What 18 months of managing other people's budgets taught us.
A campaign that is working reaches its daily budget by early afternoon and stops. This is what we built to stop that happening, and what running it on live accounts taught us.
This is the under-the-hood companion to The most expensive thing in your ad account is the budget nobody's watching.
Here is a thing that happens in ad accounts, every day, that almost nobody sees.
A campaign that is working — converting, profitable, exactly the one you'd want to spend more on — reaches its daily budget by early afternoon and stops. For the rest of the day it serves nothing. Demand keeps coming; the ads don't. Meanwhile, two campaigns over, budget sits unspent because that one was quieter than expected. The numbers that decided all of this were set weeks ago, by someone who had no way of knowing what a Tuesday in late September would look like. Google marks the campaign Limited by budget, so the fact is not hidden. The hour it stopped is, and so is the chance to act on it while the day is still running.
We spent a long time watching that pattern across accounts and came to a conclusion: it's the most fixable expensive problem in performance media. Not the bidding — the platforms have made bidding genuinely good. The budget. The one lever the automation hands back to you and then mostly leaves you to manage with a number you best-guessed a month ago.
So we built something to manage it properly, and for the past 18 months we've run it on live accounts — billing seven figures at the top end and a few thousand dollars a month at the other. It's called Orchadstrate, a budget orchestration platform for Google Ads, and this is what running it taught us.
The fix has to be intraday, or it isn't a fix
Raising a daily budget once, on an ordinary week, doesn't solve anything — it just raises spend on the days that were already fine, and the campaign still runs dry on the days that matter. What works is watching each campaign against its own day. A campaign that front-loads its spend and one that earns in the evening are different animals, and a single daily number treats them identically.
How it actually knows. Each campaign gets its own baseline curve, learned from its recent history — not a straight line, but the real shape of how it spends and converts across the hours of the day and the days of the week. Live, the engine compares what's happening now against that expected curve: is the campaign pacing ahead of its own normal? Is demand running hotter than the baseline predicts — more clicks, more impression share, at this hour? Is the conversion rate climbing, or the cost per conversion drifting out? A move worth acting on is a deviation large enough to stand clear of ordinary hour-to-hour noise — which is how it separates a genuine surge from a merely busy afternoon, and acts on the signal rather than the wobble. The failure a person can't catch at 2pm on a Tuesday, the curve sees coming at eleven.
Drag the hour.
Made-up numbers, real rules: extended funding on, redistribution on, a $40 cost-per-conversion target, no raise past twice the baseline. Demand on Campaign A surges at six in the evening, when nobody is at a desk, so it gets budget twice — allocated from Campaign B’s surplus in the morning, then added at six. Drag the hour and watch each campaign’s status change.
A campaign running ahead of its own curve has its budget extended, only if it clears the target and never past its cap, and surplus from a campaign below its curve is used first. On your account every line of this log is a real change, recorded with its reason.
Use the budget you already have, before asking for more
The instinct, when a good campaign is about to run dry, is to give it more. Most of the time you don't need to. Somewhere else in the same account, a quieter campaign is sitting on budget it won't spend today.
How reallocation decides. The engine reads every campaign against its own pace. Where one is tracking below its curve and won't use its allocation, that surplus is offered first to the campaigns that are converting and are on course to run out, within pools you define, before any increase to total spend is ever proposed. Only when a converting campaign still needs headroom after the surplus is exhausted does the engine consider ramping it. Over 18 months, that reallocation did nearly as much work as every budget increase combined — because most accounts are leaking in both directions at once, and neither leak shows up in a report until the day is over.
It only acts to rules you set, and it shows its working
The guardrails. Every action happens inside limits you set: a ceiling per campaign (a hard multiple of its baseline), a monthly cap the engine will not cross whatever happens, and performance floors — a minimum return or a maximum cost per conversion, below which a campaign is left alone rather than funded. A campaign that isn't earning doesn't get more; it gets left, and the reason is recorded. Those floors are recalculated on every pass, so a campaign that looked acceptable at nine on a handful of conversions doesn't keep a free pass for the rest of the day. A campaign too new to have a track record is given room rather than judged on numbers it hasn't had time to earn.
And every change is written down: this campaign, this hour, this much, because of this. Not a black box you're asked to trust — a log you can read, hand to a client, or check when something looks off. "It made $40 available to the campaign converting at 6pm, from one sitting idle, under your cap" is a sentence you can defend in a meeting. "It spent more on Tuesday" is not.
None of this replaces what Google does
Worth saying plainly, because it's the first question everyone asks: leave Smart Bidding on, leave Performance Max on, leave Google's own demand-led pacing on. They decide how to bid, and they pace spend across the days of the month. Orchadstrate works on the budget, underneath them, at a finer grain.
The distinction matters more than it sounds. Google's demand-led pacing shifts a campaign's own budget toward the days it predicts will be busy — useful, but it works across days, within one campaign, on Google's read of demand, and only where you have slack to move. It does nothing about the campaign that exhausts at 2pm within a peak day, and it doesn't move budget between campaigns. That is what shared budgets are for, and they exclude Performance Max. Orchadstrate works to the hour, across every campaign type, reallocates between campaigns, and does it to your rules with every move on the record. Same idea at the coarse end; a different resolution. You run both.
What it looks like when it matters
Two cases stay with us. A global charity, at end of financial year — demand stacked into the final days, campaigns going dark at exactly the wrong moment because a number set weeks earlier ran out. Making their budget genuinely always-on lifted tax-time revenue by almost 50% against the same window the year before, purely by meeting demand instead of missing it. And a string of B2B advertisers whose best lead campaigns kept tapping out whenever no one was watching — evenings, weekends, the middle of a surge.
Charity or commercial, the lesson is identical: the money arrives when it arrives. Your budget has to be there when it does.
The test is your own account, not our description of it. A free 7-day trial watches one account hourly and records everything it would have done — no card, changes nothing. If your money is already being spent at the right hours, good; you'll have paid nothing to learn that. If it isn't, you'll see exactly where.
Google's own description of daily budgets, shared budgets and the monthly spending limit is in Google Ads Help. Orchadstrate is made by The Third Way and is not affiliated with or endorsed by Google.