All guides

Where Google Ads budget actually leaks

Marketing6 min read

Open your Google Ads account. Go to the search terms report, set the range to the last 90 days, sort by cost descending. Nearly every account we audit has something in that top twenty nobody would ever have chosen to bid on, quietly spending four figures.

That report shows the gap between the keywords you bought and the searches you actually paid for. Those are different things, and the distance between them is where your money goes.

Match types are a spending decision

A keyword in exact match brackets buys that search and its close variants. Phrase match buys anything containing that meaning. Broad match buys whatever Google’s models decide is related, which in practice includes competitor names, job seekers, students writing essays, and people hunting for a free version of what you sell.

Broad match is not automatically wrong. It works when three conditions hold at once: you have enough conversion volume for Smart Bidding to learn from, your conversion tracking fires on something that represents real money, and somebody maintains negative keyword lists every week. Take away any one of those and broad match becomes an expensive random number generator.

Most small accounts fail the first condition without realising. Smart Bidding needs a steady stream of conversions to model against. Feed it eleven form fills a month and it is guessing, expensively, using your budget as the training data.

The negative list is the actual work

Negative keywords are the unglamorous half of paid search and the half that pays. Start with the obvious intent blockers: free, cheap, DIY, jobs, career, salary, template, torrent, reddit. Then work through your own search terms report and add what you find.

Do this weekly for the first two months of any campaign. After that, monthly holds. The list only grows, and it is the single asset in a PPC account that compounds, because every term you exclude keeps costing you nothing forever.

One caution. Negative keywords in broad match can block more than you expect, and adding “free” as a broad negative will also block “free consultation for commercial roofing” if that phrase matters to you. Use phrase-match negatives when the word cuts both ways.

What a lead should cost

LocaliQ publishes benchmark data drawn from thousands of Google Ads and Microsoft Ads accounts, which makes it one of the few reference points here grounded in real spend rather than vendor marketing. Their figures put the average cost per lead across all industries at $70.11, up roughly 5% year over year, with cost per lead rising in 13 of 23 tracked industries.

Averages across 23 industries hide enormous variation, so treat that as a sanity check and nothing more. Legal and finance sit far above it. If you are paying $200 a lead in a category where the benchmark is $40, something specific is wrong and it is usually match types, geography, or a landing page that does not match the ad.

The more useful number is your own, tracked backwards. Cost per lead matters less than cost per lead that turns into revenue, and those two diverge fast when broad match is feeding your pipeline with unqualified traffic.

Tracking the wrong conversion

Plenty of accounts optimise toward a conversion that costs nothing to produce. Newsletter signups, PDF downloads, a click on a phone number that nobody answered. Smart Bidding will faithfully maximise whatever you tell it to, and if you tell it to maximise low-value actions, it will find you an enormous quantity of people who take low-value actions.

Assign values. A quote request outranks a brochure download, and telling Google that changes what it buys on your behalf. Be aware that the conversion counts themselves are understated for reasons covered in why your analytics under-reports. If you can push closed-won revenue back into Google Ads through offline conversion imports, the bidding models get to optimise against money instead of form fills. That requires your CRM and your ad account to talk to each other, which is its own project and worth reading about in speed to lead.

The landing page is part of the ad

Ad spend buys a click. Everything after the click decides whether the click was worth buying, and a page that takes six seconds to render on a mobile connection will lose a meaningful share of the traffic you just paid for before it renders anything at all.

Vodafone measured this directly: a 31% improvement in Largest Contentful Paint produced 8% more sales and an 11% better cart-to-visit rate. That was an organic and paid mix, but the mechanism is identical for paid traffic, except with paid traffic you are paying for every visitor who bounces. We covered the measurement side of that in site speed and revenue.

Match the page to the ad, too. Someone who clicked an ad for commercial roof repair should land on commercial roof repair, not your homepage with a nav bar and six service categories.

A 30-minute audit

Pull the search terms report for 90 days and read the top 30 by spend. Anything irrelevant becomes a negative keyword before you close the tab. Check which conversion actions are marked primary, and confirm each one represents something you would be pleased to receive. Look at your geographic targeting settings, because the default includes people “regularly in or showing interest in” your area, which is broader than it sounds. Then load your top landing page on a phone over mobile data and count the seconds.

That is usually enough to find the leak. The fix is rarely clever, and it is almost never a new bidding strategy.

If you want someone to run that audit properly against your account, get in touch or read more about how we approach marketing.

References