Anomalies

The Anomalies page is your early warning system. leadmaps quietly watches the numbers that matter to you, learns what is normal for each one, and raises an alert the moment something moves in a way it did not expect. A sudden drop in signups, a spike in errors, a quiet day that should have been busy: instead of you checking every chart by hand, leadmaps checks them for you and only speaks up when there is something worth your time.

Think of it like a smoke alarm in your house. You do not stand in the kitchen all day watching for a fire. You let the alarm watch for you, and it only goes off when smoke actually appears. The Anomalies page is the same idea for your data. It stays quiet when things are normal, and it gets loud when they are not.

An “anomaly” just means a number that is far from what was expected. leadmaps looks at the recent history of a number, works out the range it usually sits in, and flags any point that falls well outside that range. It is not a fixed rule like “warn me if signups drop below 100”. It learns the normal shape of each number, so a quiet Sunday does not look like a problem, but a quiet Tuesday that should have been busy does.

What you see

The page is a list of alerts, newest at the top, plus a few controls to shape it.

• A page title that reads “Anomalies”, with the date range you are looking at shown nearby, for example “Last 30 days” with the two calendar dates spelled out.
• A list of alert cards. Each card is one moment when a watched number moved unexpectedly. On each card you see:
  • Which number it is. The name of the metric that moved, for example “Signups”, “Sessions”, or “Checkout errors”.
  • When it happened. The day, and where it helps, the time of day the unexpected move was detected.
  • What direction it went. Whether the number jumped up or fell down, usually shown with an up or down arrow and a color. A drop in a good number (like signups) is shown as a warning. A spike in a bad number (like errors) is also a warning.
  • How big the change was. The size of the surprise, shown as the actual value next to what was expected, and often as a percent. For example “412 signups, expected about 950” tells you both the real number and the gap.
  • A small chart of the moment. A tiny line that shows the number around the time of the alert, with the unexpected point marked, so you can see the shape of the move at a glance.
• A quiet state. When nothing unusual has happened in your range, the list is empty and says so plainly. An empty list here is good news, not a fault. It means every watched number stayed inside its normal range.

The color on each card tells you the mood. A warning color means “this is the kind of move you probably want to look at”. leadmaps tries hard to only show you moves that are genuinely surprising, so a card appearing is meant to be worth a click, not noise.

How to read it

Read an anomaly as a short story: which number moved, which way, how far from normal, and when.

Here is a worked example. Say you open the page for the “Last 30 days” and the top card shows:

• Number: Signups.
• When: yesterday.
• Direction: down.
• Size: 412 signups, expected about 950, which is 57 percent below normal.
• The small chart shows a steady line around 950 a day for weeks, then a sharp dip to 412 yesterday.

What does this tell you? Yesterday your signups were not just a little low, they were less than half of what leadmaps expected based on the weeks before. That is a real signal, not normal daily wobble. A 57 percent drop in signups overnight usually means something broke: a signup form stopped working, a button stopped submitting, or a page started erroring. Your next move is to open the funnels page and walk your signup steps to find where people are now falling away, or open the live events stream to watch whether signups are arriving at all right now.

Now compare that to a card you can relax about. Say another card shows “Sessions, up, 8 percent above normal, on the day you sent a newsletter”. That move has an obvious, friendly cause: your own email brought extra visitors. The alert is doing its job by flagging the change, but you already know why, so there is nothing to fix. Reading the size and the timing together is what tells a real problem apart from a happy bump you caused on purpose.

One thing to keep in mind: an anomaly is a flag, not a verdict. leadmaps is telling you “this looks unusual, take a look”, not “this is definitely broken”. Some surprises are good (a viral post), some are expected (a sale, a holiday), and some are real problems. The page points you to the moment. You bring the context.

Customize it

You decide which numbers leadmaps watches, how easily it raises an alert, and the period you are reviewing. Here is each control and when to use it.

• Choose which numbers to watch. Pick the metrics that matter to your business, for example signups, sessions, purchases, or a specific error count. Watch a number when a sudden change in it would mean real money or real trouble. Leave out numbers you do not act on, so the list stays focused on what you would actually respond to. A short, meaningful watch list beats watching everything and drowning in alerts.
• Set how sensitive the alerts are. Choose how far from normal a number has to move before leadmaps raises a flag. Turn sensitivity up when you want to hear about smaller wobbles early, for a number so important that even a modest dip is worth a look. Turn it down when a number naturally bounces around a lot and you only want to know about the big, clear moves. Higher sensitivity means more alerts (and more false alarms). Lower sensitivity means fewer alerts (but you might hear about a slow problem later).
• Set the date range. Use the calendar to pick the period you are reviewing, such as the last 7, 30, or 90 days, or a custom start and end day. Use a short range like “Last 24 hours” to check whether anything broke right after you shipped a change. Use a longer range like “Last 90 days” to review the bigger pattern of surprises and spot a number that has been flaky for a while.

Use cases

• Catch a broken signup form early. Add “Signups” to your watch list and set a fairly high sensitivity, because a drop in signups costs you directly. If an alert fires that signups fell far below normal, do not wait for an end of week report to notice. Action: open your signup funnel right away to find the step that broke, fix it, and watch the next day’s signups return to the normal range on the small chart.
• Hear about a viral traffic spike while it is still happening. Watch “Sessions” or “Unique visitors”. If an alert fires that traffic jumped far above normal, something you posted may be taking off. Action: open the sources page to find which channel is sending the surge, then make the most of the moment while the attention lasts, for example by pinning your best offer where those new visitors land.
• Notice a slow decline before it hurts. Watch a steady number like weekly purchases or returning visitors over a longer range. A slow drift down can hide in plain sight because no single day looks alarming. Action: if alerts start clustering around the same number over several weeks, treat it as a trend, not a fluke, and dig into the funnels and sources pages to find what changed before the decline becomes a serious dent.

Tips

• Watch the numbers you would act on, not every number you have. An alert is only useful if you would do something about it. A tight watch list of three or four numbers that really matter beats a long list that buzzes all day and trains you to ignore it.
• Read the size, not just the color. A warning card that is 5 percent off normal is a nudge. A warning card that is 60 percent off normal is an emergency. The percent next to the expected value is what tells you how hard to react.
• Not every anomaly is bad, and not every bad day is an anomaly. A sale or a holiday will trip an alert for a perfectly good reason, and a quiet weekend that was always going to be quiet will not. Bring the context the page cannot know.
• Tune sensitivity to the number’s nature. A stable number like purchases can run at high sensitivity because any real move means something. A naturally jumpy number wants lower sensitivity, or it will cry wolf every other day.
• A brand new site has nothing to compare against yet. leadmaps needs some history of a number before it knows what normal looks like, so a fresh setup may show no alerts for a while. An empty list early on is expected, not a fault.