Cookie settings

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

Skip to content
Skip to content

Compliance evidence. Internal audit.

Audit the data itself, not only the spreadsheet describing it.

IRIS records what each dated scan covered, found and couldn’t read, so internal audit can test the data part of an audit before fieldwork.

Four paid engagements and three proofs of concept.

The challenge

What makes audit data hard to test

The declared list

Dipal Panchal, co-founder and CEO, describes his last enterprise role this way: each quarterly audit found sensitive data that the declared list had missed.

Answers typed by each owner

As Dipal Panchal describes it, the chief information security officer (CISO) forwards each audit question to the application owners, who answer by hand, and the auditor files the emailed replies.

No record of what was skipped

A folder of exports shows what someone found, but not which systems nobody looked at, so your auditor has nothing to test completeness against.

In the product

The evidence behind seven audit requests

Sample report as a request list. Five rows come from an IRIS scan of a database on October 26, 2026. Row 3 comes from outside IRIS; row 7 needs a later scan.

The evidence behind seven audit requests
Auditor’s requestEvidenceProduced by, and whenHow your auditor tests it
Which systems did you check?Report section 1: enterprise_data, 2 schemas (iris_dataset, public), 16 tables, 190 columnsIRIS, at the scan on October 26, 2026Population: count the tables in those two schemas with the query below and compare the total
Is anything missing from that population?Report section 3: marketing_contacts returned 0 rows, so its columns were classified from names and types onlyIRIS, in the same scanCompleteness: test the columns classified from names alone separately, and ask the owner why the table was empty
Which sources hold data but weren’t scanned?Not in the report: your asset register, set against the sources connected to IRISYour IT team, dated before the first scan (September 18, 2026, in the sample)Completeness of the source list: every register entry is scanned, or recorded as a gap with a reason
Show me the sensitive data you foundFinding: user_credentials.api_key, credential, no protection detected in sampled values, 15 rowsIRIS, from that scan; the reviewer hasn’t decided yetAgreement: open the named table and column, and compare a few rows with the finding
Who reviewed the findings?Evidence register: classification report, generated by IRIS, under reviewYour reviewer, on the date they decideThe review control: an item under review isn’t evidence yet, so look for the reviewer’s decision on it
Does the evidence cover the audit period?Scan history: 22 completed scans of this source, September 18 to October 26, 2026IRIS, one dated record per scanPeriod: check how the scans are spaced across the year; nothing between two scans was observed
What changed after the fix?Not in the sample: a re-scan of the same scope, dated after your team changed the sourceIRIS, in a new scan dated after your team’s changeEach scan keeps its own findings, so set the column’s finding from each scan side by side

Count the tables in scope yourself

For a PostgreSQL source, the query gives your auditor a count for the first row that doesn’t come from IRIS. Run it on the scan date, or note any tables the owner created after the scan.

SQLFrom PostgreSQL documentation, page updated November 8, 2026; not yet tested by Sylox.
-- PostgreSQL 14 to 18. Tables per schema, for the schemas in the coverage record.
-- Runs as any role that can connect. Counts ordinary and partitioned tables;
-- leaves out views, materialized views, partitions and foreign tables.
SELECT n.nspname AS schema_name, count(*) AS tables
FROM pg_catalog.pg_class c
JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace
WHERE c.relkind IN ('r', 'p')
  AND NOT c.relispartition
  AND n.nspname IN ('iris_dataset', 'public')
GROUP BY n.nspname
ORDER BY n.nspname;

-- Expected: one row per schema. Compare the total with the table count in the coverage record
-- (16 in the sample report), and ask the source owner about any difference.
-- Don’t swap in information_schema.tables: it lists only the tables the role owns or holds a
-- privilege on, so an auditor’s read-only login gets a short count and no warning.
Timeline

When to have each item

Our rule for an audit with a fixed fieldwork date, so every item in the list carries a date your auditor can use.

Three months before fieldwork

Run the first scan of each source in scope. That leaves time to fix what it finds, and to ask owners about any table it couldn’t read.

A month before

Re-scan the same scope once the fixes are in, so the evidence is dated after the change. Then have your reviewers decide each new finding.

The week fieldwork starts

Export the evidence register and run the table count again. A gap between the count and the coverage record goes to the source owner first.

Between audits

Keep scans on a schedule. Next year’s audit period then has scans dated across it, which a single scan can’t give.

Overview

How internal audit tests a system report

A scan report is information produced by the entity (IPE), so internal audit tests how it was made before relying on it. The Institute of Internal Auditors’ Global Internal Audit Standards, Standard 14.1, require information that is relevant, reliable and sufficient.

In an audit of financial statements, International Standard on Auditing (ISA) 500, paragraph 9, has the auditor evaluate whether such information is reliable enough, including evidence that it is complete and correct. For US public-company audits, Public Company Accounting Oversight Board (PCAOB) AS 1105, paragraph 10, sets a similar test for information the company produces.

For a scan report, that test starts with the coverage record and the list of unread tables, before any finding. IRIS doesn’t certify the evidence: whether it is enough for the audit is your auditor’s call.

Capabilities

How the evidence is produced

Your auditor tests how a report is made, so here is where IRIS runs and what it reads.

Runs in your environment

IRIS can be deployed in your own infrastructure, on-premises or in any cloud, so your data stays within your environment.

Read access only

IRIS only reads your sources, through an account your team creates; give it a read-only account wherever the source offers one.

A sample of each table

IRIS reads a sample of each table, and small tables in full. Any table it couldn’t read is listed with the reason, and its columns stay unclassified.

Dated runs with a status

Scans run on demand or on a schedule, and each run keeps its own date and reads partial when any table in scope couldn’t be read.

What IRIS keeps

Findings and coverage records stay on your host, unless you turn on emailed reports. Stored samples are masked by default in personal-data columns; a deployment setting masks all or keeps none.

A register you can export

IRIS exports the evidence register as CSV or XLSX, with each item’s generated date, scope, reviewer decision and expiry.

Capabilities vary by edition.

Frameworks

What each law asks of your data

Each page sets out the provisions that touch the data you hold, and which IRIS evidence goes with them.

For UK GDPR, UAE PDPL, Saudi PDPL, ISO/IEC 27001, SOC 2 and PCI DSS, IRIS maps its evidence to their controls. A scored readiness view for audit work isn’t in v1; ask us if you need one.

How it works

Follow one scan from source to report

The five sample-report rows come from the October 26, 2026 run and the scan history up to it. The how-it-works page follows a scan from the connected source to the evidence, so your auditor can see where each item is written before testing it.

For the same report field by field, read the audit evidence page. Privacy and security teams read this evidence from their side: IRIS for DPOs and IRIS for CISOs.

FAQ

What audit teams ask about IRIS evidence

Protect your brand. Start with your data.

Bring the request list from your last audit. We’ll show which rows IRIS evidence answers, and which it doesn’t.