Understanding your baseline
When TagWatch says "purchases are below baseline today", what does that mean? This article explains the idea in plain language. No data science required.
What is the baseline?
The baseline is TagWatch's answer to one simple question: how many events do we normally get on a day like today?
Think of it like a shop manager who has been working at the same store for a month. On a Tuesday morning they say "we usually sell about 80 pairs of shoes on a Tuesday. If we sold only 20 by closing time, something is wrong." That mental model is exactly what TagWatch builds for your events, automatically, in the background.
A real example: the purchase event
Imagine you run a fashion webshop. Your GA4 tag fires a purchase event every time someone completes a checkout. Over the last four Tuesdays, you had roughly 120, 135, 118, and 127 purchases. TagWatch averages these out and decides: on a Tuesday, ~125 purchases is normal.
This Tuesday, your developer accidentally unpublished the GTM container at 10 am. By end of day you only have 18 purchases. TagWatch compares 18 actual against 125 expected, sees an 86% drop, and fires an alert. This happens long before your morning report the next day would have caught it.
Gray bars = expected baseline. Colored bars = actual. Red = alert threshold crossed.
The threshold: setting your alarm level
Not every dip is worth an alert. Traffic fluctuates naturally. A rainy day, a competitor running a flash sale, a bank holiday. None of these are emergencies. You do not want an alert at 11 pm because purchases are 8% lower than usual.
That is what the alert threshold is for. When you create a volume monitor you pick a percentage. The default is 80%. This means: "only alert me when the count drops more than 80% below what we normally see." A day that empties out completely is caught anyway, by the flatline rule, whatever the threshold says.
EXAMPLE: purchase event, Tuesday baseline 125
Why Monday gets compared to Monday
Most websites have a strong weekly rhythm. A B2B SaaS product might get most of its demo request form submissions on Tuesday and Wednesday, and almost none on Saturday. A consumer fashion brand is the opposite. Weekends are busy and Mondays are slow.
If TagWatch compared this Saturday against last Tuesday, the baseline would be wildly wrong. Instead, it only compares like for like: today's Monday against the last four Mondays, today's Friday against the last four Fridays. That way a quiet Sunday never triggers an alert just because your previous peak was a busy Friday.
B2B SOFTWARE: demo requests
Weekdays busy, weekends near zero. Saturday's baseline is ~3 requests, not 80.
FASHION WEBSHOP: purchases
Weekends are peak. Monday's baseline is ~45 purchases, not 95.
The first four weeks: building up the baseline
When you add a new event to a monitor, TagWatch needs history before it can compare. Flatline detection starts after the first days of data. Percentage drop alerts wait forfour weeks of same weekday data, because that is when the baseline becomes reliable.
Think of it like a new employee. On their first day, they have no idea what "a normal Tuesday" looks like yet. After a month, they have a solid feel for the rhythm of the business and can confidently say when something looks off.
What to do when an alert fires
You get an email (or Slack message) saying purchases are down 70% from the baseline. Work through this checklist:
Check if your tracking tag is still live
Open your website, right-click, choose Inspect, go to the Network tab, and reload the page. Can you still see your GTM or GA4 tag loading? If not, your tag broke. That is likely the cause.
Check if your campaigns are still running
No campaigns means no traffic means no events. Look at your ad platform. Did a budget run out? Was a campaign paused? A sudden drop in ad spend directly causes a drop in conversions.
Check for a website outage
If your checkout page is down or returning an error, purchases will go to zero immediately. Try completing a test purchase yourself, or check your uptime monitoring tool.
Check if there is a legitimate business reason
Sometimes there is a simple explanation: a public holiday in your main market, a product going out of stock, or a promotional period that just ended. If the dip makes business sense, you can resolve the incident manually in TagWatch.
Common questions
Q: Will I get a false alarm during Black Friday?
Black Friday and Cyber Monday send purchase volumes through the roof. Sometimes they are 5 to 10 times above normal. TagWatch will not alert you for events that are higher than the baseline, only lower. So a spike is fine. However, in the weeks after Black Friday, the baseline may temporarily be elevated by those peak days. Keep this in mind when reviewing alerts in late November and December.
Q: My event fires once a week. Does the baseline work?
Volume monitors work best for events that fire at least 10 to 20 times a day. If your event fires only a handful of times per week, the baseline is too uncertain to be useful. A single missed order would look like a 100% drop, and percentage alerts stay paused below an average of 10 events a day. Use a Pixel monitor instead to verify that the tag itself is loading correctly.
Q: I just launched. I have no historical data yet.
That is completely fine. Add your monitor now and let it collect data. Flatline protection starts within days, and percentage alerts join once four weeks of history exist. TagWatch never alerts on events without enough history.
Q: The baseline looks too high. We ran a big promo last month.
The baseline uses a rolling four week window, so an unusual promotion from six weeks ago has already dropped out of it. If a recent promotion is still in the window, you may see slightly elevated baselines for the weeks immediately following the promo. This is expected and settles on its own.
Key things to remember
The baseline compares like for like
Monday is compared to the last four Mondays. Friday to the last four Fridays. This automatically handles your weekly traffic pattern.
The default 80% threshold is a good starting point
It catches real breakage and ignores normal fluctuation. Lower it for an event whose count is stable enough that a smaller dip already means something.
New monitors need a few weeks to learn
Flatlines are caught within days; percentage alerts wait for four weeks of same weekday history. This is by design and prevents false alarms in week one.
Spikes do not trigger alerts. Only drops do.
If your events are up, something good is probably happening. TagWatch only alerts when things go down far enough to suggest breakage.