Warn before you blow the budget

A budget report at month end tells you what already happened. Useful for learning, useless for deciding, because every decision that produced the number is behind you.

The version worth having warns you while there is still something to do.

Set the threshold below the limit

The instinct is to alert at a hundred percent. By then the money is committed.

Pick a level where you would still change behaviour: often seventy or eighty percent of a monthly budget, or a run rate that would exceed it if it continued. The point is to arrive early enough to act.

Watch the rate, not just the total

Every weekday at 8am, check spend against budget by category.

Tell me about a category only when one of these is true:
  it has passed 75 percent of its monthly budget
  its run rate this month would exceed the budget if it continued
  it has grown more than 40 percent against the same point last month

For each: the category, spent so far, the budget, the projected month-end
figure, and what changed.

Say nothing about categories that are fine. If nothing qualifies, say
"nothing to flag" and stop.

The run rate condition is the one that earns its place. A category at 40 percent on the tenth of the month is on track to be at 120 by the end, and a threshold alarm will not notice until the twenty-fifth. Projecting forward is the difference between a warning and a receipt.

Send it where a decision can be made

Into the channel where the people who could actually stop the spend are, not into an inbox. If it goes to one person who is travelling, the alert is a log entry.

Give the projection a caveat

Early in the month a run rate projection is nearly meaningless — three days of spend extrapolated is noise. Ask it to say how confident the projection is based on how much of the month has elapsed, and to stay quiet on rate alone in the first week.

Without that you will get an alarming projection on the second of the month, every month, and you will learn to ignore it.

What good looks like

You find out about an overrun on the twelfth with a fortnight to react, rather than on the first of the following month.

The measure is whether any alert changed a decision. If none did in a quarter, either your budgets have too much slack or the thresholds are too high to catch anything actionable.