Cookie settings

Necessary, always on: they keep the site working and remember your cookie choice.

Skip to content
Skip to content

Glossary. Scan coverage.

What is scan coverage, and why isn’t “not measured” zero?

Coverage is the denominator under every finding: what was read, what failed and what was left out, each with its reason and owner.

By Dipal Panchal, co-founder and CEO · Updated November 7, 2026

What the scan read, and what it didn’t

Scan coverage is the record of what a sensitive data scan read and what it didn’t, by source, table and file, with the reason for every gap. Unlike sampling depth, which says how much of each table was read, coverage says which objects were read at all. A source that wasn’t read never counts as clean.

The word has a precedent: NIST’s glossary defines coverage, for security control assessments, as an attribute of an assessment method that addresses the scope or breadth of the objects assessed.

A coverage record for one scan

Adapted from Test Org 02, Sylox’s demo bank estate. Four of the six rows can’t support a scan’s “nothing found”, and the other two support it only for what the scan could see; the excluded share rests on its migration record instead.

Records
SourceStatusReasonOwner
Core banking databaseScannedAll three tables readCore banking lead
Cards databasePartialcard_customer read; card_accounts timed out, so its columns are unclassifiedCards platform team
CRM databaseScannedcrm.person readCRM owner
Loans archiveFailedHost unreachable, so nothing was read; its count is “not measured”Loans operations, to restore the connection
Branch document scans (cloud storage)Partial12 password-protected PDFs listed by name and not readBranch operations, to supply the files unprotected
Marketing file shareExcludedMigration record attached: the share was emptied on June 30, 2026Marketing operations; exclusion approved by the CISO on July 2, 2026

Five statuses, and what each lets you say

These five are a reporting vocabulary: a scan marks each source completed, partial or failed, and owners supply the exclusions and the unconnected sources.

Five statuses, and what each lets you say
StatusWhat you can sayWhat you can’t say
ScannedFindings hold for the tables read, within the sample, as of the scan dateThat nothing exists outside the sample, or in copies made after the scan date
PartialFindings hold for the tables, files or parts of files that were readThat the unread tables, files or parts are clean
FailedThat the source exists and is in scopeAnything about what it holds
ExcludedWhy it was left out, who decided and whenThat it holds no sensitive data, unless the owner’s evidence shows it
Not connectedThat it is in scope and nobody has connected it yetAnything about what it holds

Why “not measured” is never zero

A scan that couldn’t open a source has no result for it. Counted as clean, the loans archive in the record above would vanish from the gaps and the total would look better than the estate is.

“Whatever is linked, I can scan,” said Dipal Panchal, Sylox’s co-founder and CEO, on an advisory call in 2026. A coverage report should say how many sources are in scope, how many are connected and how many were scanned, and that the findings come only from the last group. The count is only as complete as the in-scope list, so take that list from an independent register and record where it came from.

Breadth and depth are two separate claims

Coverage is breadth: which sources, tables and files were read. Sampling is depth: how much of each was read. A database scan can reach every table in scope and still read a sample of rows from each one.

State both in a report: “all tables, sampled” and “some tables, read in full” are different claims.

Where a scanned source can still hide data

A source marked scanned was read. That says nothing about how well each value was classified, and a card number split across two columns can pass without a match.

Coverage of a source also isn’t coverage of its copies. An export taken from the core banking database to a laptop is outside the record until that location is in scope. A scan also sees only the tables its account can see, so a table it has no privilege on may not be listed at all.

Report every count with its denominator

Our rule: write every finding count as “found in X of Y sources read, with Z not measured”, and never add a failed, partial or unconnected source to the clean total. Applied to the six-row record above, the sentence counts two rows read in full, two read in part, one not measured because it failed and one excluded by its owner.

Record three things beside every exclusion: the owner’s name, the reason with its evidence, and the date. Without them there is nothing for an auditor to test. Each record covers its own scan date, so an audit period needs one record per scan.

With IRIS

In the sources you connect, IRIS reads a sample of each table, and small tables in full, and records each scan’s coverage with its date. Capabilities vary by edition.

A scan status for each source in a scan: IRIS produces this.

Tables and files it couldn’t read, listed with the reason: IRIS produces part of this, for the tables and files the scan account can list.

Unknown outcomes shown as unknown: IRIS produces part of this, for the sources and tables in a scan.

SaaS apps and other systems nobody connected: Not produced by IRIS.

A task, owner and due date for each gap: Not produced by IRIS.

See how a scan runs

Read the coverage before the findings

Next time a scan report lands, yours or a vendor’s, find its coverage section first. If there isn’t one, ask for each source’s status before reading a single finding, and check that every failure and exclusion is listed.

An auditor testing completeness starts from the in-scope register and traces each source to the coverage record. Audit evidence for data security explains what makes each item testable. To keep coverage per source over time, use how to build a source coverage register.

FAQ

Percentages, partial sources and sign-off

Protect your brand. Start with your data.

See a sample report: what was scanned and what couldn’t be checked come before the key findings.