SEO risk register

Audience: Merchant·Plan: starter·Read: 7 min·Updated: 2026-08-03

In 30 seconds

• One backlog across four sources: crawl issues, broken links, keyword revenue risk and SLA breaches — ranked together rather than four separate lists.
• A 0–100 exposure score built from four capped bands, so no single dimension can pin it at 100 and hide the others.
• Items are read from the source tables, with assignment and resolution as an overlay — so the register never claims a backlog is clear because nobody has triaged it yet.
• The register says when it is truncated and when a section failed to load. An error is never rendered as a zero.
• Crawl issues are scoped to your latest completed crawl, because issues are regenerated per crawl and summing every crawl triple-counts.

How to use it

1. Read the exposure score and its four bands. The bands tell you which dimension is driving it.
2. Work the list in priority order. It is already ranked across all four sources.
3. Assign items to the person who will fix them.
4. Watch the trend for whether the score is actually moving.

⏱ Weekly review · 💳 Starter+ · 🎯 One backlog, honestly counted

What feeds it

Source	What counts as an item	Priority rule
Crawl issues	Findings from your most recent completed crawl	The crawler's own severity, with info collapsed to low
Broken links	Open broken links found on your site	5xx is critical; a 4xx is high only when it demonstrably costs something — attributed revenue, organic clicks or real 404 hits; otherwise medium
Keyword risk	Tracked keywords carrying a risk flag	Set by the keyword risk sweep
SLA breaches	Items past their due date	Escalated

The broken-link rule is worth understanding. A dead URL nobody visits and nothing links to is a housekeeping item. A dead URL with organic clicks against it is lost revenue. Grading both as "broken" produces a list where the important row is buried behind two thousand unimportant ones.

Source rows, not work items

The register reads the source tables and treats assignment and resolution as an overlay on top of them.

Why this matters

Work items in this suite are created lazily — a row exists only once a human assigns, snoozes or resolves something. A register that queried work items therefore reported "all resolved" on a store with thousands of open crawl issues and broken links, because nobody had triaged any of them yet.

Reading the sources means the register reflects reality on day one, and matches what the Crawler and Broken links tabs already show you.

An item leaves the register when its overlay reaches a terminal state — resolved or dismissed. Anything else, including no overlay at all, leaves the underlying row live in the register.

Assignment is mirrored back to the source row, so an item you own here does not appear unowned on the tab it came from.

The exposure score

The score is 0–100, built from four capped bands:

Band	Contribution	Reaches its cap at
Critical crawl and link issues	up to 40	10 issues
High crawl and link issues	up to 20	20 issues
SLA breaches	up to 20	5 breaches
Keywords at risk	up to 20	20 keywords

Capping each band is the point. An uncapped formula against a real technical backlog — a live store carries thousands of open crawl issues — pins at 100 on the first crawl and never moves again, which tells you nothing. With bands, each dimension retains visible headroom and the score responds to work you actually do.

One-time step in the trend

Snapshots written before August 2026 used an earlier, uncapped formula. The exposure trend therefore has a single step at the changeover. Comparisons after that point are like for like.

Scoping and counting

• Crawl issues are scoped to the latest completed crawl. Issues are regenerated on every crawl, so summing across jobs counts the same problem once per crawl that found it.
• Suppression rules are applied in the query, not after fetching, so counts over a 50,000-issue crawl are exact rather than exact-over-the-first-page.
• Truncation is disclosed. When the list shows the worst N of a larger set, it says so.
• An empty register explains itself. "No crawl has completed yet" and "no risks found" are different states and are shown differently.
• A section that failed to load shows a banner, not a zero. A failure that renders as zero is indistinguishable from good news.

Trend

The register snapshots itself daily, so the exposure score plots over time. A trend line needs at least two points — a store that has only ever taken a manual snapshot will see the chart appear on the second day.

Plan tiers

Capability	Free	Starter	Pro	Agency	Enterprise
Unified register across all four sources	—	✓	✓	✓	✓
Exposure score + band breakdown	—	✓	✓	✓	✓
Assignment and resolution	—	✓	✓	✓	✓
Suppression-aware counts	—	✓	✓	✓	✓
Daily exposure trend	—	✓	✓	✓	✓
Keyword revenue risk band	—	—	✓	✓	✓

Keyword revenue risk depends on the Search Console connection and the keyword scoring pipeline.

Best practices

• Work critical before quantity. Ten critical issues cap a band that two thousand medium ones cannot touch.
• Assign, do not just resolve. An unassigned item that breaches its SLA still counts against you, and there is no way to dodge that by leaving it unowned.
• Dismiss with a reason. A dismissal is a decision you are recording; six months later the reason is the only thing that makes it reviewable.
• Re-crawl before declaring victory. The register reflects the last completed crawl — fixes are invisible to it until the next one runs.
• Check the trend monthly. Day-to-day movement is mostly crawl variance.

See also

• Site crawler — where crawl issues and the remediation workbench live
• Site audit — the page-level audit and the broken-link worklist
• Technical health — regression incidents and their own SLA clock
• Keyword portfolio — where keyword risk flags are set