Issue with Exact Type Match in Column A Causing Unexpected Clustering in Zingg Predictions

Aditya J. · 2026-03-31T16:21:15.114Z

Hey Im facing some issue when using zingg. I have used column "a" as exact type match but after prediction, im getting some records where column a is exact match but still they are in different cluster.

Zingg Community

Issue with Exact Type Match in Column A Causing Unexpected Clustering in Zingg Predictions | Zingg Community

Padam P.
·
·
What you are looking for is https://docs.zingg.ai/latest/stepbystep/configuration/deterministicmatching In Zingg, EXACT match and deterministic matching are quite different in purpose. EXACT match is defined at the field level . On that particular field, records will be considered the same only if they are an exact match but other fields also decide if two records match. On the other hand, deterministic matching is a rule-based approach at the record level. Here, you define strong conditions like “if email matches exactly, then these two records are the same,” and Zingg will directly mark them as a match without involving the ML model at all So in short, the EXACT match contributes to the ML scoring process, while deterministic matching acts as a hard rule that can directly decide matches.
Aditya J.
·
·
Aditya J.
·
·
If you see this example record, my Z_Score depend on these five columns: ("linkedin_url", "string", MatchType.NULL_OR_BLANK, MatchType.EXACT), ("full_name", "string", MatchType.NULL_OR_BLANK, MatchType.FUZZY), ("email", "string", MatchType.NULL_OR_BLANK, MatchType.EXACT), ("company_name", "string", MatchType.NULL_OR_BLANK, MatchType.FUZZY), ("domain", "string", MatchType.NULL_OR_BLANK, MatchType.EXACT),
Aditya J.
·
·
based on the conditions, it should give some score only on linkedin url,
Aditya J.
·
·
other components are also matching if you will see then why null zscores and why they are in different cluster
Aditya J.
·
·
Padam P.
Padam P.
·
·
Could you please check your labelling? Maybe you have marked such a pair of records as No Match
Aditya J.
·
·
I mostly added labelling pairs manually via df. Still I will check once if you say so
Aditya J.
·
·
I have added like 600 pairs based on 15 diff conditions as matching and no matching values
Aditya J.
·
·
Padam P.
Aditya J.
·
·
Hey Padam P. i tried those but still same result
Aditya J.
·
·
is this the issue with blocking tree zingg internally creating. Also, FYI, im using zingg 0.3.4
Aditya J.
·
·
@here
Aditya P.
·
·
firstly update your zingg version
Aditya J.
·
·
okay so is this the zingg version issue?

118 comments