How Spikely decides a number is unusual
A fixed threshold either fires all night or misses the morning that matters. Here's how Spikely learns a usual range for every metric, by hour and day, instead.


Why fixed thresholds fail
Say you alert when signups drop below 20 an hour. At 3 AM that fires every night, because 4 signups at 3 AM is perfectly normal. On a busy Tuesday afternoon, 25 might be a real drop, and it says nothing.
The same number means different things at different times. So Spikely doesn't ask you for a threshold first. It learns one.
A usual range for every metric
For each metric, Spikely looks at how it behaves at that hour on that day of the week, and builds a usual range around it. The range is narrow at 3 AM, when not much happens, and wide on a busy afternoon, when numbers swing more.
When a number steps outside that range, up or down, that's a move. Spikely tells you which way it went, how far outside it is, and since when, like Errors up 16× since 9:00 AM.
Sensitivity: how far is far enough
Most people start on Balanced. Big moves only suits a noisy product; Every move suits a launch week, when you want to hear about everything.
- Big moves only: only moves far outside a metric's normal range, like a 50% drop.
- Balanced: moves bigger than a metric usually makes at that hour and day.
- Every move: any change past a metric's normal range, however small.
Good moves count too
Not every move is bad news. Visitors up 3× from a front-page post is worth knowing about. Spikely knows that for errors and refunds, lower is better, so a drop there reads as good news rather than an alarm.
Then: why it moved
When you open a move, Why lists what your other connected tools recorded around the same time: a deploy on Vercel, a new error in Sentry, a spike of traffic from one site. It reports what happened and never guesses at a cause it can't see.
Want a hard line as well?
Add your own rules for the numbers you care about: when MRR drops below $5,000, or signups fall by 30%. Mark a rule urgent and it breaks through quiet hours.