Data cleaning software that proposes changes instead of making them

On one side there are free desktop tools. OpenRefine is the best known: open source, capable, and something you install and then learn to drive with its own expression language. It runs on your machine, which is good for privacy and bad for sharing the work with a colleague.

On the other side there are enterprise data quality and analytics platforms. They are thorough and they are bought through procurement, with onboarding measured in weeks, which is why teams paying for one often want an Alteryx alternative that does the prep step alone. Both are reasonable answers to a question most people are not asking. The common case is one person with one file that has to be right this afternoon.

Anyone can write a script that lowercases a column. The hard part is defending the result. When someone asks why the March number moved, the answer cannot be that the tool cleaned it.

Because every change is approved by a person and written to an audit log, you can hand over the file and the list of what changed in it. That is what makes the output usable in a report, a migration or a board pack. If your file starts life in a spreadsheet, Excel data cleaning covers what the built in commands do and where they stop, data deduplication tools cover the matching that Remove Duplicates cannot reach, and the rule by rule view of a single repair pass is on data scrubbing software. If the cleaned file is headed into an ERP, the rules for one common destination are on NetSuite CSV import.

Here is what the engine did to a real eight row customer export, the kind that comes out of a store or a billing system before it goes into QuickBooks Online or a CRM. Columns: customer, company, email, last order date, open balance, state, ZIP code and country. The faults are ordinary ones: the same customer twice, a name exported as Marco Díaz, balances typed as $1,284.50, 450.5 and $2,310, dates in four formats, states written Maine, N.J., minn and Mass, a ZIP code that lost its leading zero, and USA, US, U.S.A. and United States in one column.

It proposed 33 field changes and two duplicate groups, flagged one email it would not guess about, and changed nothing until each proposal was approved. The breakdown: 8 currency values, 7 country spellings, 6 dates, 6 states, 2 ZIP codes, 1 encoding repair, 1 name in lowercase, 1 email in capitals with a trailing space and 1 name with stray spaces. Run on 2026-10-19 with the same engine the demo uses.

Most buyers are not choosing between products. They are choosing between kinds of product, and the right one depends on where the data lives and who has to sign off on the result.

The buyer is rarely a data engineer. It is the controller whose customer master has to reconcile before month end, the RevOps manager about to import an event list into the CRM, or the consultant running a QuickBooks or NetSuite migration who has to show the client what was changed in their data. Each of them has the file today, a deadline this week, and a person who will ask why a number moved.

That is why the review step and the audit log come first here. The software proposes, a person approves, and the log of every decision goes out with the cleaned file. If the job is a whole contact database rather than one export, CRM data cleansing covers that workflow, and if the question is buying a service instead of software, data cleaning services sets out when each makes sense.