"Metadata" is one of those terms that gets used constantly but rarely explained clearly. Most people have a vague sense that it means "data about data," which is accurate but not particularly useful.
Here's a practical explanation of what metadata is, why it matters for data quality, and what you should actually do with it.
What Is Metadata?
Metadata is information that describes, contextualizes, or classifies other data. It tells you what the data is, where it came from, when it was created or modified, who created it, and how it should be interpreted.
A customer record in your CRM has data fields (name, email, phone, company). That same record also has metadata: when was it created? Which user entered it? When was it last modified? Which data source imported it? Is this record marked as verified or unverified?
Sohovi finds gaps, duplicates, and format errors in your CRM data — so your team is working from records they can trust.
Metadata doesn't describe the customer — it describes the record itself.
The Three Types of Metadata You'll Encounter
Structural metadata — Describes the structure of a dataset: field names, data types, allowed values, relationships between tables. A database schema is structural metadata.
Descriptive metadata — Describes what the data contains or represents: what does the "status" field mean? What are the valid values for the "region" field? This is what a data dictionary captures.
Administrative metadata — Describes the lifecycle of the data: when was it created, who owns it, when was it last updated, what's its access policy, what are its retention rules?
Why Metadata Is Critical for Data Quality
Here's the practical connection: most data quality problems are metadata problems in disguise.
Sohovi shows you exactly what is wrong with your data — completeness gaps, type mismatches, duplicates — in one clear report.
You can't validate what you haven't defined. If there's no metadata saying that the "phone" field must contain a 10-digit number in E.164 format, you have no standard to check against. Data quality rules are, fundamentally, metadata — they're assertions about what valid data looks like.
You can't find problems without provenance. When a field has inconsistent formats, the fix depends on knowing where the data came from. If your metadata says "this column was imported from System A on March 15th," you know exactly which import to re-run or correct.
You can't trust what isn't documented. A dataset with no metadata — no field definitions, no source information, no ownership — is a dataset you can't fully trust, because you can't verify your interpretations of it.
Metadata and Data Quality in Practice
A few specific ways metadata gaps cause quality problems:
Unnamed or poorly named fields. A column called "dt" could mean date entered, date of transaction, date of last touch, or something else entirely. Without metadata explaining the field, different users interpret it differently and populate it differently.
Missing data types. A field that's supposed to hold integers but has no enforced type will quietly accept text values, which corrupts aggregations and sort orders.
No last-modified tracking. Without administrative metadata tracking when a record was last updated, you can't assess timeliness — one of the core data quality dimensions.
Sohovi measures all 10 data quality dimensions — completeness, validity, uniqueness, accuracy, consistency, and more — automatically across every column.
No source tracking. When you import data from multiple sources and merge them, records from each source may have been collected differently. Without source metadata, you can't trace quality problems back to their origin.
Practical Steps for Better Metadata
You don't need a formal metadata management platform to start. Three practical steps:
1. Create a simple data dictionary. For your most important datasets, document each field: what it means, what format it expects, what the valid values are, who owns it. A shared Google Sheet works. The discipline matters more than the tool.
2. Add source and timestamp fields to every import. When you import data, add a column for the source (e.g., "CRM export 2024-03-15") and the import date. This gives you the minimum provenance to investigate quality problems.
3. Define and document your quality rules. Every data quality rule you apply (email must be non-null, phone must match this pattern) is metadata. Write it down. When a rule changes, update the documentation.
The Bottom Line
Metadata is the foundation of data quality management. You can't reliably measure quality without definitions. You can't trace problems without provenance. You can't maintain standards without documentation. The businesses that have the cleanest data are usually the ones with the best-maintained metadata — they know what their data means, where it came from, and who is responsible for it.
