How to cut wasted ad spend: six checks, in the order to run them

Cristian Bragau, Founder & Lead Auditor · Reviewed Oct 2026 · verified against the base of 74 audits

Where the waste hides

Across our 74 published audits, wasted paid spend was 22% of audited marketing spend, and non incremental paid spend turned up in 68% of the audits, more often than any other leak. Most of that waste does not look like waste. It converts, it reports a healthy return, and nobody switches it off. The wasted ad spend page defines it, and the benchmark breaks the figures down by industry and spend.

This guide is the order of work: six checks to run on your own Google Ads and Meta accounts, where to find each one, and how to decide what to cut. The order matters. The first checks find spend you can cut this week without a test, because it loses money on any reading of the numbers. The last ones find spend that may be waste, and that needs a test before you touch it.

Set the yardstick before you cut anything

Platform ROAS is the wrong ruler for this job, for two reasons. The platforms count the same sale more than once: attribution inflation showed up in 54% of the audits. And revenue is not margin: a sale whose product cost, shipping, payment fees and returns eat most of the price can lose money at a ROAS that looks fine.

So do two things first. Reconcile one month of platform revenue with your actual orders (the steps are in how to reconcile GA4 and ad platform revenue), so you know how much of the reported revenue is real. Then work out what an order keeps after its variable costs, its contribution margin, and turn it into a break even ROAS. If your products carry very different margins, do it per product group. Every later check compares a search term, a placement, a product group or a campaign with that number.

Check 1: Cut search terms that never convert

Your keywords decide which auctions you enter. The search terms report shows the actual searches that triggered your ads, and Google notes that it leaves out terms with too little query activity, to protect privacy. Performance Max has its own search terms report, with data from March 2023 onward.

Export at least one full buying cycle of search terms, and no less than 90 days, with cost, conversions and conversion value. Sort by cost. A workable rule: a term that has spent twice what an order keeps after its variable costs, without a single sale, has already lost money and becomes a negative keyword. For example, say an order keeps $40: a term that has spent $80 with no sale goes. Terms that convert, but below your break even ROAS, get a lower bid or a tighter match type, not a negative.

Two details save money here. Negative keywords do not match close variants, so add the plurals and synonyms yourself. And compare the cost of the terms the report lists with the campaign’s total cost: the difference went to searches you cannot see one by one, and negatives and tighter match types are the controls for it. The Google Ads audit checklist scores the related settings.

Check 2: Switch off what runs by default and earns nothing

Some spend is waste only because nobody chose it. When you create a Google Search campaign, search partners are included by default; you can opt out in the campaign settings, and the Network (with search partners) segment compares their results with Google Search. On Meta, an ad set with no placements chosen runs across Meta’s default platforms, which can include Facebook, Instagram, Threads, Messenger and the Audience Network. Break the results down by platform and position before you accept that mix.

Then look at where and when the ads ran. In Google Ads, account level placement exclusions apply to every campaign in the account, Performance Max included, and since March 2024 they also cover search partner sites. Ad scheduling sets the days and hours your ads show and lets you bid up or down by time. In case 12, a B2B SaaS company’s $1.2M ad budget hid a 40% wasted impression rate, because placements and dayparting were never optimized after launch: $74k a year of margin identified, findings in 7 days.

The decision rule: a network, placement or time slot that has had a full buying cycle and returns far below your break even on its own numbers comes out. One that sits close to break even gets a lower bid.

Check 3: Take low margin products out of the bidding

Shopping and Performance Max campaigns can spend across the whole catalog, and the campaign’s ROAS averages good products with bad ones. A campaign can report 4.0x while half its spend sells products that need 5.0x to break even. To see it, give every product a margin group in your feed. Merchant Center has five custom labels, custom_label_0 to custom_label_4, for this kind of grouping, and Google’s own examples include low margin and high margin. Google recommends using custom labels to group items and target them in listing groups; from there you can exclude a group, and in a standard Shopping campaign you can also bid it lower.

Merchant Center also takes a cost of goods sold attribute, and with conversions that carry cart data, Google Ads can then report gross profit. That is a useful start, but gross profit comes before shipping, payment fees and returns, so your own contribution margin per group is the number to compare with spend.

A worked example, not a client figure: say a shopping campaign spends $10,000 a month and reports $40,000 of sales, a 4.0x return. Split by margin label, the high margin group spent $5,000 and sold $28,000 at a 40% contribution margin, so it kept $11,200, or $6,200 after its ads. The low margin group spent $5,000 and sold $12,000 at a 20% contribution margin: it kept $2,400 and lost $2,600 after its ads, because at 20% the break even is 5.0x and it ran at 2.4x. Excluding that group lifts profit by $2,600 a month even if every reported sale was real, so it needs no test.

The same pattern shows up at campaign level. In case 01, two prospecting campaigns at a DTC brand were buying revenue below break even after shipping and returns, a margin leak of $134k a year, with findings in 6 days.

Check 4: Stop your own accounts bidding against each other

Inside one account, the platforms keep you from bidding against yourself. Google says that when several of your keywords match the same search, they do not compete with each other in the auction, and only one of them can trigger an ad. Meta does the same for ad sets with overlapping audiences: it enters the one with the best performance history and holds the others back. Overlap inside an account mostly costs delivery and clarity, not price.

The expensive version runs across accounts: several locations, franchisees or regional teams, or an agency and an in house team, each buying the same keywords in an account of its own. In case 26, every location of a services business bought the same keywords against each other, and the internal competition pushed up the cost per click across the group: $39k a year of margin identified, findings in 5 days.

The check is a list. Name every account that bids on your brand and your core terms, compare their keyword lists, give each term one owner, and share one negative list across the rest.

Check 5: Audience overlap and frequency on paid social

On Meta, two questions decide whether social spend reaches new buyers or pays again for the same people. First, how much do your audiences overlap? Meta’s audience overlap tool compares up to five audiences (in Audiences, tick them, then Actions and Show audience overlap), and it needs at least 10,000 accounts in an audience to say anything useful. The overlap that costs most is between prospecting and the people who already bought or already sit in retargeting: exclude recent buyers from prospecting unless you mean to sell to them again.

Second, how often does the same person see the ad? Meta reports frequency as the average number of times each person saw your ad, and marks it as an estimate. Meta allows a hard frequency limit only on ad sets whose performance goal is reach or ThruPlay, so a sales campaign has to be controlled through its audience and its budget. Read frequency by week next to cost per order. When both rise together, the extra spend is buying more impressions on the same people, not more buyers.

The overlap also crosses platforms. A visitor retargeted on Meta and on Google can be counted as a sale by both. The Meta ads audit checklist scores those checks.

Check 6: Test brand terms and retargeting before you cut them

Brand search and retargeting usually report the best return in the account, often because the buyer was already on the way. That is where non incremental spend concentrates, and it was the most common finding in the audits. In case 50, a marketplace’s branded and category search paid for clicks its organic listings would have captured for free: $66k a year of margin identified, findings in 6 days.

Do not cut these on a hunch. Some brand spend protects your name when competitors bid on it, and some retargeting does bring back buyers who would have lapsed. The question is incrementality: how much of the result disappears when the spend stops. The test design, a geo or audience holdout, is in how to measure incrementality. While the test runs, hold the budget flat rather than letting it grow.

Performance Max needs one setting before the test means anything. It can serve on your brand searches, and Google’s brand exclusions keep it from doing so on Search, Shopping and YouTube search inventory. When the goal is new customers, Google advises focusing on the new customer acquisition goal and brand exclusions.

Then decide what to cut, cap or keep

Put every candidate from the six checks in one list with its monthly spend, and sort it into four groups. Cut what spent through a full buying cycle and converted nothing: the search terms, networks, placements and hours from checks 1 and 2. Cut or exclude what loses money even if every reported sale is real: anything below its break even ROAS on reconciled revenue, like the low margin group in check 3. Neither needs a test, because no test result could make them profitable.

Cap and test what clears break even on reported sales but sits next to the purchase: brand search, retargeting and spend on existing customers. Keep, and consider adding to, what clears break even on the revenue it actually caused.

Then check the result in MER, total revenue divided by total marketing spend. After a real cut, spend falls and revenue holds, so MER rises. If revenue falls far enough that the margin lost is larger than the spend saved, the spend was doing more than the reports showed: put it back and test it properly. Rank what you cut by dollars a year. To size the total before you start, the wasted ad spend calculator models a range for your spend; it is a model, not a measurement. Paid media is one of several places a budget leaks, and where SMB marketing budgets leak covers the rest.