How a Million-Pageview Media Site Catches Editorial Bugs in 5 Minutes

The team and assumptions

Imagine a B2B-focused owned media property, "B Media," with 1,000,000 monthly pageviews. They publish about 50 articles a month, run by an editorial team of 7 plus one part-time engineer. The CMS is in-house, images are served via Cloudflare R2, and ~60% of revenue-driving traffic comes from Google organic search.

For a site like this, the scariest outage isn't "the homepage is down" — that gets noticed immediately. The real risk is editorial mistakes that quietly drop you out of Google, or broken images that no one catches for hours. Miterl was deployed specifically to catch this category of "silent" failures within 5 minutes.

Three editorial-side incidents you actually see

1. Accidental noindex tag in production

A staging-only <meta name="robots" content="noindex"> got copy-pasted into a production template tweak. Nobody noticed for a week. Search rankings cratered. Recovery took over a month.

2. Image CDN signed-URL expiry

Signed URLs for image delivery expired silently. Whole categories of articles started rendering with broken images. Nobody on the team noticed until users started tweeting that the site was broken.

3. Post-launch 404s

A URL scheme migration shipped without one of the redirect rules, leaving old URLs returning 404. Google traffic to those articles vanished overnight, but the homepage was fine, so basic ping-style monitors saw nothing wrong.

A three-layer monitoring setup

Layer 1: DOM Integrity (noindex detection)

The homepage, category pages, and the top 5 articles are watched with DOM Integrity monitors checking:

• noindex meta is absent
• canonical points at the right domain
• og:image is non-empty
• The <title> element contains the brand suffix (| B Media)

The moment an editor accidentally publishes a noindex page, an alert fires into Slack #editorial-alerts within 5 minutes.

Layer 2: Keyword monitoring (per-article survival check)

The top 20 articles by traffic are individually registered as keyword monitors. Each one checks that a distinctive phrase from the article (e.g. "2026 SaaS market forecast") still appears in the rendered HTML. This catches accidental unpublish, URL changes that break the article, and bad redirects — at the article level, not the site level.

Layer 3: Image CDN liveness

A representative URL pattern from the image CDN (e.g. https://cdn.b-media.com/sample.webp) is watched with 1-minute HTTP checks. CDN outages, expired signing tokens, and rate-limit blocks all surface immediately.

Splitting alerts so editors aren't paged at 3 AM

Editors and developers want very different alerts. Splitting them is non-negotiable.

Miterl's Alert Contact feature lets each monitor pick its own notification destination, so editors never get paged for a 3 AM CDN blip.

Monthly review cadence

The first Monday of each month, the editorial team reviews the Miterl monthly report (PDF) together:

• Total downtime and SLA achievement
• DOM Integrity violations and their root causes
• Response time trends

The granularity ("last month's noindex incident traced back to a copy-paste from the staging template") feeds directly back into editorial workflow improvements — required PR reviews, separated templates, and so on.

CMS deploy windows

For larger CMS deploys, the deploy script wraps work in a maintenance webhook to suppress false alerts.

# Before deploy
curl -X POST https://miterl.com/api/v1/webhooks/maintenance/$TOKEN/start \
  -H "Content-Type: application/json" \
  -d '{"duration_hours": 2, "name": "CMS upgrade v3.4.0"}'

# Deploy + smoke tests
./deploy_cms.sh
./run_smoke_tests.sh

# Resume monitoring
curl -X POST https://miterl.com/api/v1/webhooks/maintenance/$TOKEN/end

DOM Integrity rules are strict enough that template churn during a deploy almost always triggers false positives, so suppression is essential.

Related reading

• Miterl Documentation — monitor types and setup
• How to use DOM Integrity monitoring — noindex incident prevention
• Pricing — Plus and Pro plan comparison
• All use cases — playbooks for other industries