How often should a budget rule run?

The hard part of writing a budget rule is not the threshold. The threshold is usually obvious: you have a target acquisition cost, and you want to intervene when a campaign crosses it. The hard part is deciding how often the rule should look — and in most accounts that decision is never made. Whatever the default was, stays.

The interval looks like an innocent setting, but it is what gives a rule its character. The same threshold reads as anxious on an hourly rule and as negligent on a daily one.

What the interval is really asking

How often a rule runs is an answer to the question: how long does this campaign need before its data means anything? In a campaign with three conversions a day, checking hourly means seeing zero conversions most hours — and the rule will read that zero as poor performance. In a campaign with two hundred conversions a day, carrying a decline that started in the morning into the next day can cost a full day of budget.

The signal to read is this: how many events does the rule need to evaluate its condition, and how long does it take to accumulate them? The interval cannot be shorter than that.

• If daily conversions are in single digits, the interval should be at least a day; checking more often means mistaking noise for signal.
• With tens of conversions a day, a twelve-hour interval is reasonable — a campaign that breaks in the morning is caught before the evening.
• Hourly becomes meaningful only in campaigns with hundreds of conversions, and even then the threshold has to tolerate intraday swings.
• Rules that read spend instead of conversions (how much of today’s budget is gone) are not waiting for events, so they can run more often.

The learning phase and an hourly rule do not sit together

A budget change can push a campaign back into learning. A rule that runs hourly and touches the budget every time it runs may never let the campaign leave that phase: the system recalibrates on every pass and no calibration ever finishes.

A practical order to start with

1. Set the rule up in suggestion mode first and spend a week only reading what it proposes.
2. Count how often you thought “yes, I would have done that”. If it is less than half, the problem is the threshold, not the interval.
3. Pull the interval down to the shortest window that still accumulates enough data to evaluate the condition.
4. Before switching to automatic, make sure the rule is one you can undo.

Not running the rule at all

Sometimes there is no correct interval. A campaign launched yesterday, a holiday week, a product that just went out of stock — in all three a threshold built on past data gives the wrong answer. Pausing the rule during those periods is more honest than loosening the threshold: a loosened threshold is waiting to be tightened again, and it usually isn’t.