Guide. Audit evidence. Coverage.
How to build a source coverage register an auditor can reconcile
One row per source in scope, with a status, reason, owner and date, plus the two-way check that shows the list itself is complete.
Co-founder and CEO · Updated November 8, 2026

On this page
The register, the inventory and the rule between them
Applies to any program that scans data sources, whatever the tool. The two queries below list databases on PostgreSQL 16 to 18 and SQL Server 2022. Sylox ran the PostgreSQL query in its lab. The SQL Server query comes from vendor documentation that Sylox has not yet run.
Rule: every source on the inventory has exactly one register row, and every register row traces back to the inventory. The register is complete only as far as the inventory behind it, so take the inventory from a list the scan team did not write.
The scan owner keeps the register. Source owners supply their rows, and a reviewer outside the scan team does the reconciliation. Seeding the first rows takes about half a day for a small estate, once the inventory exists.
- Pull the inventory from an independent list
- Give each source one row, an ID and an owner role
- Set one of five statuses, defined in writing
- Reconcile the register to the inventory in both directions
- Seal the version when the scan cycle closes
The fields every row carries
Each field answers a question the auditor will ask about that row.
| Field | What to write | What it answers |
|---|---|---|
| ID | A stable number, such as SRC-01, never reused | Which row is this, in this cycle and the next |
| Source | System, environment and instance, such as billing, production, host B | Exactly which thing was or was not read |
| Status | Scanned, partial, failed, excluded or not connected | How far the findings from this row reach |
| Reason | One sentence for anything other than scanned | Why the gap exists, in words a stranger follows |
| Owner | A role, not a person, with a named deputy | Who answers for the row and fixes it |
| Scan date and cycle | The date the source was read and the cycle it belongs to | Which period the row describes |
| Next action and due date | The step that changes the status, and when | Whether the gap is being closed |
A register for one scan cycle
One organization and one scan cycle, with owners shown as roles. Three rows cannot support a clean result for the source they describe.
| Source | Status | Reason | Owner |
|---|---|---|---|
| SRC-01 Customer database (PostgreSQL) | Scanned | All schemas read, a sample of each table; scanned November 2, 2026 | Head of customer platforms |
| SRC-02 Billing database (SQL Server) | Partial | Archive schema denied, listed as a gap; grant due for the re-scan on November 20, 2026 | Finance systems lead |
| SRC-03 Human resources file share | Excluded | Mid-migration; approved October 28, 2026, review January 15, 2027 | HR director |
| SRC-04 Analytics warehouse | Failed | Network path from the scan host blocked; re-run due November 13, 2026 | Data platform lead |
| SRC-05 Legacy orders database (Oracle) | Not connected | No read-only account yet; request raised, account due November 27, 2026 | Operations lead |
| SRC-06 Marketing shared drive | Partial | Some files unreadable, listed with reasons; owner reviews them | Marketing operations lead |
What each status lets you claim
Read the register the way an auditor reads it. In this register, one source was read in full, two in part, and three produced no findings at all. Results from this cycle say nothing about the warehouse, the human resources share or the legacy orders database.
NIST SP 800-53A describes a coverage attribute that addresses the scope or breadth of the examination, interviews and testing in an assessment. Where an assessor cannot obtain enough information, the finding for that part is other than satisfied. The status column is how you tell them first.
List the databases on a PostgreSQL server
This is the PostgreSQL half of the inventory check. Run it on every server in scope and keep the output with the register.
-- Read-only. Any login that can connect; pg_database is readable by everyone.
SELECT datname AS database_name,
pg_catalog.pg_get_userbyid(datdba) AS owner,
datallowconn AS accepts_connections
FROM pg_catalog.pg_database
WHERE NOT datistemplate
ORDER BY datname;
-- Completeness anchor: the number of rows, to compare with the inventory for this server.
SELECT count(*) AS databases_listed
FROM pg_catalog.pg_database
WHERE NOT datistemplate;List the databases on a SQL Server instance
The same check for SQL Server. A restoring database is listed with its state, but an offline one is hidden from a login without enough rights, so use a login that sees every database.
-- Read-only. Databases the login may see; VIEW ANY DATABASE is granted to public by default.
-- database_id 1 to 4 are the system databases (master, tempdb, model, msdb).
SELECT name AS database_name,
SUSER_SNAME(owner_sid) AS owner,
state_desc,
create_date
FROM sys.databases
WHERE database_id > 4
ORDER BY name;
-- Completeness anchor: compare with the inventory for this instance.
SELECT COUNT(*) AS databases_listed FROM sys.databases WHERE database_id > 4;The reconciliation, in both directions
Run it once per cycle, before sealing the register. The right-hand column shows what it finds on the register above.
| Direction | The check | What it finds here |
|---|---|---|
| Inventory to register | Every database from the queries above has a register row | orders_archive on the billing host has no row; add SRC-07 as not connected |
| Inventory to register | Every SaaS app on the finance and procurement lists has a row | The support desk tool is on the finance list only; add SRC-08 as not connected |
| Register to inventory | Every row traces to an inventory entry | SRC-05 matches the inventory but is on the decommission list; ask its owner |
| Register to scan results | Every scanned or partial row has results dated in this cycle | SRC-06 has results dated in the last cycle; mark it partial until re-read |
| Statuses to evidence | Every excluded row has the owner’s reason attached | SRC-03 has the approval and the review date; the exclusion stands |
Seal each version and keep the series
When the cycle closes, the scan owner signs the register and the reviewer countersigns the reconciliation. Store the sealed file read-only, with the cycle ID in its name, and open a new version for the next cycle instead of editing this one.
Keep a change log for rows altered after sealing: who changed what, when and why. NIST SP 800-53A asks that when earlier assessment results are reused, the date and type of the original assessment be documented. A register series gives you that for free.
Open the findings report with the coverage summary. State the number of sources in scope, how many were read and which were not, before any finding. Link each finding to its row.
What this method misses
A row covers one instance. Replicas, snapshots, restored copies and exports of a source need rows of their own, and the catalog queries list only databases on servers you already know about.
Sources nobody put on a list stay invisible to the reconciliation. Cross-check with network and cloud account inventories, and treat a new account as a new row. The queries also do not tell you what a database holds, only that it exists.
A status is a claim by a person. Have the reviewer open one scanned row and one excluded row in full each cycle, to check the evidence behind the word.
IRIS records each scan’s coverage for the sources you connect, with the scan date. Capabilities vary by edition.
A scan status for each connected source, with a reason for anything it could not read: IRIS produces this.
Evidence items with their source and scan date, exported as CSV or XLSX: IRIS produces this.
Sources nobody connected, such as a SaaS app: Not produced by IRIS.
An owner, due date and reminder for each gap: Not produced by IRIS.
Explore audit evidenceReconcile one server, then plan exclusions
Run both directions on one server this week. The check passes when every database on it has a row or an owner’s written reason, and no row points at a database that is gone.
For rows that end up excluded, the record that stands up in an audit is set out in how to document exclusions when a source can’t be scanned.
Official sources
- NIST SP 800-53A Rev. 5: assessment methods, coverage and reuse of results, NIST. Opened November 8, 2026.
- pg_database system catalog, PostgreSQL 18, PostgreSQL documentation. Opened November 8, 2026.
- sys.databases (Transact-SQL), Microsoft Learn. Opened November 8, 2026.
Start with a free assessment of one database you choose.
Get a free assessmentCoverage register questions
Yes. Version it by cycle, keep sealed copies read-only and log changes after sealing. Every row still needs a status, a reason, an owner and a date.
Someone outside the scan team, such as internal audit or the privacy office. The people who built the register will read their own list as complete.
Yes, and its status tells the reader why there are none: read and clean, partly read, or not read at all. A row with no findings and no status is the one that misleads.
Once per scan cycle, with a sealed version each time, and again after a significant change such as a migration or a new application going live.
Protect your brand. Start with your data.
See the coverage record IRIS produces in our sample report, or start with a free, read-only assessment.


