Why does my cost per lead go up when I spend more on Google?

Why does cost per lead rise when a contractor increases spend on a channel?

Because the cost of marketing a report shows is an average at last year’s spend, and the next dollar into the same channel buys the next-best click, not the same one. At the eight-figure Toledo exteriors company where I direct marketing, Google Local Services ran 6% in its lowest-spend year and 32% in its highest, and Google Ads revenue sat in a 24%-wide band for four years while the demo doubled in price.

What is the difference between average and marginal cost of marketing?

The report gives one number per channel: spend divided by the net revenue attributed to it, for the year. That is an average, and it describes the dollars already spent at the volume already run. The number a budget decision needs is different. It is what the next $20,000 will return. The next $20,000 goes to the search a competitor is also bidding on, the homeowner who was going to call anyway, the zip code that was excluded last year for a reason. The average understates the cost of the next dollar in every channel with a finite audience, which in a metro the size of Toledo is every channel.

I did not think about this clearly for years, and the reason is that the report never shows it. It shows 12%, and 12% looks like a rate, and a rate looks like it should hold at twice the volume.

What did Local Services do each time we turned it up?

Google Local Services Ads is the pay-per-lead product at the top of the results. I have turned it up and down five times since 2022, mostly because its cost of marketing swings with platform changes in ways I could not predict. Here is what the swings look like when the spend is put next to them, from the Cost of Marketing report in Lead Perfection. $64,171 of spend at 11.2% in 2022. $175,183 at 31.8% in 2023. $82,636 at 21.0% in 2024. $19,385 at 6.0% in 2025.

The cheapest year is the smallest one. The most expensive year is the biggest one, by almost three times, and it is not close. Part of it is the platform, which keeps changing how leads are priced and dispatched and is the reason I have never been able to predict this line. But the shape is the shape of a channel with a limited number of good leads in it. The first $20,000 buys the homeowners who were searching and the last $100,000 buys the ones who were not.

What did Google Ads do across four years?

The paid search line ran between $138,780 and $194,126 of spend a year from 2022 through 2025, and the net revenue it produced sat between $745,288 and $928,809, a band 24% wide, in every one of those years. Net revenue per dollar was $4.92, $4.78, $5.46 and $5.29. Four years, a spend range of $55,000, and the return per dollar barely moved, which is what a channel looks like when it is already near the volume its audience will support. Cost per demo on that line went from $797 to $1,498 over the same window, so the leads we were buying at the margin were getting worse even as the average held.

I would not have seen any of this on the annual report, because the annual report is one dot per year. It took putting the dots on one chart, spend on one axis and cost of marketing on the other, to see that each channel has a region it has actually been run in and a cliff at the edge of it.

Where this is wrong

For a channel with a tiny history, the first dollars are unrepresentative in the other direction. Our first year on one aggregator ran at 137% cost of marketing on fourteen demos and settled at 14.5% the year after (It Launched at 137% COM). A launch year is not a marginal read either. The rule only works inside the range a channel has actually run, and outside that range the honest answer is that nobody knows, including the vendor.

What to do

• Tonight: for each paid channel, write down annual spend and cost of marketing as a pair, one pair per year, and look at whether the expensive years are the big ones.
• This quarter: when adding budget to a channel, add it in increments the channel has run before, and read the marginal result at ninety days, not the blended one.
• Ongoing: treat any vendor projection that applies this year’s rate to double the spend as a guess until the channel has run at that spend.

Sources: Lead Perfection “Cost of Marketing by Source,” Toledo, 2022 to 2025 full year, first party. Platform-change attribution is the author’s read, not documented in the report.

Related notes