A company's website, directory page, social profile, filing, and search result may disagree about its phone, address, domain, or operating status. The safest response is not to select the newest-looking value automatically. W3C PROV provides a model for describing entities, activities, and agents behind data, while NIST information-quality guidance emphasizes objectivity, utility, and integrity. This lesson turns those principles into a practical contact-list conflict queue.
VISUAL LESSON
What you will learn
- 01Create field-level provenance records.
- 02Prioritize identity, authority, and recency appropriately.
- 03Publish resolved values with confidence and review states.

ILLUSTRATIVE WORKED EXAMPLE
Score an illustrative phone-number conflict
PRACTICAL INTERFACE MAP
Move from disagreement to documented resolution
Keep business and location identity stable; list each candidate value, source URL, date, context, and collection method.
Check directness, authority, branch fit, current context, consistency, corroboration, and explicit update dates.
Publish chosen value, alternatives, reason, confidence, reviewer, verification date, and next action.
STEP-BY-STEP LESSON
Conflict → provenance → evidence comparison → resolved or deferred
THE LEAD ATLAS METHOD
Lead Atlas Data can research a custom list for a customer's exact campaign, market, locations, and business categories, with source evidence and field review aligned to the requested data scope.See how custom list research works ↗Isolate the exact conflict
Create one conflict record per field and business unit. Do not mix a headquarters address disagreement with a branch phone disagreement or a legal-name question. Preserve candidate value, source URL, page context, access time, collection method, business identity, location identity, and the raw text surrounding the value.
Assign stable business and location keys before choosing a winner. If identity itself is uncertain, resolve or defer that question first; perfect scoring cannot make two different businesses the same entity.
Build the provenance ledger
For every candidate, record who or what published it, where it appeared, when it was accessed, any stated effective date, how it was transformed, and which researcher or process collected it. This creates a practical trace from the export field back to the source activity and responsible owner.
Keep snapshots or quoted fragments only within lawful, necessary limits and preserve the source URL and page title. A search snippet, aggregator, or cached result can be a discovery clue but should not silently outrank the current page it points to.
Evaluate source context
Score directness, authority for the specific field, match to the correct location, recency evidence, internal consistency, corroboration, and known limitations. An official corporate page may be authoritative for headquarters but less useful than an official branch page for a local phone. A newer directory timestamp may reflect recrawl, not a business update.
Write deterministic rules for common fields and an override path for unusual evidence. Include a reason code such as official location source, official contact source, two-source corroboration, stale source, branch mismatch, disconnected clue, or unresolved.
Resolve, retain, or defer
Choose a value only when evidence supports the intended use. Preserve rejected alternatives, source links, decision rule, reviewer, and date. Use conflict unresolved, needs direct verification, or do not publish when uncertainty is material. Avoid filling an export simply because the schema expects a value.
Assign confidence as a documented evidence state, not a promise of future validity. If direct verification is authorized, record method and date without implying that a receptionist, voicemail, or automated message confirms every other business field.
Monitor the resolution over time
Set review intervals based on field volatility and campaign risk. Phones, locations, staff roles, and websites can change at different rates. Reopen a conflict when a newer credible source appears, the campaign reports a failure, or the business identity changes; never overwrite the history.
Deliverable: field-level provenance schema, conflict queue, evidence scoring rules, branch-versus-parent logic, source hierarchy with exceptions, chosen and alternate values, reason and confidence states, reviewer record, verification date, versioned export, and refresh triggers.
Turn this lesson into a research brief.
Apply “Resolve Conflicting B2B Contact Data With a Source-Provenance Ledger” to one campaign before requesting or using a list.
- 01Market boundary
Name the locations and business categories this decision applies to.
- 02Fit evidence
Write the public signals that would make a business relevant enough to review.
- 03Exclusions
List the business types, markets, and records that should not enter the campaign.
- 04Outreach use
State who will review the list, personalize the message, and record outcomes.
THE TAKEAWAY
Resolve contact data one field at a time, weigh identity and source context before recency, preserve alternatives and reasons, and set a verification date rather than pretending conflict has disappeared.OFFICIAL REFERENCES