$32M in Google Ads Spend: Lessons Learned

One Google Ads account. An Australian residential solar installer, running since 2017. Over nine years it spent $32,277,473, bought 7.6 million clicks from 836 million impressions, and recorded 251,780 conversions across 507 campaigns.

In August 2026 we were asked to audit it. The blended cost per lead had drifted to $300–400 and the brief was to get it to $150. What follows is what the data said: pulled directly from the Google Ads API, campaign by campaign, search term by search term, back to the first dollar spent.

Each post stands alone. Together they are the audit.

A suggested reading order

If you read them in number order you get the audit roughly as it happened: what was found in the campaigns, then in the match types, then in the conversion count, then what a bidding change cost, then the tracking, the experiments, the lead forms, and then the general lessons. Parts 1 through 8 are case studies; parts 9 through 16 are positions.

If you only read three: part 4 (what the conversions column is made of), part 13 (why most tests can't detect what they claim), and part 9 (the cheapest channel was the one nobody trusted).

What this is not

The client is anonymised throughout. No company name, no domain, no spend figures beyond what the ratios need. The account's conversion tracking double-counted web form submissions during the audit window (part 4 explains how), so absolute conversion counts are the platform's figures, not unique leads. Ratios and comparisons hold, because the same counting applies to both sides of every comparison.

Every number was pulled via the Google Ads API v25 between 26 August and 6 September 2026. Nothing is illustrative.