Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Google Ads & PPC › Bidding & Budgets

Scaling Google Ads Spend

How to grow a Google Ads account without wasting the increase: where headroom sits, budget-limited campaigns, looser targets, wider queries, new campaign types.
Guideadvanced10 min readUpdated

Where headroom comes from

Every campaign is eligible for a certain volume of auctions, and its Impression Share says how much of that volume it took. The rest was lost for one of two reasons, which the account reports separately: the budget ran out, or the ad's rank was too low to show. Beyond both sits a third pool that no column measures, the searches, placements and places the campaigns are not set up to enter at all.

Source of headroomWhere it showsLeverRisk
Demand lost to budgetSearch lost impression share due to budget; limited-by-budget statusRaise the budgetLowest: the campaign already wants to spend
Demand lost to rank under a targetLost share due to rank; spend below budget; a target stricter than recent resultsLoosen the target in stepsModerate: buys auctions the model judged less likely to pay
Demand lost to rank from qualityLost share due to rank; weak Quality Score componentsBetter ads and pagesLow cost, slow; the Quality Score guide owns it
Searches the keywords do not reachSearch terms near the edge of the list; keyword researchNew keywords, wider match typesModerate: new queries convert less predictably
Inventory the account does not useCampaign types not yet runShopping, Performance Max, video, displayHigher: new learning, new measurement questions
Places and people not targetedLocation reports; markets not servedNew areas, languages and audiencesHighest: closer to a new launch than to scaling

Read the table from the top. The first rows release demand the account already understands, at prices close to those it already pays. The lower rows reach people and searches the account has no history with, where the cost of the next conversion is a guess until measured.

Budget-limited campaigns first

The cheapest headroom sits in a campaign that is limited by budget and profitable at the margin. Google reports the status with a recommended budget, and the search lost impression share due to budget shows how much eligible demand went unbought. The budget simulator Google provides estimates conversions at other budget levels, which the budgets guide recommends as the quickest check on whether the next tranche of spend is worth buying.

Three conditions make a budget increase the right first move. The campaign's marginal cost, not its average, is inside the ceiling the business can pay. The lost share to budget is material, so there is demand to buy. And the campaign's conversion tracking measures the real outcome, so the extra conversions are real ones. Where all three hold, raising the budget buys more of something the account already knows how to buy.

Two details from the documentation shape the increase. A campaign whose budget is not the constraint will not spend more if the budget rises: under a target, the budget only matters once the campaign is limited by it. And a large budget change is among the changes Google lists as able to start a learning period, so the increase is made in steps rather than one jump, with time between them for results to settle.

Loosening targets in steps

A campaign that leaves budget unspent under a target CPA or target ROAS is held back by the target, not by money. Google documents that Smart Bidding sets a bid for each auction from a prediction of whether it will convert and what the conversion will be worth, and that the target is an average the strategy aims for. A stricter target therefore means fewer auctions where the prediction clears the bar; a looser one admits auctions the model judged slightly less likely to pay. The underspending campaign diagnosis traces the same mechanism from the symptom.

Loosening is a controlled way to buy those auctions. Raise a target CPA or lower a target ROAS by a small amount, leave the campaign long enough for the conversions from the new level to arrive, then measure the extra spend and the extra conversions. Each step costs more per conversion than the last, because the model is reaching further down its list of auctions, and that is what the next section measures. The limit is the economics the bidding hub derives from margin: the maximum cost per acquisition or the break-even return, never the figure the campaign happened to achieve last quarter.

The Smart Bidding strategies guide sets out the practical rules that follow from the learning period: change one thing at a time, in steps, and not during a sale or a holiday. Loosening a target and raising a budget on the same day leaves two causes for one result.

Widening the queries: keywords and match types

When campaigns already buy most of the demand their keywords can reach, the next source of volume is searches they do not reach. Two routes exist and they differ in control.

New keywords from the search terms report. Searches that converted through a loose match and are not yet keywords in their own right are the safest additions: the account has evidence they convert. Adjacent services, product names and problems the business solves come next, from keyword research.

Wider match types. Google documents that broad match reaches searches related to the keyword, beyond the meaning phrase and exact match require, and recommends it with Smart Bidding because the strategy can judge each related search at auction time. The match types guide adds the condition the documentation does not spell out: broad match needs enough conversion data for the strategy to be selective. Moving proven ad groups to broad match is better run as a test than as a switch; the broad match expansion test describes a campaign experiment that holds the bid strategy fixed, compares value at target and sets guardrails on the queries.

Either route brings searches that convert less predictably than the core, so negatives have to keep pace with the search terms the new keywords bring, as the negative keywords guide describes.

New campaign types and placements

Search reaches people at the moment they search. Other campaign types reach different inventory: Shopping puts products with images and prices on Google's results, Performance Max runs across Google's channels from one campaign, and video and display reach people who are not searching yet. Each is a new source of volume and a new set of measurement questions.

The comparison of Performance Max and Search campaigns sets out how the two interact in one account, which is the first question when a Performance Max campaign joins an established Search account: some of what it reports may be demand the Search campaigns were already buying. Shopping needs a product feed in good order, which the product feeds guide covers. Video and display, covered in the YouTube and display sections, create demand rather than capture it, so their returns build more slowly and should be judged on a longer horizon and with view-based conversions reported separately.

A new campaign type starts with no history and its own learning. Give it its own budget, a target set from the same economics as the rest of the account, and a reading of its marginal contribution at the account level, not only of its own reported results.

New markets and audiences

The largest headroom is people the account does not reach at all: new areas within a country, new countries and languages, and audiences that are not yet looking. They are also the furthest from scaling and the closest to a launch. A new country brings new pricing, competition, regulation and language, and the testing and scaling guide treats it as the most demanding of the scaling paths. Within a country, extending location targeting to areas the business can serve but has not targeted is the gentler version, read area by area in the location report. For a local service business taking its work to a new town, the guide to expanding into a new service area runs that extension as a trial with its own campaign, budget, cost ceiling and decision date.

Audiences that are not yet in market, reached through video or display, are demand creation. They suit an account whose search demand is largely bought, and they should be funded and judged as a separate job from the search campaigns, with patience about how long their effect takes to show.

Judging each step at the margin

The account average is the wrong measure of a scaling step, because it blends the cheap conversions the account already had with the expensive ones the step bought. The step is judged on its own figures.

Marginal CPA = Extra spend ÷ Extra conversions(currency)

Marginal ROAS = Extra conversion value ÷ Extra spend(ratio)

Compare the marginal figure with the ceiling the business can pay: the maximum CPA for a lead account or the break-even return for a sales account, as the bidding hub derives them. While the marginal figure is inside the ceiling, the step adds profit; once it crosses, the step loses money even if the average still looks healthy. Diminishing returns mean each step will be worse than the last, and the marginal returns guide explains why the average always flatters the margin. The marginal ROAS entry defines the value version. The CPA calculator and the ROAS calculator do the arithmetic.

Four rules keep the reading honest. Compare periods of the same length and season, before and after the step. Read them only after the conversion delay has passed, because the newest days are incomplete. Read at the account level, because a step in one campaign can move conversions from another, which the Smart Bidding guide warns about when judging strategy changes. And remember that platform-attributed conversions include some that would have happened anyway: for a large increase, the budget increase incrementality test measures what the extra spend caused rather than what it was credited with. Writing each step as a hypothesis, a measured result and a decision follows the experimentation loop, and the record becomes the account's own curve of cost against volume. Each step's marginal CPA, written into that record, is one point on the curve. The academy checkpoint on thin data, sales, outages and scaling spend tests where an account's headroom sits first and how each step of growth is judged at the margin.

A fabric shop grows over two quarters (illustrative)

Illustrative example — not a real result

Bobbinmoss Fabrics raises spend in four steps

Bobbinmoss Fabrics, an invented online fabric shop, spends £6,000 a month on Google Ads and records 400 sales, an average cost per sale of £15. Orders are of similar value, so the campaigns bid on target CPA. From its margin the shop can pay at most £26 for an additional sale. Over two quarters it takes four steps, each held for about six weeks and read after the conversion delay had passed.

StepLeverExtra spend a monthExtra sales a monthMarginal cost per saleAverage cost per sale after
OneBudget raised on two Search campaigns limited by budget£1,20060£20.00£15.65
TwoTarget CPA loosened from £15 to £18 in two moves£1,50065£23.08£16.57
ThreeBroad match tested on the best-selling ad groups, then adopted£90038£23.68£17.05
FourNew Performance Max campaign across the catalogue£2,00055£36.36£18.77

The first three steps each cost more per sale than the one before, as expected, and all stayed under the £26 ceiling. The fourth step is the instructive one: the account average after it was still a comfortable £18.77, but the 55 extra sales had cost £36.36 each, and part of what the new campaign reported turned out, on comparison with the Search campaigns' figures, to be sales they had previously bought. Judged on the average, the step looked like success; judged at the margin, it lost money on every extra sale. The shop removed the fourth step, kept the first three, and recorded £9,600 a month and 563 sales as its new baseline, with a note to retest Performance Max on the product lines Search does not cover.

Common mistakes

  • Raising the budget on a campaign that is not limited by it. Under a target, the budget is not the constraint and the increase changes nothing.
  • Judging a step on the account average. The average blends old cheap conversions with new expensive ones.
  • Taking several steps at once. A budget rise, a looser target and new keywords in one week leave one result with three possible causes.
  • Reading a step before its conversions arrive. The newest days always look weak.
  • Loosening the target to last quarter's figure instead of the margin limit. The ceiling comes from economics, not history.
  • Switching to broad match without a test. Run it as an experiment with the bid strategy held fixed.
  • Counting a new campaign type's reported results as all new. Some of them were already being bought.
  • Expecting demand creation to pay back like search. Video and display need a longer horizon.

Frequently asked questions

How much can I raise a Google Ads budget at once?

Google does not publish a safe percentage. Its Smart Bidding documentation lists large changes to budgets and targets among the changes that can start a learning period, and the account's own history is the best guide to what counts as large for a given campaign. Raise in steps, leave each step long enough for its conversions to arrive, and read the marginal cost of the step before taking the next, so that a disappointing increment can be identified and reversed on its own.

Should I scale by raising the budget or by loosening the target?

It depends on what is holding the campaign back. A campaign limited by budget, with a large share of impressions lost to budget, will spend more if the budget rises. A campaign with a target that leaves the budget unspent will not: the target is the constraint, and raising the budget changes nothing until the target is loosened. Check the bid strategy status and the lost impression share columns before choosing the lever.

Why did my cost per conversion rise when I scaled?

Because each increment buys auctions the campaign had previously passed over or lost, which are less likely to convert or more expensive to win. That is diminishing returns, and it is expected. The question is whether the extra conversions cost less than they are worth, which only the marginal figure answers: the extra spend divided by the extra conversions, compared with the most the business can pay for one.

Terms used on this page

Tools to try

Frameworks

Where this page is used

Part of these learning paths

Also related

Sources

  1. About average daily budgets — Google Ads Help
  2. About Smart Bidding — Google Ads Help
  3. About keyword matching options — Google Ads Help

Related