Most wasted Google Ads spend comes from a handful of settings and habits, not bad luck in the auction. Start with conversion tracking, because every bid optimizes toward what you count. Then work through the 30 checks below, each rated by how much budget it typically leaks.
- High: misdirects bidding across the account or wastes a large block of spend.
- Medium: leaks money in part of the account, or slowly.
- Low: hygiene. Rarely the reason results are bad.
Before you audit: check conversion tracking
Smart Bidding, Performance Max and every CPA in your reports rest on conversion data. If it counts the wrong things, or counts them twice, every other fix is tuned against a broken number, and that waste never shows up as a line item.
| # | Check | Leak |
|---|---|---|
| 1 | Only real business outcomes (purchases, qualified leads, booked calls) are primary conversions; page views and clicks are secondary | High |
| 2 | No action is counted twice, such as a Google Ads tag and an imported GA4 key event both set as primary | High |
| 3 | Lead actions count “One” per click; purchases count “Every” | Medium |
| 4 | Conversion values reflect real revenue, or a realistic value per lead stage | High |
| 5 | Lead gen accounts import qualified leads or opportunities from the CRM | High |
| 6 | Auto-tagging and enhanced conversions are on, and the click ID survives redirects and forms | Medium |
Check 1 is the leak I find most often. If a contact page view is primary, Smart Bidding happily buys visitors who glance at that page and leave.
Check 5 separates B2B accounts that scale from ones that fill the CRM with junk. Without offline data, Google optimizes for form fills, and the cheapest form fills are often the worst leads.
Google Ads and GA4 will still disagree after these fixes; the guide on why GA4 and ad platforms don’t match explains which number to trust. If tracking needs a rebuild, that’s attribution work to finish before you touch bids.
Account and campaign structure
| # | Check | Leak |
|---|---|---|
| 7 | Search campaigns are opted out of the Display Network | High |
| 8 | Location option is “Presence”, not “Presence or interest”, unless you sell to people researching your area from elsewhere | High |
| 9 | Auto-applied recommendations can’t add keywords, upgrade to broad match or change bidding without review | Medium |
| 10 | Brand and non-brand search run in separate campaigns and budgets | Medium |
| 11 | Each campaign gets enough conversions to learn; no sprawl of tiny campaigns | Medium |
Checks 7 and 8 leak because they’re defaults on new Search campaigns. Display traffic behaves very differently from search but hides inside the same campaign totals. “Presence or interest” can show a local service’s ads to people who merely searched about the area.
Check 10 is indirect waste. Cheap brand conversions mixed with generic terms hide what new-customer acquisition really costs, so the generic side gets overfunded. Judge Search Partners as its own segment too, and keep it only if it converts at an acceptable cost.
Search terms and negative keywords
Waste is most visible here. Open the last 90 days of search terms, sort by cost, and start at the top.
| # | Check | Leak |
|---|---|---|
| 12 | High-cost search terms with no conversions are reviewed monthly and excluded | High |
| 13 | Shared negative lists block job seekers, freebie hunters, DIY and existing-customer queries | High |
| 14 | Broad match runs only with Smart Bidding and trustworthy conversion data | High |
| 15 | Brand and competitor queries don’t leak into generic campaigns | Medium |
| 16 | Campaign-level negatives route each query to the campaign built for it | Low |
For a B2B software account, the standing negative list usually covers jobs (“careers”, “salary”), learning (“course”, “certification”) and existing customers (“login”, “support”, “cancel”).
Check 14 matters because broad match relies on bidding signals to decide which loose matches are worth buying. With broken tracking or manual bids it mostly buys volume, so stay on phrase and exact.
Three cautions. Google only lists queries that meet its privacy thresholds, so some spend never maps to a visible term. Negative keywords don’t match close variants, so add plurals and common misspellings yourself. And a broad or phrase negative on “free” also blocks “free trial”; if you offer one, exclude narrower phrases like “free download”.
Bidding and budgets
| # | Check | Leak |
|---|---|---|
| 17 | Target-based bidding runs only where a campaign has steady conversion volume | High |
| 18 | Target CPA and target ROAS come from unit economics, not platform suggestions | Medium |
| 19 | Budget goes to campaigns with the best marginal return, not spread evenly | Medium |
| 20 | Bid strategies aren’t edited every few days; seasonality adjustments cover short promotions | Low |
My working threshold before moving a campaign to target CPA or target ROAS is around 30 conversions a month. It’s a rule of thumb, not a Google requirement, but below it bidding tends to swing.
On check 18, a loose target lets Google pay more than a customer is worth. A tight one starves the campaign, and teams then “fix” volume with broad keywords, which reopens check 14.
On check 19, look for campaigns limited by budget at a good CPA while weaker ones have room to spare; shifting money between them is often the fastest win.
Performance Max: brand cannibalization and overlap with Search
Performance Max spends where conversions come easiest, often people already searching your brand or returning to your site. That can flatter its ROAS while adding little new revenue.
| # | Check | Leak |
|---|---|---|
| 21 | Brand exclusions are applied, or brand traffic in PMax is a deliberate choice | High |
| 22 | PMax is judged on new customers, not blended ROAS, using the new customer acquisition goal where possible | High |
| 23 | Final URL expansion is off, or excludes blog, careers, support and login pages | Medium |
| 24 | Core non-brand terms also exist as Search keywords you control | Medium |
Brand cannibalization
Check PMax’s search term reporting for brand queries, then compare brand campaign clicks before and after PMax launched. If they fell while PMax conversions rose, PMax is likely claiming customers who would have come anyway. Brand exclusions fix the symptom; a holdout test shows whether any brand spend is incremental, and the incrementality testing guide explains how to run one.
Should PMax run alongside Search?
For ecommerce with a healthy product feed, usually yes: PMax reaches Shopping inventory Search can’t. For lead gen and B2B it’s a closer call, because PMax can find cheap, low-quality leads on Display and YouTube unless it optimizes on CRM-qualified conversions.
When both are eligible, Google prioritizes a Search campaign whose keyword is identical to the query; otherwise the higher Ad Rank wins. Keep your most valuable terms as exact or phrase keywords in Search, where you see queries and control copy, and let PMax work the rest. Add PMax only once tracking is clean and Search covers the core terms well.
Ads, assets and extensions
| # | Check | Leak |
|---|---|---|
| 25 | Responsive search ads repeat the ad group’s core term and offer; pins are rare | Medium |
| 26 | Sitelinks, callouts, structured snippets and relevant call or image assets are live | Low |
| 27 | Automatically created assets and generated text are reviewed for claims and prices | Low |
Weak ads rarely burn money outright. They lower click-through rate and relevance, which tends to raise cost per click. Ad Strength rates asset relevance and variety, not results, so don’t rewrite a converting ad to chase “Excellent”. Pin only what must appear, since every pin cuts the combinations Google can test.
Landing pages and post-click
| # | Check | Leak |
|---|---|---|
| 28 | Every final URL returns a 200, loads fast on mobile and keeps tracking parameters through redirects | High |
| 29 | Each ad group lands on a page matching its query and offer, not the homepage | High |
| 30 | Forms, checkout and phone numbers are tested end to end, and tests show up as conversions | Medium |
For check 28, crawl every final URL in ads, sitelinks and PMax asset groups; ad disapprovals don’t reliably catch broken pages fast. For check 29, someone searching “SOC 2 compliance software for startups” should land on exactly that, not a generic product tour.
Prioritizing fixes by wasted spend
Estimate each finding’s monthly waste as affected spend multiplied by the share you judge wasted, then rank by that figure and by effort.
Worked example with made-up round numbers: an account spends $20,000 a month.
| Finding | Affected spend | Judged wasted | Est. monthly waste |
|---|---|---|---|
| Display opt-in on two Search campaigns | $2,000 | Most of it | ~$1,500 |
| No negative lists | $4,000 | About a third | ~$1,300 |
| PMax spending on brand queries | $3,000 | About half | ~$1,500 |
| Target CPA above customer value | $6,000 | Hard to size | Fix after tracking |
Here, the first three fixes take an afternoon and could recover about $4,300 a month. The bidding fix waits because it depends on accurate conversion data.
The order I use:
- Fix conversion tracking (checks 1-6)
- Change leaking defaults: Display opt-in, location, auto-apply (7-9)
- Add negative lists and review top search terms by cost (12-15)
- Apply PMax brand and URL exclusions (21, 23)
- Fix broken or mismatched landing pages and test forms (28-30)
- Only then change bid strategies, targets and budgets (17-19), one campaign at a time
Log the date of every change so you can read its effect later. This sequence is the first month of most performance marketing engagements I take on: stop the leaks, then scale what’s left.
Get it built
If you want an outside read on where your Google Ads budget is going, I can run the audit and hand your team a ranked fix list. The Growth Audit is $1,500 fixed and credited if we continue. See pricing or get in touch.