True Cost Per Lead Case Study: 8 Simulated Businesses, Every Number Checked

An ad platform tells you what you paid the platform. Then it divides that by leads. That is the platform’s cost per lead.

However, a lead costs more than media. For example, a partner may keep a share of revenue. Also, a refund or a chargeback may reverse a sale. A vendor may charge a fee too. In short, all of those belong to the lead.

So CDAI divides every cost by the number of leads. That is the true cost per lead. Dividing the same true cost by sales gives the true customer acquisition cost (CAC).

Each row is one simulated business, shown on its own numbers. Also, nothing is added together across businesses. The gap is how much higher the true figure is than the platform’s.

Simulated data. Seed 1 of three simulated worlds per business. Gap = true CPL versus platform CPL. The full report shows all three seeds for every business.

The gap changes with the business. Where partners, refunds and fees are large, the platform number understates the cost by a wide margin. Because of that, a campaign can look cheap in the ad account and yet lose money. The clinical-trial recruitment business is the control: its only cost is media spend, so its true and platform figures match at $48.40. The engine adds a gap only when a cost exists.

Over 90 simulated days, the roofing contractor generated 2,263 leads and 463 sales. The ad platforms alone would have reported $8.77 per lead. The true cost per lead was $99.97.

Then the table shows where the rest of the cost came from. Every dollar is a cost the simulated business actually carried.

True CAC for this business was $488.64. This is what the business produced in the simulation. However, no decision was acted on, so it measures the business, not an uplift from the engine.

A case study is only worth reading if the arithmetic is right. So the numbers were checked in two independent ways.

1. The engine’s own scripts. True CPL, true CAC, contribution margin and every cost layer were recomputed from the simulated world’s ledger and matched the database (966 of 966 checks) and the engine (966 of 966 checks) on the 24 seed runs.
2. A separate recalculation. After the data was signed, every figure was recomputed with SQL straight from the raw tables: 6,131 checks, 0 mismatches.

In addition, the database was reconciled to the simulated world line by line. Unexplained differences: 0. Our 30-day retest methodology explains how outcomes are graded.

Two of the 26 runs used real developer accounts. One ran on a real HubSpot portal and one on a real Salesforce org. The simulation created the leads and won deals in the real CRM. Then the nightly adapters pulled them back.

A verifier compared the real CRM with the engine’s database record by record. HubSpot passed 14 of 14 required checks. Salesforce passed 14 of 14. In both runs the fingerprint of the CRM’s won-deal list was identical to the database’s.

The runs also found a defect, so we named it and published it. For example, when a CRM deal and a payment charge both describe one sale, the engine stores it twice. The full report prints the exact counts.

• It is a simulation. The 8 businesses, their deal sizes and their costs are ours. The profit levels are properties of those worlds, not what a client should expect.
• No decision was acted on. The results measure each simulated business, not an uplift from following CDAI.
• The vendors were mock vendors. Also, real ad, call-tracking and payment vendors can behave differently.
• Samples are small. Each business ran in three simulated worlds. Read the full report’s intervals, not single figures.

Instead, the point of the exercise is narrower. It tests that the engine’s arithmetic is right on worlds where the answer is known. See the full validation report for per-decision results.

• Tag every lead with its campaign. Keep campaign or UTM fields on the CRM record. See our guide to lead quality.
• Connect closed revenue. The calculation needs to know which leads became customers.
• Add the costs outside the ad account: partner payouts, refunds, chargebacks and fees. Our guide to marketing costs lists them.
• Connect one CRM and one ad account. That is the minimum for true CPL and true CAC. See how it works.