TimecardPastely

How many alerts would you actually get?

Paste a metric and test fixed limits and adaptive ranges against your own history, before you switch anything on.

Runs in your browser. Your numbers aren't uploaded or stored.

Alert below 2,029 or above 2,746. That's the fixed-limit starting point.

How often each setting would have fired on this data
SettingWould have fired
Fixed limits12 points
1 alert
1 alert
1 alert

Showing Balanced. The shaded band is the adaptive range for this setting.

How the fixed limits are chosen

Limits = median ± max(3 × 1.4826 × MAD, 15% of median)

The suggested limits start from your median, then add a margin based on how much your numbers usually vary. It uses median absolute deviation, so one huge spike in your history doesn't push the limits wider.

Fixed limits or an adaptive range?

A fixed limit is simple, but it goes stale. As a metric grows, it starts firing constantly or stops meaning anything, and it can't tell a quiet Sunday from a broken Monday.

An adaptive range compares a number with what's normal for it right now. That's why the table counts both, so you can see the difference on your own data.

How to read the results

Fewer isn't always better. Check each alert against what really happened: would you have wanted to know? If the answer is mostly no, use a less sensitive setting. If it missed something real, use a more sensitive one.

Skip the tuning. Let Spikely learn it.

Spikely builds a range for every metric, adjusts as you grow, and only speaks up when a number breaks pattern.

3 days free on yearly · iPhone, iPad, Mac and Apple Watch

Questions, answered plainly.

Several weeks of daily numbers at least, so the test covers weekly patterns. The first stretch is used to learn what's normal, so it can't produce alerts.

For adaptive ranges, a run of consecutive points outside the range counts as one alert. For fixed limits, the table counts each point outside them.

Start with whichever would have fired on the moments you truly cared about, and little else. There's no universally correct sensitivity, because it depends on how noisy the metric is and how bad a miss would be.

No. Everything runs in your browser, including reading a CSV file.

More free tools