Compare Two CSV or XLSX Files
Compare two spreadsheet files using a selected key and group records into matched, missing, duplicate and conflicting buckets.
Build dependable lookup, filter and counting formulas from clear inputs instead of guessing spreadsheet syntax.
ReconNimble’s formula generator turns a business question into an explicit spreadsheet formula without sending the request to a generative AI service. Choose Excel or Google Sheets, select the formula family and enter the ranges exactly as they appear in your workbook. The tool shows the resulting formula, explains what each part does and creates one reusable share link for identical inputs. This deterministic approach is especially useful for HR, IT, finance and operations teams that need repeatable formulas they can review before placing them into a live workbook.
Start by defining the business question rather than the button you plan to press. For Formula Generator, a typical situation is that an operations analyst needs a dependable lookup or counting formula without sending workbook details to an AI service. Before running anything, record the source date, the owner of each dataset and the result you expect from at least three known records. That small control makes it easier to spot a wrong key, wrong delimiter or wrong source file before the same mistake affects thousands of rows. ReconNimble deliberately keeps the free workflow visible: the browser shows the selected inputs, summary counts and result rows instead of turning the task into a black box.
The central rule for this workflow is to state the lookup value, ranges, target application and expected missing-value behaviour explicitly. Stable identifiers are preferable to descriptive text whenever identity matters. A field can look unique in a ten-row preview and still repeat later, so inspect blanks and duplicates before treating it as a key. Keep an untouched copy of the source files and do not “clean” away exceptions merely to make counts agree. If two systems encode the same identifier differently, correct or normalize that known format explicitly; do not assume that similar-looking values refer to the same person, asset, invoice or transaction.
Use the supplied fictional sample first, then repeat the same steps with a small extract from your own authorised data. Write down one record that should succeed, one that should appear as an exception and one edge case such as a blank key or quoted field. Run the tool and confirm those three cases before reviewing totals. For this task the useful output is a formula that can be copied, explained and tested against known records. If the known cases do not land in the expected groups, stop and inspect headers, selected keys, source scope and data types rather than accepting a plausible-looking total.
Treat counts as navigation, not proof. Open several result rows from each status and compare them with the original source values. Pay special attention to duplicates because a repeated key changes the meaning of one-to-one matching and joins. Where the tool creates a local download, open that new file separately and confirm its row count, headers and a few values at the beginning and end. Keep the original sources unchanged so another reviewer can reproduce the result. If the output will drive payroll, payment, asset ownership, compliance or another high-impact action, obtain the appropriate business-owner sign-off.
A recurring failure mode is invented sheet names, wrong absolute references, approximate matching and locale-specific separators. Another is treating normalization as data correction: trimming spaces or comparing text without case sensitivity can remove harmless presentation differences, but it cannot establish that two distinct identifiers are equivalent. Also remember that browser XLSX support is intentionally tabular. It reads worksheet cell values for utility tasks; it is not a full Excel rendering or calculation engine and it does not promise preservation of macros, charts, pivot tables, conditional formatting or every workbook feature. Use the right tool for the evidence you actually need.
The public Formula Generator workflow processes selected source files in browser memory. Running the tool does not send those files to ReconNimble, and supported exports are created locally. Browser memory varies by device, so the configured row cap rejects oversized jobs with a clear message rather than risking a tab crash. Publishing a result is a separate, explicit action that sends only the limited displayed result data used for the share page; never publish confidential rows. When the workflow needs durable operational evidence, use the authenticated Reconcile app when the real task is comparing whole datasets rather than calculating one cell. That separation keeps the high-traffic utility layer private while preserving a more controlled path for saved reconciliation work.
Use the sample file to test the workflow before using business data. It contains fictional records created only for demonstration.
Download sample fileThe generator cannot inspect your workbook structure from typed references. Sheet names, ranges and business rules remain your responsibility. Locale-specific separators may differ in installations that use semicolons instead of commas.
Need saved jobs or advanced matching? The free browser tool is designed for fast local checks. ReconNimble Reconcile adds authenticated saved jobs, multi-source comparison, controlled fuzzy matching, profiles, server exports and an audit trail.
No. The default generator is deterministic and browser-based, so the same inputs produce the same formula without a paid AI call.
ReconNimble canonicalises the normalized formula request to avoid creating duplicate public pages.
Yes. Select Google Sheets; the generator adjusts formula behavior where Excel and Sheets differ.
Compare two spreadsheet files using a selected key and group records into matched, missing, duplicate and conflicting buckets.
Inspect a pasted formula or the first worksheet in an XLSX file for broken references, fragile lookups and performance risks.
Suggest column mappings between two files using normalized header names, business synonyms and sampled data types.