Guide · Spend control
Budget pacing: the mechanics nobody reads until it's expensive
Written 19 November 2025 by Wout Peeters · thresholds re-checked 3 August 2026
Google Ads budgets do not behave the way the word "budget" implies, and the gap between assumption and mechanics funds a lot of unpleasant Mondays. This guide covers the overdelivery rules as they actually work, where genuine runaways come from, and the pacing alerts that turn a €4,000 surprise into a €90 anecdote.
How does Google actually treat your budget?
As an average, not a ceiling. On any given day a campaign may spend up to twice its average daily budget, and Google balances the excess across the billing month so total charges stay within roughly 30.4 times the daily figure. Within a month this usually nets out; within a week it absolutely does not, which matters for cash flow, for promotions, and for anyone reading a three-day report. The practical translation: your "€100/day" campaign is licensed to have €200 days whenever the auction looks tasty, and no alarm accompanies that.
Where do real runaways come from?
- Bulk edits: a spreadsheet upload or script that strips caps or fat-fingers a decimal. The classic generates a 10× budget with no visual drama in the interface.
- Shared-budget inheritance: a new campaign joins a shared pool and quietly outbids its siblings for the whole allocation.
- Currency and import mixups: an account imported across currencies where 5,000 meant something very different at origin.
- Forgotten experiments: a test slice left running after its verdict, spending dutifully toward nothing.
- Auction shocks:a competitor's exit or a news spike drops CPCs, and volume floods in at 2× license. The one genuinely platform-side cause, and even it only hurts unwatched accounts.
Note what is missing: malice. Invalid traffic is real money at industry scale ($63 billion globally in 2025, with 8.51% of paid traffic flagged invalid), but the runaways that produce four-figure single-day damage are almost always a legitimate account change nobody was watching.
The pacing math worth automating
Two comparisons, computed hourly. Daily: today's spend so far against a same-weekday 28-day baseline, alerting at ±25%, because Monday and Saturday are different animals in every account we watch. Monthly: month-to-date spend against the planned curve, alerting at 110%, with the curve shaped to your calendar rather than divided flat: a retailer's promo week is supposed to spend heavy, and a flat plan cries wolf exactly when everyone is too busy to care. Both checks are buildable with an Ads script and an afternoon, per the monitoring checklist, and both run natively in our monitoring layer every 30 minutes.
What did this look like in a real account?
The cleaning franchise in this outcome ran shared budgets across 14 territories with a flat weekly plan. Nothing was broken by the usual definition, and the account still bled: quiet Saturdays absorbed spend the hot Monday mornings needed, an artifact only visible once pacing was measured against weekday-shaped baselines. Re-curving the plan and adding the ±25% alerts contributed a meaningful slice of the €47 → €26 lead-cost drop, and cost nothing but attention. Pacing work rarely stars in case studies because it looks like bookkeeping; the ledger disagrees.
The runaway response drill
- Cap or pause the bleeding entity first; diagnosis can wait eleven minutes.
- Preserve evidence: spend by hour, current settings, the change history view.
- Find the trigger in change history; it is almost always a change.
- Reconcile damage in euros against the plan, for the log rather than for blame.
- Add the alert that would have caught it. The second incident should cost coffee money.
Accounts that run this drill once tend to become monitoring believers; the arithmetic in the €9,400 write-up explains why faster than any pitch.
Asked before signing, usually
Yes, by design. Google may spend up to twice your average daily budget on any single day to chase traffic, balancing it across the month so you're not charged beyond roughly 30.4 times the daily figure. The monthly math usually holds; the cash-flow surprise on a heavy Tuesday is yours to absorb.
Human edits, mostly: a bulk change that removes a cap, a currency mixup on an imported campaign, a shared budget quietly inheriting a new campaign, or an experiment left running. Platform overdelivery gets the blame; the incident logs say account changes cause the expensive cases.
They make it smoother and less visible at once. A shared pool stops single campaigns from starving, and it also lets one broken campaign quietly drink the pool dry while the others fade. If you use shared budgets, per-campaign pacing alerts stop being optional.
Alert at ±25% of the daily plan and at 110% of the month-to-date plan, measured against a same-weekday baseline so weekends don't cry wolf. Retail with strong seasonality needs event-aware plans; a flat monthly number divided by 30 misfires every promotion week.
Cap first, diagnose second: pause or cap the bleeding entity, screenshot spend by hour, then find the triggering change in the change history. Reconcile the damage in euros, and only then decide whether to request an invalid-activity review. Diagnosis before capping is how a bad hour becomes a bad day.
Want the alerts without the afternoon?
Pacing is one of the five checks in the €390 monitoring tier, running every 30 minutes with a human on call.
Get a free strategy call