measuremy.site
measuremy.site / Blog

How to Set Up Traffic Alerts for Sudden Drops

November 5, 2026 · 5 min read

Cinematic view from a car interior driving through Manchester at night in the rain.
Photo by Nathan J Hilton on Pexels

Start with a baseline, not a guess

A useful traffic alert needs something to compare against. For most small sites, a daily visitor or pageview count over the previous 14 to 28 days is a good starting point. Use the median rather than the mean if your traffic has occasional spikes from launches, newsletters, or social posts. A single unusually large day can pull an average up and make normal days look like drops.

If your site has a clear weekly pattern, compare weekdays with weekdays. For example, Tuesday should be compared with recent Tuesdays, not with Sunday. If traffic is low, expect more day-to-day noise. A site with 30 visitors per day may vary by half without anything being wrong.

Define a drop that matters

A percentage drop alone is not enough for small sites. A 50% drop from 10 visitors to 5 visitors is often noise. A 50% drop from 500 visitors to 250 visitors is worth investigating. Use two conditions together: a percentage below the baseline and a minimum absolute difference.

A practical starting point is:

alert if today < baseline * 0.6 and baseline - today >= 20

That means alert when traffic is at least 40% below the 14-day median and at least 20 visits below that median. Tune both numbers after a few weeks. If you get too many false alarms, raise the percentage or the absolute floor. If you miss real drops, lower them.

Choose the right time window

For a small site, a daily alert is usually better than a real-time alert. Checking every hour creates noise because traffic is uneven throughout the day. An afternoon lull can look like a collapse when it is only a time-of-day pattern.

Alert on a completed day if you want stable numbers. If you need faster warning, wait until at least half the day has passed and compare the same partial period with previous days. Also be clear about which time zone defines a day. A drop may look larger or smaller depending on whether your analytics day starts at midnight UTC, your local time, or the visitor's time.

Monitor segments, not just totals

Total traffic can hide the real story. An organic search drop might be offset by a direct traffic spike from a newsletter. If you only watch the total, you miss the part that needs attention.

Useful segments to watch include:

Start with one overall alert, then add one or two segments that matter most for your business. Do not create twenty alerts on day one.

Turn the alert into a daily check

If your analytics tool has custom alerts, the basic setup is the same: choose the metric, set the condition, pick the lookback window, and choose the notification channel. Check your tool's documentation for the exact steps because menus and names differ.

If built-in alerting is limited, a scheduled daily report can serve the same purpose. Some people prefer a simple email every morning with yesterday's visitors and the 14-day median. A technical founder can also run a small script on a schedule:

yesterday = pageviews(day - 1)
baseline = median(pageviews(day - 14 .. day - 2))
if yesterday < baseline * 0.6 and baseline - yesterday >= 20:
    send_alert(yesterday, baseline)

The point is not the code. It is that the comparison runs automatically and only interrupts you when the condition is met.

Write the alert message so you can act

A bad alert says only: "Traffic dropped." A useful alert tells you where to look first. Include the site name, date, current count, baseline, percentage change, and the segment if the alert is segment-specific.

For example:

That message gives you a starting point without opening five tabs.

Test it before you need it

Do not wait for a real drop to find out the alert is broken. Lower the threshold temporarily so the condition triggers, then confirm the notification arrives in the right place. If you use email, check that it is not going to spam. If you use Slack or another channel, confirm the message is readable on a phone.

Testing also shows whether your baseline is sensible. Run the alert against a quiet day from last month and see whether it would have fired. If it would have fired on several ordinary days, tighten the threshold.

Keep it quiet enough to stay useful

Alert fatigue is the most common reason traffic alerts get ignored. If every small fluctuation sends a notification, you will eventually mute the channel or delete the rule.

Use a cooldown or deduplication so the same condition does not fire repeatedly. One alert per condition per day is enough for most small sites. Prefer a single morning summary over continuous pinging. If two consecutive days are both below the threshold, that can be a second, higher-priority alert rather than two identical ones.

Pair the alert with a short runbook

An alert is only useful if it leads to a response. Keep a short runbook separate from the alert itself. A basic version includes:

You do not need to automate all of this. You need to know what to open when the alert arrives.

With a hosted cookieless tracker like measuremy.site, the dashboard shows visitors, pages, sources, countries, browsers and systems. Point your alert message at that dashboard so the first step is reading the source breakdown rather than guessing.

Revisit the threshold

A threshold that works in March may not work in July. Traffic changes with seasons, product launches, publishing frequency, and search behavior. Review the baseline and alert condition once a month or once a quarter. If your normal traffic has grown, an old absolute floor may hide real drops. If traffic has settled lower, the alert may fire too often.

A traffic alert is not a set-and-forget setting. It is a small monitoring loop that should stay close to the actual pattern of the site.

A good sudden-drop alert has a realistic baseline, a threshold that fits your traffic level, a segment that points to the cause, and a quiet notification channel you will actually read. Start simple, test it, and adjust the threshold until it only interrupts you when something is genuinely worth your time.

See it on your own site

measuremy.site is cookieless analytics your coding agent can read. 30-day free trial, no card. Plans from $9/month.

Start free trial See the live demo

Add measuremy.site analytics to this project: connect the measuremy.site MCP server at https://a.measuremy.site/mcp for all my projects (claude mcp add --scope user; it signs in with OAuth, no key, and one sign-in covers every site), then tell me how to sign in. Details: measuremy.site/llms.txt

Free tools

All free tools

More from the blog