Scrubly
Sign inGet started

data profiling & cleaning

Find what’s wrong with your data.Fix it with rules you can read.

Upload a CSV or Excel file. Scrubly profiles every column, scores the file’s quality, and flags what needs fixing. You write the fixes as rules, or ask Claude to suggest some, and preview every change on your real file.

Start cleaning Try the demo

Free plan · files up to 25 MB · no card needed

customers.csv
A sample customers file with seven problems — stray whitespace, mixed case, currency symbols in a number column, empty plans, and a duplicate row — cleaned by five rules applied in order: trim whitespace, lowercase, find and replace, fill empty values, and remove duplicate rows.

profiling

A quality score you can check the arithmetic on.

Every column gets its type, missing and repeated values, formats, and outliers. The score out of 100 blends four parts with fixed, published weights, so a low number always points at a cause.

How the quality score is weighted
partweightmeasures
Completeness35%Share of cells that hold a value.
Validity30%Share of values that parse as their column's type.
Consistency20%How closely each column sticks to a single format.
Uniqueness15%Repeated rows, and repeats in columns that should be keys.

rules

Every change is a rule you can read.

A rule is a condition that picks rows and an action that changes them, both from a fixed catalog. Rules run top to bottom in the order you set. Nothing changes your data that isn’t written down.

conditions12

  • Is empty
  • Has a value
  • Matches pattern
  • Does not match pattern
  • Outside a range
  • Is one of
  • Is not one of
  • Greater than
  • Less than
  • Equals
  • Is a duplicate
  • Text length outside bounds

actions12

  • Trim whitespace
  • Lowercase
  • Uppercase
  • Title case
  • Convert type
  • Fill empty values
  • Find and replace
  • Map values
  • Remove duplicate rows
  • Remove matching rows
  • Clip to a range
  • Flag for review

Preview before you commit

See the first 100 rows before and after, with counts taken from the whole file: cells changed, rows removed, rows left. The preview runs the same code as the real thing, so what you see is what you get.

ai suggestions

Claude suggests rules. You decide which ones run.

Ask for suggestions and they arrive switched off, beside your own rules. Accept the ones you want; the rest never touch your file. Everything else in Scrubly works without AI.

sent

  • Column names and their types
  • How many values are missing or repeated
  • The patterns and formats found in each column
  • The quality score and its parts

never sent

  • Every row and every cell of your file
  • Any value from a column flagged as personal data

The one exception is the most common values of a small, non-personal category, like the handful of options in a status column. Requests go through OpenRouter to Anthropic’s Claude, only via providers that don’t keep or train on them. You see all of this before anything is sent.

free plan

Start with a real file today.

The free plan covers the whole product: profiling, rules, previews, and AI suggestions. No card, no trial clock.

Start cleaning Sign in
Largest file
25 MB
Rows per file
500,000
New datasets a month
20
AI suggestion runs a month
15
Uploaded file kept for
7 days
Your profile and rules stay after the file is deleted.