# Ad measurement

**Eight months of ad bidding, trained on conversions that weren't real.**

Divine Power · For a home-services brand

A home-services brand was paying for leads through automated ad bidding that learned from its own tracking. I orchestrated several existing AI skills as subagents to audit that tracking and work through large Google Analytics and Ads exports, and directed the audit with follow-up questions. The ad platform reported 7.3 conversions for every real one. Separately, the real cost was $946 a lead, against a $91 benchmark. The audit is done; the tracking rebuild is in progress.

- **Problem:** The ad platform reported 7.3 conversions for every real one, and automated bidding learned from that inflated data for eight months.
- **My role:** Founder of Divine Power. I orchestrated several existing AI skills as subagents and directed the audit with follow-up questions.
- **What I did:** Orchestrated existing AI skills as subagents to audit the tracking schema and digest large Google Analytics and Ads exports, then set the conversions the platform reported against the real ones.
- **Outcome:** A real cost of $946 a lead against a $91 benchmark, and a projected \~$700K growth opportunity identified. The tracking rebuild is in progress.

## Results

- **$946** real cost per lead, against a $91 benchmark. A real cost, measured in the audit; $91 is the benchmark
- **7.3x** conversions reported for every real one. Automated bidding had learned from that signal for 8 months
- **\~$700K** growth opportunity identified. Projected, not realized: the tracking rebuild is in progress

_Client details are omitted or blurred for confidentiality._

![Two bar charts. Cost per lead: $946 real, from the audit, against a $91 benchmark. Conversions reported per real one: 7.3 reported by the ad platform against 1 real.](https://maujimenez.io/work/ad-measurement/cost-per-lead-and-conversions.png)
_Built from the audit's figures. The same numbers are in the table below\._

## Why the numbers misled

Automated bidding spends toward whatever the tracking reports as a conversion. At this brand, the ad platform reported 7.3 conversions for every real one. For eight months the bidding learned from that inflated signal.

## How the agents ran the audit

The evidence was spread across a tracking setup and exports too large to read by hand. I orchestrated several existing AI skills as subagents to audit the brand's tracking schema and to digest its Google Analytics and Google Ads exports, and I directed the audit with follow-up questions. The agents made the volume possible.

## What the audit found

| Measure                                   | Finding |
| ----------------------------------------- | ------- |
| Conversions reported for every real one   | 7.3     |
| Months of bidding trained on that signal  | 8       |
| Real cost per lead                        | $946    |
| Benchmark cost per lead                   | $91     |
| Growth opportunity identified (projected) | \~$700K |

The audit found two separate gaps. The real cost per lead was $946, against a $91 benchmark. Separately, the ad platform reported 7.3 conversions for every real one. The \~$700K is an opportunity identified from the audit, not revenue earned.

## Where it stands

The audit is complete. The tracking rebuild is in progress, so that the bidding learns from real leads instead of inflated ones.

---

- Web page: https://maujimenez.io/work/ad-measurement
- Index for AI tools: https://maujimenez.io/llms.txt