Missed match
1 person · 2 identities
M. Rivera
not linked
Morgan Rivera
You see only part of Morgan Rivera's activity.
Platform / Entity Resolution
Resolve the identity.
Resolve the identity.
Follow the activity.
Identify which records refer to the same person, organization, or other entity. See why they match, then follow their activity and relationships across sources in DataWalk's enterprise knowledge graph.
Native Entity Resolution runs within DataWalk. You can resolve records and explore the results in the same platform, without deploying or integrating a third-party matching engine.
The challenge
The wrong identity changes the answer.
Different systems hold different pieces of an entity's activity. Names and identifiers can vary, be missing, or be shared. That makes it hard to tell which records belong together.
False match
2 people · 1 identity
J. Park · ID-221
wrongly linked
J. Park · ID-735
You attribute one J. Park's activity to the other.
Matching & resolution
Match beyond exact identifiers.
Names change. Addresses vary. Identifiers may be missing, inconsistent, or shared. Specialists configure matching rules and thresholds to balance the risk of missed matches and false links.
Compare identifiers and attributes using appropriate exact, fuzzy and phonetic methods.
Evaluate combinations of names, addresses, contact details, identifiers and other relevant attributes.
Set matching conditions and thresholds to reflect your data, use case and tolerance for different types of error.
Narrow the comparisons to relevant candidates before evaluating whether records match.
Platform advantage
Follow the identity across sources.
Start with a resolved identity in DataWalk's enterprise knowledge graph. Follow the transactions and relationships held in the records linked to it.
The source records retain their original attributes and relationships. Analysts can inspect them individually or explore them together.
Use the result your way.
Retain linked source records, work with resolved-entity representations, or use supported merging approaches where the application requires them. How identity is represented can be configured for the work that follows.
Linked source recordsGrouped for visual analysisResolved-entity representationMerged records
Available approaches differ between established configurable ER and Native ER.
Fragmented relationships
Connected activity and relationships
Explainability
Inspect what made the match.
The evidence stays with the match.
See which records were linked and which details and rules supported the match. When records stay separate, inspect whether DataWalk compared them and what evidence kept them apart.
The same inputs and configuration produce the same verdict. Language models play no role in matching.
Operational resolution
Resolve identities as your data changes.
Build a resolved population from your records, update it as data changes, and check whether a person or organization is already known.
What do we know across this entire population?Establish identity across a large dataset so analysts can work from a prepared population rather than reconstructing it record by record.
What changes when new records arrive?After the initial batch, Native ER processes new or changed records in subsequent runs.
Does this new subject match anyone we already know?Check a person or organization against the population DataWalk has already resolved. The API returns a verdict and supporting evidence. The check is read-only.
What changes if we apply different matching rules?Run alternative configurations in the same environment and compare their results before operational use.
Measured scale
Proven at population scale.
162M
Organization records
3.69B
Candidate pairs evaluated
3h 04m
Active processing on 12 compute nodes
The same workload took 7h 03m on 3 compute nodes. Some matching work requires data to move between nodes, so adding nodes does not reduce processing time in direct proportion.
Active processing covers normalization through matching. Result import is outside the timed window.
FAQ
Frequently Asked Questions
Find your answer.
What is entity resolution?
Entity resolution is the process of determining which records refer to the same real-world entity, even when names, identifiers and other attributes differ across data sources. It uses matching methods, rules and available evidence to identify records that belong together. This helps organizations establish a more consistent understanding of people, businesses and other entities across fragmented data.
Why is entity resolution important for a knowledge graph?
A knowledge graph connects entities through their relationships and activity. When the same person or organization appears under different identities, important connections can remain fragmented. False matches can create relationships that shouldn't exist. Entity resolution helps establish which records represent the same real-world subject, improving the reliability of connected analysis.
How is entity resolution different from deduplication?
Deduplication is a common application of entity resolution, focused on identifying duplicate records. Entity resolution also establishes identity across different sources and contexts. Once records are identified as representing the same entity, organizations can choose whether to retain them separately, connect them through identity relationships or consolidate them. Matching records and merging records are different decisions.
Does DataWalk merge or delete original records?
Not necessarily. DataWalk can represent resolved identities through relationships between source records, allowing their original attributes and connections to remain available for analysis. Whether records are consolidated, preserved or represented through resolution links depends on the implementation and configuration. Native ER links records and preserves their source attributes and relationships for analysis.
How can entity resolution accuracy be evaluated?
Entity resolution should be tested using representative records with known matching and non-matching outcomes. Evaluations should measure false matches, missed matches and the quality of candidate selection, rather than relying on a single accuracy figure. DataWalk can be evaluated against these requirements using customer-relevant data and matching rules.
Can DataWalk resolve new records as data changes?
Yes. Subsequent batch runs process new or changed records. The on-demand API checks one person or organization against the population already resolved.
What does implementing Native ER involve?
Prepare and map source fields, configure matching rules, and validate the results on representative records. The quality of the source fields and mappings affects which records can be matched.
See DataWalk in action
Test the identities that matter.
Bring representative records, difficult matches and known non-matches. Test how DataWalk establishes identity, examine the supporting evidence and explore what the results make possible for analysts working across connected data.