A monthly cleanup should be a recipe, not a ritual.
A repeatable sequence is useful when the same report arrives with the same kinds of problems. Here is how to make it explicit—and when to stop and review.
Open the local workspace →1. Establish what stays the same
Choose one representative monthly report. Confirm which row contains headers and whether one record means an invoice, an invoice line or a customer. An invoice number alone may not be a safe duplicate key for a report with multiple product lines. That is a business decision; Migrate cannot infer it from matching column names.
Note the expected date convention before adding a rule. A value such as 05/11/2026 could describe May or November. Choose the source format explicitly. If the input mixes conventions, preserve the findings and fix the source rather than treating one successful conversion as proof that the whole column is correct.
2. Put the rules in a deliberate order
Cleaning spaces before comparing duplicate keys can produce a different result from removing duplicates first. For a recurring report, a practical starting sequence is trim whitespace, normalize selected text where appropriate, compare the chosen keys, replace explicitly defined missing markers and normalize dates. Adjust this sequence to the report's meaning.
Each step has visible settings. Avoid lowercasing identifiers or replacing every dash with an empty string merely because those changes look tidy. A dash can be a meaningful value, and mixed case can be part of an identifier. Preview removed records as well as changed cells before exporting.
3. Save the instructions, not the report
Name the recipe for its purpose, such as “Monthly sales preparation.” Saving locally stores the rule configuration in this browser. It does not store the imported rows, schedule a job or synchronize with another device. Export recipe JSON if you need a portable copy, and review it before sharing: header names and replacement settings can still be sensitive.
Recipe schema version 1 describes the file format. It is not an edit history. Saving again under the same name replaces that local recipe. Keep a separate exported copy before making significant changes.
4. Reuse it with a review checkpoint
A new report may reorder columns without changing their names. Recipes match normalized headers rather than positions, so that reordering is supported. A renamed required column needs an explicit mapping. Extra columns are shown and require confirmation; missing required columns block processing.
Matching headers do not prove matching units, types or business meaning. If “Amount” switches currency or a date column changes convention, review the recipe even when the schema is compatible. Migrate does not validate currency, email addresses, phone numbers or accounting rules.
Try two months safely
Start with the synthetic sample, add whitespace and duplicate rules, generate a preview and save the recipe. Then use “Try next-month sample” inside the workspace. Its reordered columns demonstrate the matching behavior without uploading any business file. Load the saved recipe if you opened a fresh workspace.
The current beta is local and manual. Batch processing, unattended scheduling and team approval are not available. Treat repeat use as a way to test whether the saved instructions actually save effort, before asking for automation.