Website monitoring

If the checkout form stops working, the analytics code is removed by mistake or robots.txt blocks the pages, the fault can go unnoticed until someone complains or the monthly figures drop.

That is why you need systems that raise an alert when a change to the website could harm sales, measurement or visibility, whether the change is intentional or not.

With continuous monitoring, problems are detected quickly, instead of weeks later when the revenue has already been lost or search visibility has fallen drastically.

Why monitoring is critical

Detect problems quickly

The problem without monitoring: Something breaks on the website. It is only discovered when someone complains or the monthly figures show a dramatic drop.

With monitoring: Uptime monitoring and error tracking raise an alert within minutes. Google Analytics 4 (GA4) can evaluate custom insights at most once an hour. Data in Search Console normally arrives with a delay of two to three days. The problem can be fixed before it has a major impact.

Prevent lost revenue

Every hour that a critical process is down costs money.

Example

Scenario

An online shop with sales of SEK 1,000 an hour.

Problem

The checkout does not work for 24 hours.

Loss

SEK 24,000.

With monitoring, the problem is detected within 10 minutes and fixed within one hour.

Loss: SEK 1,000 (96 per cent less).

Protect SEO positions

It can take months to recover from SEO problems.

Example

Problem

Misconfigured canonical tags make every page point to the home page.

Without monitoring

The fault is discovered after two months, by which time organic traffic has fallen by 60 per cent. It takes four to six months to win the rankings back.

With monitoring

The fault is detected on day 1 and fixed on day 2. The rankings are barely affected.

Ensure data quality

Decisions are based on data. If the data is wrong, the decisions will be wrong.

Example

Problem

The conversion tracking code is broken.

Without monitoring

The reports show zero conversions. The team believes the campaigns are performing poorly and pauses them, even though the campaigns are performing well and it is the data that is not being collected.

With monitoring

An alert is sent when the number of conversions suddenly drops to zero. Troubleshooting shows that the tracking is not working. The fault is fixed before any decision is made.

What should be monitored?

Overall monitoring

Check that the website is up

Tools

Pingdom, UptimeRobot, StatusCake

Alert

Email or text message when the website is down or the response time exceeds three seconds.

Check that the traffic sources have not dried up

Tools

GA4 custom insights, custom scripts

Alert

Traffic from a source falls by more than 30 per cent from one day to the next.

Check that the traffic flow follows the forecast

Use

Identify when traffic deviates from what was expected (positively or negatively).

Alert

Traffic is more than 20 per cent below the forecast.

Check that every component works

Tools

Synthetic monitoring (simulating user flows), Selenium tests

Alert

A critical flow (checkout, for example) is not working.

Check that analytics tools have not been removed by mistake

Tools

Custom scripts that check that the tracking code is present on the pages

Alert

The tracking code is missing from critical pages.

Check that the purchase flow and checkout work smoothly

Tools

Real User Monitoring (RUM), error tracking (Sentry)

Alert

Checkout abandonment increases by more than 10 per cent.

The three types of monitoring

1. SEO monitoring

Purpose: Monitor organic visibility and technical SEO health.

Why it matters

SEO problems are often discovered too late. Rankings can fall dramatically before anyone notices.

Typical alert

“Search phrase X has fallen from position 3 to position 15, a drop of 12 positions.”

Read more about SEO monitoring →

2. AI monitoring

Purpose: Monitor visibility in AI tools (ChatGPT, Gemini, Perplexity).

Why it matters

A user who asks an AI tool instead of Google only encounters the companies mentioned in the answer. If the company does not appear in AI answers, it loses attention.

Typical alert

“The company is no longer mentioned in ChatGPT answers for [important topic], where it was previously in the top 3.”

Read more about AI monitoring →

3. Conversion monitoring

Purpose: Monitor the conversion rate and identify problems in the conversion funnel.

Why it matters

A small drop in conversion can cost enormous sums if it is not detected quickly.

Typical alert

“The conversion rate has fallen from 2.5 per cent to 1.8 per cent (minus 28 per cent) over the past 24 hours.”

Read more about conversion monitoring →

How to set up monitoring step by step

Step 1: Identify critical metrics

Not everything can be monitored. Prioritise what matters most to the business.

Example (e-commerce)

Step 2: Set up monitoring

Tools and configuration

Uptime monitoring

Pingdom and UptimeRobot, which check the website at intervals of one to five minutes.

SEO monitoring

Semrush and Ahrefs, with daily checks of rankings.

AI monitoring

Ahrefs Brand Radar and custom scripts, with weekly checks of mentions.

Conversion monitoring

GA4 custom insights, with daily checks of the conversion rate.

Error tracking

Sentry and Rollbar, in real time for JavaScript and back-end errors.

Step 3: Define thresholds

When should alerts be sent?

Threshold too low

Too many false alarms (alerts for normal variation).

Threshold too high

Problems are detected too late.

Recommendations

Uptime

Alert immediately when the website is down.

Traffic

Alert on a drop of more than 30 per cent from one day to the next.

Conversion rate

Alert on a drop of more than 15 per cent from one day to the next.

Number of conversions

Alert on a drop of more than 20 per cent from one day to the next.

SEO rankings

Alert when an important search phrase drops more than five positions.

Adjust based on experience

If there are too many false alarms, raise the threshold. If problems are missed, lower it.

Step 4: Configure alerts

How should alerts be sent?

Email

Good for non-critical alerts (SEO reports, weekly summaries).

Text message

For critical alerts (the website is down, the checkout is not working).

Slack/Teams

For teams that work in these platforms every day.

PagerDuty

For large organisations with an on-call rota.

Who should receive alerts?

Critical alerts (checkout down)

DevOps, the technical lead and the CTO.

SEO alerts

The SEO team and the marketing manager.

Conversion alerts

The product team and the marketing department.

Step 5: Create action lists

When an alert comes in, what should be done?

An action list is a troubleshooting checklist.

Result: The problem is fixed faster. The team knows exactly what to do.

Step 6: Test regularly

Monitoring can stop working, and without regular tests nobody notices.

Step 7: Review and adjust

Monitoring tools

Uptime and performance

SEO monitoring

AI monitoring

Conversion and error tracking

All-in-one monitoring

Common monitoring scenarios

Scenario 1: The website is suddenly down

Alert: Pingdom sends a text message saying “The website is down”.

Goal: The website back up within 30 minutes.

Scenario 2: Organic traffic falls by 40 per cent

Alert: GA4 custom insight “Organic traffic minus 40 per cent compared with the previous day”.

Goal: Identify the cause within 24 hours and produce an action plan within 48 hours.

Scenario 3: The conversion rate halves

Alert: “Conversion rate down from 2.5 per cent to 1.2 per cent (minus 52 per cent).”

Goal: The problem identified and fixed within two hours.

Scenario 4: An important keyword has dropped 10 positions

Alert: “Keyword X has fallen from position 3 to position 13 (minus 10).”

Goal: An action plan within a week and the position regained within one to three months.

How to get the most out of monitoring

Prioritise critical metrics

Do not monitor everything. Focus on what affects the business most.

The 80/20 rule applies: a few metrics affect results most, so pick those out and monitor them.

Avoid alert fatigue

If there are too many alerts, people start ignoring them, including the important ones.

Document alerts and actions

Alert log: Document every alert, what was found and how it was fixed.

Test regularly

Monitoring systems can stop working.

Once a quarter: simulate a fault and confirm that the alert arrives.

Continuous improvement

How to handle an alert from start to finish

An alert needs to lead to action.

Process

1. Alert sent → Problem detected
2. Verify → Is the problem real?
3. Diagnose → What is the cause?
4. Fix → Correct the fault
5. Document → Log what happened and how it was fixed
6. Prevent recurrence → Can the problem be prevented in future?

Contact us about website monitoring

Carnaby helps companies set up monitoring of uptime, SEO, AI visibility and conversion, so that problems are detected before they affect results.

Contact us →