Data profiling for CSV and Excel — see every column before you trust it.

Before you can score, clean, or build rules against a dataset, you need to understand what's actually in it. Sohovi's data profiling tool reads a CSV or Excel file and, in seconds, tells you exactly what each column contains — without you writing a single formula or query.

What gets profiled

Drop in a file and Sohovi computes, per column: null rate and non-null count, distinct value count and cardinality, inferred data type (email, phone, date, numeric ID, free text), the most frequent values, and statistical outliers using a ±3σ threshold on numeric columns. Wide files with 100,000+ rows are automatically sampled for the interactive profiling view, while your DQ rules still run against every row.

Built-in PII detection

Every profiling pass also flags columns that look like they contain personal data — emails, phone numbers, SSNs, credit card numbers, street addresses, or API keys and secrets — using pattern matching combined with entropy analysis for opaque tokens. You see this before you export, share, or build a rule against the column, so sensitive data never accidentally ends up in a report you send outside your team. Read more on our PII detection page.

Runs entirely in your browser

Profiling happens inside a Web Worker in your browser tab — your file is never uploaded to a server. That means profiling a spreadsheet full of customer records doesn't create a new copy of sensitive data anywhere outside your machine. Only the aggregated profiling summary (not your raw rows) is ever saved, so you can come back and compare profiles across runs. Full details on how this is engineered are on our security architecture page.

From profile to rule to score

Profiling is the first step in Sohovi's workflow: once your columns are profiled, Sohovi's ML rule suggester proposes quality rules based on the inferred type of each column, and every rule you accept feeds into a full 10-dimension DQ score you can track over time.