Product Record Diff (Attribute Comparison)

When two records look like the same product, the question is never “are they similar?” but “where exactly do they differ?” Paste two records below to compare product records attribute by attribute and see, field by field, what matches, what conflicts, and what only one side has — the groundwork every merge or deduplication decision depends on.

What it checks

The diff aligns two records by attribute name and evaluates each field independently. For every attribute it reports:

• Exact matches — values that are byte-for-byte identical after trimming whitespace (e.g. both records carry GTIN 04003318012341).
• Normalized matches — values that differ only in formatting: case, punctuation, leading zeros, unit notation (3/4 in vs 0.75"), or spacing in a part number (HBR-12 05 vs HBR1205). These are likely the same value expressed differently.
• Conflicts — fields where both records have a populated value and they genuinely disagree (one furniture SKU says Oak, the other Walnut; one MRO fitting lists thread 1/2-14 NPT, the other 1/2-13 UNC).
• One-sided fields — attributes present in only one record, so a merge would inherit them rather than choose between them (record A has a UNSPSC code, record B has a hazmat flag).
• Missing on both — fields defined in your header set but empty in both records, useful for spotting enrichment gaps before they propagate.
• A per-record field count and fill rate, plus an overall similarity summary so you can gauge at a glance whether the pair is a near-duplicate, a partial match, or two distinct products.

How it works

The tool performs a field-level set comparison rather than a single whole-record string diff. After parsing both records (CSV row, JSON object, or pasted key–value pairs), it builds the union of attribute keys, then compares each key’s values through a small normalization pass: case folding, whitespace collapse, punctuation stripping for identifiers, and common unit harmonization. A raw mismatch that survives normalization is classified as a true conflict; one that resolves is flagged as a formatting-only difference. This mirrors how a deterministic matcher treats identifiers and how data normalization is applied before any probabilistic step — the same logic behind a fuzzy match score, but exposed transparently so you can audit each field yourself.

All processing happens client-side, in your browser. Nothing is uploaded, stored, or sent to a server, so you can compare confidential CPG cost fields, industrial-distribution supplier records, or customer catalogs without a data-handling review. Refresh the page and the data is gone.

Related resources

FAQ

How do I compare two product records to see what's different? +

Align the records by attribute name and compare each field on its own rather than diffing the whole row as one string. For every attribute you want to know whether the values are identical, formatting-only variants, genuine conflicts, or present on just one side. The tool above does this automatically and labels each field, so you can scan a pair of MRO, CPG, or furniture records and immediately see the handful of attributes that actually disagree.

What's the difference between a record diff and a fuzzy match score? +

A fuzzy match score collapses a comparison into a single number that estimates how likely two records are the same product — useful for ranking thousands of candidate pairs. A record diff does the opposite: it expands one pair into a full field-by-field breakdown so a human can review exactly why the score landed where it did. You typically use scoring to find candidates and a diff to adjudicate the borderline ones.

Are conflicting values always a sign these are different products? +

No. Many conflicts are formatting or vocabulary differences — 0.75" vs 3/4 in, Oak vs OAK, or a part number with and without a hyphen. The diff separates these formatting-only differences from true conflicts, where both sides carry incompatible real values (a different thread spec, a different color, a different pack quantity). True conflicts on identifying attributes are the ones that suggest two genuinely distinct products.

Is my product data uploaded anywhere when I use this tool? +

No. The comparison runs entirely in your browser using client-side code. Your records are never transmitted to or stored on a server, which is why you can safely diff confidential cost, supplier, or customer-catalog fields. Closing or refreshing the tab clears the data.

What do I do after I find a difference? +

Decide which record’s value should survive for each conflicting field, then merge into a single canonical record while preserving the source of each value so the merge stays reversible. For setting consistent rules across many pairs — including when to merge automatically versus route to review — see the auto-merge thresholds playbook and the reversible-merges guide linked above.