Synthetic examples: reuse-a.csv reuse-b.csv
LOCAL SPREADSHEET TOOLS
Look up spreadsheet values using multiple key columns
Join local Excel or CSV tables with explicit LEFT or INNER rules, selected return columns, duplicate checks and unmatched-row audit.
Find related records without uploading files
Use a single identifier or a composite key such as Order + Line to retrieve fields from B into A. The column names can differ: map each A key to its corresponding B key. Unlike Append, Lookup relates rows by keys; row order and column positions do not establish a match.
How it works
Select A and B, choose each sheet and confirm each header. Add one or more key mappings and explicitly select B columns to return. LEFT keeps participating A rows with blank B values for unmatched records; INNER keeps matched A rows. Both modes include a separate unmatched audit.
Empty keys and duplicate B matches block by default. You may explicitly treat empty as a value or exclude it with an audit. Explicit first/last selects an original B row when duplicates exist. Repeated A keys keep every A row; they are not silently deduplicated.
Synthetic example
A has ID 00123 and 00999. B maps 00123 to Beta. LEFT returns Beta for 00123 and an unmatched row for 00999; INNER returns only 00123, while 00999 remains in Unmatched. The downloadable examples use invented IDs and values.
If B has two 00123 rows, the default blocks. First/last makes a documented selection and reports the potential expansion count, but this version never expands one A row into many matches. Full outer and group matching are not supported.
Exact by default, explicit changes only
Text 00123 differs from numeric 123. Text IDs longer than 15 digits are not coerced to numbers. Opt in separately to trim, case-insensitivity, numeric normalization, date normalization, blank/empty equivalence or line-ending normalization. Ambiguous 03/04/2026 requires MDY or DMY when normalizing dates.
Source values and one-based A/B rows remain reviewable. Formula expressions and cached values are data, never executed or recalculated. Duplicate/blank column names remain distinct through visible source column positions and stable IDs.
Privacy, export and resource limits
No file contents, filenames, keys or results are sent over the network or stored in browser databases. The workbench has no advertisements, analytics or runtime CDN. Cancel/Clear terminates the Worker and clears the session.
XLSX includes Data, Unmatched, Duplicates_B, Summary and Run_Settings, with selected rules and a local timestamp. Original filenames are omitted; source sheet names remain only in local audit data. Formula/cache and date serial/epoch export as safe audit text. CSV exports Data only and warns about type/audit loss.
Default approximately 10k × 30 per file; 100k remains advanced. There is no fixed original-file size cap; ZIP/XML, text and column safety gates remain. At most two input budgets are retained. Result and unmatched/duplicate audit cardinality are checked against 650k default / 6.5m advanced output cells and 256 columns before materializing. Wide, high-unmatched reports can block even when input files individually fit.
Frequently asked questions
Why are apparently identical keys unmatched? EXACT distinguishes raw types, leading zeros and blank/empty values. Review the input preview types before enabling normalization.
Can duplicates multiply my result? Not in this version. Potential one-to-many cardinality is reported; block/first/last are the only strategies.
Is 100k guaranteed on a laptop or phone? No. Multiple models and output may use high memory. Use a desktop, start small and review resource warnings; advanced mode is not a device-capacity or all-browser guarantee.