How to Set a Google Ads Budget That Is Not a Guess
Most Google Ads budgets are a round number someone felt comfortable with. A budget you can defend starts at the revenue you need and works backwards through margin, conversion rate and cost per click — and it usually tells you something uncomfortable before you spend a penny.
Start at the revenue target, not the spend
The right sequence is revenue, then margin, then conversions, then clicks, then spend. Working in the opposite direction — picking a monthly figure and hoping — is why so many accounts run for months without anyone able to say whether they are working.
Say you need 50,000 in additional monthly revenue and your average order value is 250. That is 200 orders. If your landing pages convert at 2.5 percent, you need 8,000 clicks. At an average cost per click of 1.80, that is a 14,400 monthly budget.
The value of doing it this way is that every assumption is now visible and arguable. If 14,400 is more than you can spend, you know exactly which lever to pull: raise average order value, improve conversion rate, or find cheaper traffic. A round number gives you nothing to work with.
Break-even ROAS is the number that governs everything
Before judging any campaign you need to know the point at which you stop losing money. Break-even ROAS is simply one divided by your gross margin.
At a 25 percent margin you need 4:1 just to break even. At 50 percent you need 2:1. At 80 percent, typical for software or services, you break even at 1.25:1. This is why comparing your ROAS to an industry benchmark is close to useless unless the margins match, and why a campaign that looks strong in one business is a disaster in another.
Work out your own figure first, then set a target ROAS above it with enough headroom to cover the costs that sit outside cost of goods. Our ROI Calculator and Ad Spend Calculator will run the arithmetic across different assumptions so you can see how sensitive the plan is.
Budget by intent, not evenly
Splitting budget equally across campaign types is a common and expensive habit. Different campaigns do different jobs and deserve different money.
Branded search is cheap and converts brilliantly, which makes it look like your best campaign. Be sceptical: most of those people were going to find you anyway. It is usually worth running defensively, but it is capturing demand rather than creating it, and its reported ROAS is flattering.
High-intent non-branded search — people searching for what you sell without knowing you — is where incremental revenue actually comes from. This deserves the largest share.
Remarketing is efficient but strictly limited by the size of your audience. Budget it to the audience, not to a percentage; pushing more money at a small list just increases frequency until people resent you.
Broad prospecting and Performance Max need enough volume to learn. Underfunding them produces a campaign that never exits the learning phase and generates data no one can act on.
The learning phase is a real constraint on how you budget
Google needs roughly 30 to 50 conversions in a rolling period before automated bidding performs predictably. If your budget generates 8 conversions a month, smart bidding will not work well and no amount of tuning fixes that.
This has a direct budgeting consequence: concentration beats spread. Three campaigns each getting 40 conversions will outperform ten campaigns each getting 12. When budget is tight, cut the number of campaigns rather than the budget per campaign.
It also means you should resist the urge to change things constantly. Every significant edit to bidding, targeting or creative can reset learning. Set a review cadence — typically every two weeks — and let campaigns run between reviews.
A useful floor: if a campaign cannot realistically produce 30 conversions a month at your target cost per acquisition, either consolidate it into another campaign or do not run it.
Build in a test budget, and protect it
Set aside roughly 10 to 20 percent of spend for testing new keywords, audiences, creative and landing pages. The purpose is not immediate return — it is finding the next thing that works before your current winners decay, which they will.
The reason to ring-fence it is that test budget is always the first thing cut when a month looks tight, which is exactly when you most need to find something better. An account with no test budget slowly optimises itself into a corner: rising costs, shrinking audiences, and no candidate to replace the campaign that is fading.
Judge test spend on learning rather than ROAS. A test that conclusively proves an audience does not work has done its job.
Seasonality and pacing
Monthly budgets get spent unevenly, and daily budget caps interact badly with that. Google can spend up to twice your daily budget on a high-traffic day, balancing across the month, which is usually fine — but if your demand is genuinely spiky, plan for it rather than letting the algorithm smooth it.
Look at last year for genuine seasonality, then budget above baseline in your strong periods rather than holding a flat monthly figure. Competition rises too, so your cost per click will increase in peak season; a flat budget quietly means less volume exactly when demand is highest.
Finally, watch for budget-limited campaigns. If a campaign is hitting its cap daily and still converting at target, it is leaving money on the table — that is the clearest signal in the whole account that more budget is justified.
Key takeaways
- Work backwards: revenue target, margin, conversion rate, cost per click, then budget.
- Break-even ROAS is one divided by gross margin. Benchmarks from other businesses are meaningless without matching margins.
- Branded search flatters its own numbers — it captures demand rather than creating it.
- Concentration beats spread: automated bidding needs 30 to 50 conversions per campaign to work.
- Ring-fence 10 to 20 percent for testing, and judge it on learning rather than return.
Frequently asked questions
What is a good starting budget for Google Ads?
There is no universal figure, but there is a useful floor: enough to generate around 30 conversions a month in your main campaign, because that is roughly what automated bidding needs to perform. Work it out from your own numbers rather than a benchmark. Multiply your target conversions by your expected cost per acquisition, which is your cost per click divided by your conversion rate. If that figure is far beyond what you can spend, the honest answer is usually to narrow your targeting to fewer, higher-intent keywords rather than spreading a small budget thinly.
What is break-even ROAS and how do I calculate it?
Break-even ROAS is the return on ad spend at which you neither make nor lose money, and it equals one divided by your gross margin. At 25 percent gross margin you need 4:1, at 50 percent you need 2:1, at 80 percent you need 1.25:1. Anything above that figure is profit before overheads. This is the single most useful number in paid media planning, because it turns "is 3:1 good?" from an opinion into arithmetic. Calculate it once and put it at the top of every campaign report.
Should I bid on my own brand name?
Usually yes, defensively, but with clear eyes about what it is. Bidding on your brand is cheap and converts extremely well, which makes it look like your best-performing campaign, when in reality most of those people already intended to buy from you. The genuine reasons to run it are keeping competitors from appearing above you, controlling the message and landing page, and covering terms where your organic listing is weak. Test it with a geographic holdout if you can, because many businesses find most of that revenue arrives organically anyway.
How much should I allocate to testing?
Between 10 and 20 percent of total spend is a sensible range for most accounts, and the important part is protecting it. Test budget is invariably the first thing cut in a tight month, which is precisely when finding something better matters most. Judge it differently from your core budget: a test that definitively proves an audience or message does not work has succeeded, even though its ROAS is zero. Without a standing test budget an account slowly optimises into a corner, with rising costs and nothing ready to replace a fading winner.
Why is my campaign not spending its full budget?
The usual causes are that your bids are too low to win auctions, your keywords have genuinely low search volume, your targeting is too narrow, or your ad quality is limiting impression share. Check the impression share lost to rank versus lost to budget in your reporting: those point to different fixes. Lost to rank means bidding or quality, lost to budget means you are actually capped. An underspending campaign is not automatically a problem, but it usually indicates the plan assumed more available demand than the market has.
How often should I change my budget?
Review every two weeks, change deliberately, and avoid constant tinkering. Significant edits to budget, bidding strategy or targeting can push campaigns back into the learning phase, during which performance is unstable and the data misleads you. If you need to increase spend on a working campaign, do it in increments of roughly 20 to 30 percent rather than doubling it, so the algorithm can adjust without fully relearning. The exception is genuine seasonality, where planned step changes are correct.
Is Performance Max worth the budget?
It can be, but it needs enough volume to learn and it is much harder to diagnose than search campaigns because reporting is limited. Two practical cautions. First, it will happily absorb branded traffic and report it as its own success, so exclude brand terms if you want a clean read. Second, it needs strong creative assets and a properly configured product feed, so it rewards preparation. If your budget only supports one well-funded campaign, a tightly targeted search campaign is usually the safer first investment.
How do I forecast cost per click before I start?
Use Google Keyword Planner for a range, then treat it as optimistic. Planner figures come from historical auction data and do not know your quality score, which materially affects what you actually pay. A reasonable planning approach is to take the top-of-page bid estimate, assume you will pay somewhere near it initially, and build in room for costs to rise 20 to 30 percent during the first month while quality scores establish. Revisit the whole model after four weeks of real data rather than treating the forecast as fixed.