← Eddie J. Smith

Results

What the numbers actually said once someone opened them.

Here are six accounts I've been inside.

Before you read them.

Every one of these has a note saying where it stops. Best month rather than average, estimate rather than money in the bank, one channel out of ten. You can open those notes on each card.

I went through 39 accounts. Not all 39 were clients: the set includes prospects who never signed and ventures I own. Most didn't hold up once I was in the data, so they're not here. Which is also why a few of these look modest next to what you'll see on other people's sites.

CRM-measured 24 months captured (2024-07 to 2026-06); the story uses 2025-06 to 2026-06

A home-services contractor

Its Google Ads account received job values from its field-service software, and about 87% of the value it tracked over 24 months came from them. From October 2025 those job values thinned, and from January to June 2026 the job-level actions recorded almost none.

The number

Tracked conversion value, mostly job values sent by the field-service software for booked jobs, ran $60,045 to $101,930 a month from June to September 2025, the strongest four months in the 24 captured; it fell to $12,504 in October 2025 and was under $35 a month from January to June 2026 on about $46,000 a month of spend, while lead and call conversions kept being counted.

What this one does not cover
  • This describes a past engagement that ended in December 2025. It is not a current client.
  • The drop did not begin when the engagement ended. Tracked value fell to $12,504 in October 2025 and $21,718 in November as job-level conversions from the field-service software thinned, then read $60,121 in December, about a third of it from an uploaded lead action rather than the field-service job actions.
  • Over the 24 months captured, about 87% of tracked value came from job values sent by the field-service software; the rest was assigned values on calls and leads, which are not revenue.
  • Tracked conversion value shows what the tracking recorded, not how well the ads performed. Campaign mix and tracking setup changed over the 24 months, and the data cannot separate the two. No improvement and no comparison between managers is claimed.
  • From January to June 2026 conversions were still counted (371 to 478 a month), but they were lead and call actions carrying almost no value ($11 to $34 a month in total), and the job-level actions recorded almost none. Booked revenue after October 2025 is not in the evidence, so whether demand changed cannot be measured here.
  • In the field-service export for the 365 days to October 23, 2025, about 64% of revenue on call-linked jobs arrived with no source recorded, and most of the rest was not paid search, so none of it is credited to the ad account and any per-channel figure is a floor.
  • No change log exists in the evidence, so nothing is attributed to a specific action.
Modeled calendar 2025

A regional internet provider

They could tell you their ad spend to the dollar. Ask what a customer actually cost and nobody had ever worked it out. It's in their own sales records, so we did: $246 on Google in their best market.

The number

Modeled cost per acquired subscriber of $246 and $323 on Google in two markets and $467 on social, from the client's own 2025 lead-to-sale rate of 11.99%.

What this one does not cover
  • These are estimates. Nobody bought anything at this stage. Every conversion is a lead; no purchase event exists, so no revenue figure and no return can be stated.
  • The 11.99% close rate (2,533 sales from 21,131 leads) comes from the client's own 2025 lead and sales records, as reported in an analysis prepared for the client. It is quoted, not estimated, and it is one rate for all sources and both markets, applied to each platform's 2025 cost per lead; actual close rates by platform are not known.
  • The figures cover calendar 2025. Our campaigns did not carry all of the Google spend for the whole year (in the larger market they carried part of it from May 2025 and all of it from August), so the figures describe the accounts over the year, not our work alone.
  • The best figure is one platform in one market. The same model gives $323 in the other market and $467 on social. Read the table, not the best number in it.
  • No lifetime value, payback or return is available: the client's reporting records no churn model.
  • Lead volume itself carries a caveat, because several form actions are counted at once.
  • Costs have risen since in the larger market. On that account's own conversion count (a different lead count from the client's records), cost per lead was $24.03 from May to December 2025, $45.44 in August 2026 and $50.56 for September 1 to 28, 2026, so the 2025 model is not a current run rate.
Platform-reported 7 months, 2025

A tree-service company

Calls and form fills were the whole measure here. Nothing in the account tracked revenue, so the honest number is what each one cost, and how the count moved as the budget grew.

The number

255 calls and form fills on $16,500.64 of ads over seven months (March to September 2025), $64.71 each.

What this one does not cover
  • Leads are calls and form submissions as the account counted them at the time, with no revenue value attached.
  • Counting changed inside and after this window. Until 2025-10-22 two form actions were counted and every call from an ad counted, including repeat calls, so one inquiry could count more than once; the form event was tightened to one per ad click on 2025-06-20. Later months are not comparable and this is not a current run rate.
  • Seven months from a small base: monthly spend rose from about $1,258 to about $4,617 over the window, and monthly cost per lead moved between $52 and $84.
  • No improvement is claimed. Lead volume rose with spend.
Cash-verified 3 clean months

An apparel brand being told it made money it never made

A setting inside Google was quietly adding a fixed value to new-customer orders. In March 2026 that put CA$1,693.56 into their reported sales that no order backed, turning a 3.59x return into a reported 3.85x. Dropping it to a nominal $0.01 was the recommendation on file, and from April the added value was under CA$7 a month.

The number

CA$63,146 in ads produced CA$385,651 in tracked store sales, over three clean months.

What this one does not cover
  • Three months back to back with no gaps, stated in Canadian dollars. The return is the same either way; the dollar totals are not.
  • The earlier month affected by value inflation is excluded rather than restated, so no platform-added value reaches this total.
  • No lift is claimed. The padded period is not a 'before' to improve on, because its numbers were not measuring the right thing.
  • The final month was captured on its closing day and is still settling.
  • Monthly returns were 5.38x, 6.66x and 6.16x, so no upward trend is claimed.
Cash-verified 14 months straight

An apparel brand, fourteen months straight

Fourteen months of tracked store sales with no gaps. Every dollar of it traces back to an actual order in their store, not to a value the ad platform assigned.

The number

$139,019 in ads produced $1,386,770 in tracked store sales across 14 straight months (July 2025 to August 2026), 9.98x.

What this one does not cover
  • Fourteen back to back complete months of tracked store sales, averaged across the whole period. The 15.43x peak month is a peak and is labeled as one.
  • No improvement is claimed. There is no engagement document and no change history, so this is the account held up while we ran it, which is not the same as us lifting it.
  • Nothing here speaks to incrementality. Tracked return is what the platform measured, not what the business would have lost without it.
  • The latest months are weaker than the window average. August 2026 returned 8.3x and September 1 to 28 returned 6.7x (outside the window, still settling), and tracked revenue in both ran below the same days a year earlier: about 23% lower in August on about 2% more spend, and about 46% lower in September 1 to 28 on about 5% more spend (about 38% lower when purchases are dated by conversion date). Read the 14-month figure as a record of that window, not a current run rate.
  • July 2026 was re-pulled two months after it closed and settled about 12% higher than its closing-day capture; the figure uses the settled value.
Cash-verified first 41 days

A membership brand, first 41 days

Forty-one days. Small money. The point of including it is that the purchase tracking was right from day one, so the cost per sale is a fact rather than a guess.

The number

$3,427 in ads produced $16,628 in sales. About $19 to make a sale.

What this one does not cover
  • Forty-one days (January 1 to February 10, 2026). Short window, small absolute spend, and nothing here extrapolates to a month or a year.
  • Tracked purchase revenue from the site's purchase event, not an assigned value.
  • Purchases are any tracked purchase on the site during the campaign, which may include more than memberships.

Want to know what your account really says?

Your dashboard will not tell you. The audit tells you what a lead is actually worth to you, and how much of your revenue you currently cannot trace at all. Most owners have never seen either number.

Book the Owner's Math audit