How should I set my alert threshold?

A common approach is a multiple of your baseline — for example 3x your normal daily spend — so routine variation does not page you but a genuine spike does. The simulator lets you test a candidate threshold against several spike sizes before committing.

What counts as a runaway scenario?

A runaway is an unbounded loop or retry storm — a bug that keeps calling the API without backing off. It is modelled here as a large multiplier of baseline (e.g. 50x) so you can confirm your threshold and a hard cap would catch it quickly.

Why not just set the threshold very low?

A threshold too close to baseline fires on normal daily variation, training people to ignore alerts. The goal is a threshold high enough to avoid false alarms but low enough to catch a real spike within a day.

Does a threshold alert stop the spending?

No. An alert only notifies you. To actually stop runaway spend you also need a hard spending cap or kill switch at the provider or gateway level. Use the alert to investigate and the cap as the backstop.

Is my data uploaded anywhere?

No. All simulation runs in your browser. Nothing you enter is stored or sent to a server.

Cost Spike Alert Simulator

Simulate your AI cost alerts before they matter

A spend alert is only useful if it fires on real spikes and stays quiet on normal days. This simulator takes your baseline daily spend and a candidate alert threshold, then projects what a 3x, 10x, and runaway spike would cost per day — and flags which ones would actually trip the alert.

How spike detection works

Most teams alert on daily spend crossing a threshold. The threshold is usually expressed as a multiple of baseline:

fires_alert = projected_daily_spend >= threshold
projected_daily_spend = baseline x spike_multiplier

A 3x spike often comes from a legitimate traffic surge; a 10x spike usually signals a bug or abuse; a runaway (tens of times baseline) is almost always a retry storm or infinite loop. Your threshold should sit above normal variation but below the smallest spike you care about.

Tips for good guard rails

Layer alerts and caps. An alert tells you something is wrong; a hard spending cap or kill switch is what actually stops a runaway. Use both.
Alert on rate, not just total. A daily total alert can be hours late. Pair it with an hourly burn-rate alert so a runaway is caught fast.
Tune to avoid fatigue. If the 3x scenario fires on a normal busy day, your threshold is too low — raise it until only genuine incidents trip it.