You scale Google Ads without a higher CPA only by a step from a ceiling, not by a percent copied from someone else. The rules do not name such a percent. The ceiling is the lead price the deal margin can still hold. The Conversions column has to show the action you get paid for, not a click and not a funnel step. Change the budget or the target one at a time. Judge the leads from this step, not the average CPA. Above the ceiling, roll the step back even when the average still looks calm.
I have worked in Google Ads for 19 years, since 2006. The account breaks when someone accepts the recommended budget and leaves the target looser than the price the business can pay. Next comes one lever and a rollback. This sits inside the system in my article on the Google Search formula.
Check the deal economics and tracking first
The ceiling does not come from the account. Take the deal and subtract the goods, the work, delivery, and the fee. Leave the share without which the deal is not worth keeping. What remains is the CPA ceiling. The point where ads only pay back variable costs is not profit yet. I show the arithmetic in my article on ROAS in Google Ads.
A primary action goes into bidding and into the Conversions column when the campaign optimizes for that goal. A secondary action stays in All conversions. That is how the page on conversion goals puts it. Do not set every funnel step as an account default goal. In Google’s example, a store that wants sales keeps Purchase. A secondary action does enter bidding once it sits in a custom goal assigned to the campaign. Maximize conversions will not turn on without conversion tracking. On a campaign that was spending much less than its budget, turning it on can jump the spend.
How to judge demand headroom and traffic quality
Impression share is not every query in the country. It is an estimate of auctions where the ad was already strong. A small swing does not by itself call for a change. That is on the impression share page. Search lost IS (budget) is time without an impression for lack of budget, at campaign level only. Search lost IS (rank) is time without an impression because of weak Ad Rank. The rank column is hidden on ad groups if the budget ran out in the date range. You add the columns from this help page. They update in 1-2 days.
In the Search campaigns from an archive I keep for this kind of review, impression losses from rank in 2025 exceeded losses from budget by about 24 percentage points, at the median of the linked groups. The calculation is weighted by spend. This is not a market reading and not a forecast of what a bigger budget returns. Rank includes the bid and other factors. Adding budget often treats the smaller of the two losses.
Do not read the Lost IS (budget) column for Maximize conversions or Maximize conversion value with no target. Google says the column is incompatible with those strategies. The headroom question is whether the ceiling still holds when spend reaches the budget. The search terms report shows queries that actually triggered the ad. It hides rare queries. Clean the bad ones before any step.
What to increase: budget, bids, or audiences
Read the strategy in the campaign settings. From June 2026 the label Maximize conversions with a Target CPA is renaming to Target CPA. The behavior stays the same. Target CPA is an average price you want for a conversion, not a cap on each click. Some leads will cost more than the target. Some will cost less.
Raise the budget when the loss from budget is larger than the loss from rank, and the target is already at the ceiling. On the Campaigns page, in the Budget column: pencil, amount, Save. Divide a monthly amount by 30.4. A recommended budget does not check your margin. The path is on the budget change page. For most campaigns a day can reach two daily budgets, and the month is counted as the daily budget times 30.4. Pay for conversions has no daily cap. The limits are on the page about spending limits. A shared budget is one budget across several campaigns, so a step raises the ceiling for every campaign on it.
Touch the target when the campaign is Limited by budget and actual results beat the target, or when there is a gap in Ad Rank. From 17 August 2026 a target strategy that is limited by budget stays near the target you typed, including after a budget increase. On the update page a $10 target with a $5 actual result pulls delivery toward $10 until you lower the target. Google will not fix it for you. Set a higher target only inside the ceiling. A target that is too low drops clicks.
An audience bid adjustment will not add volume. On Target CPA, adjustments other than device are ignored on Search, and a device adjustment changes the CPA target, not the bid. On Maximize conversions, adjustments do nothing except minus 100 percent on a device. Learning lasts up to 7 days after a strategy is created or reactivated, campaigns, ad groups or keywords are added to or removed from the strategy, or conversion goals or actions change significantly. Changing the budget, target, or audience does not turn that status on. Ads keep serving.
Which numbers to check at each step
One day changes one lever. Otherwise you cannot tell what moved the price. Estimate the size of the step using the budget simulator in your account or the Search lost IS (budget) metric at campaign level. The simulator gives an estimate, not a guarantee; the loss metric does not suit Maximize conversions or Maximize conversion value without a target. Choose an amount the business is prepared to test within its CPA ceiling.
| What you see | What you change | What you do not do |
|---|---|---|
| Limited by budget, and the target is looser than actual CPA | Set the target to the margin ceiling first, then one budget step | Do not add budget on the old loose target |
| Budget loss is larger, and the target is already at the ceiling | Only the average daily budget | Do not change keywords and the target the same day |
| Rank loss is larger | The target, if the new target is still inside the ceiling | Do not treat budget as the fix for rank |
| Maximize conversions or value with no target | The budget, if you accept a different CPA | Do not read Lost IS (budget) as demand headroom |
| The search terms report is full of junk queries | Negatives before any step | Do not buy more of the same queries |
| Both losses are small | New queries as a separate test | Do not expect volume from the same set |
A worked example, not a report from an account. The owner of a plumbing company set an $80 ceiling for a qualified lead from the margin. The figures below are only so the arithmetic is visible.
| Figure | Start, 14 days | After one step |
|---|---|---|
| Spend | $2,400 | $3,600 |
| Leads | 40 | 48 |
| Average CPA | $60 | $75 |
| Leads from this step | the step had not happened | 8 leads at $150 |
| Ceiling from the margin | $80 | $80 |
| Decision | one step is allowed | stop: $150 is above the ceiling |
Marginal CPA = (spend after − spend before) / (leads after − leads before) = ($3,600 − $2,400) / (48 − 40) = $150 per lead. The average of $75 is still under the ceiling. The leads from this step cost $150. The average misleads. In my Search archive for 2024-2025, months in 2025 with spend growth of 10 percent or more had an average click price about 50 percent higher than in 2024. Months without that growth showed about 3 percent. After the campaign mix was lined up, the gap was about 38 percent and 19 percent. Click price rose even without spend growth. That is a link in the archive, not an effect of the budget and not a market reading.
When to stop and how to roll the change back
Stop if the new leads cost more than the ceiling, the queries got dirtier, or spend reached the budget with no new leads. Put the budget back in the same column. Serving uses the new amount at once. If you raised the amount earlier that day, the daily cap stays at 2 times the highest amount of the day. The rest of the month is what you already spent plus the new budget times the days left, including the day of the edit. That is on the page about when a budget change takes effect.
If you touched the target, put that back too. Learning does not turn on because of it. A strategy settles to a new target over 1-2 conversion cycles. Judge the step over 2 cycles or 2 weeks, whichever is longer, and include the conversion delay. The update page asks for the same 1-2 cycles after a budget step. I lay out how many days a lead takes in my article on how long Google Ads takes to work. Do not touch keywords the same day.
- Write down the ceiling from the margin. Calculate the break-even point separately.
- The Conversions column has to show the action you get paid for.
- If actual results beat the target on a Limited by budget campaign, fix the target first.
- Record spend, conversions, CPA or value, and both impression losses.
- Change one lever.
- Judge over 2 cycles or 2 weeks, whichever is longer.
- If the step’s leads cost more than the ceiling, put the budget back.
Scenarios for an online store, services, B2B, and SaaS
Run an online store with uneven order values on target return, if you pass the real value and not one CPA for every cart. I show that calculation in my article on target ROAS. A step on an add to cart goal grows carts. From 17 August 2026 a loose target on a campaign that is limited by budget pulls toward the number you typed.
For a service business, a plumber or an HVAC shop, take the ceiling from the finished job. An all calls action buys empty calls at the same average call price once the budget is larger. In B2B a form as the primary action holds the form price while the deal gets more expensive. Bid on a qualified lead, and do not make the window shorter than the deal cycle. In SaaS a trial start is not a payment. Keep the payment in the bidding.
When this does not apply
This order fails when there is no primary conversion tied to money. The budget multiplies the wrong figure, and the ceiling is built on the wrong event. It also fails when you need an answer this week and the sale takes longer than the window you use to judge. A verdict the morning after the edit is noise.
Do not carry this over to a campaign total budget, a fixed amount for 3-90 days. There is no daily cap of two budgets. The system spreads what is left through the end date, and you cannot switch an existing campaign onto this type. The line about a target that holds when you add budget is not about App campaigns, video reach, or video views. The 17 August 2026 update does not cover them.
How to keep control as you grow
The step-by-step version is in my book on Google Ads for business owners. There you learn to hold bids, campaign structure, and profitability while spend grows: from a random budget increase to one step with a ceiling and a rollback.
My e-book “19 Google Ads Secrets”: how to get the most out of Google Ads without draining your budget, drawn from real campaigns and tests.
See the book →Questions and answers
Is there a safe percent for raising the budget?
No. The rules do not name one. The step comes from the ceiling, and the stop comes from the price of the new leads.
Does a bigger budget turn on Learning?
No. A new strategy, a change of keywords or ad groups, and a significant change to conversion goals or actions can show it for up to 7 days. The budget does not turn the status on. You still do not judge it the next morning.
Actual CPA is better than the target. Can I just add budget?
No, if the strategy has a target and the campaign is limited by budget. From 17 August 2026 delivery pulls toward the target you typed. Set the target you are willing to hold first.
When do I judge whether the step worked?
Not the next morning. Wait 1-2 conversion cycles, and judge over 2 cycles or 2 weeks, whichever is longer, with late conversions included. Count the price of the leads from the step.