference set and a defined evaluation method.
Track reintroduced conflicts and incorrect downstream associations as well. A system can resolve a row today and overwrite it tomorrow. Reviewing recurrence helps reveal whether source precedence or integration behavior undermines the accepted decision.
Keep operational outcomes distinct. Fewer bounces, more complete records and correct person matches are different measurements. A complete row can still be wrong, and a correctly identified person can still have no usable contact channel.
Frequently asked questions
Should the newest enrichment result win?
No. Separate lookup time from the age and specificity of the underlying evidence. Use recency as context while verifying the person and employer relationship.
Does agreement between providers establish identity?
It can provide corroboration, but shared upstream data can repeat the same mistake. Review identifiers and evidence that directly address the disputed match.
Can a valid email belong to the wrong lead?
Yes. Channel validity and intended-person identity are separate questions. Resolve both before treating the address as ready for outreach.
What if the person is verified but the employer is not?
Preserve the partial conclusion and keep employment unresolved. Do not use the uncertainty to select an employer-specific channel or personalize a current-role claim.
Is an incomplete record always a failed enrichment?
No. Leaving an unsupported value unknown can be the correct result. Judge whether the evidence supports the intended action rather than rewarding field completion alone.

