Which metrics matter most in an AI-controlled auction?
Use a four-layer model.
| Layer | Core question | Example metrics |
|---|---|---|
| Business economics | Did advertising create profitable growth? | Contribution profit, gross profit, CAC, payback, margin |
| Incrementality | What happened because of the ads? | Incremental conversions, lift, iCPA, holdout results |
| Blended acquisition | Is total marketing becoming more efficient? | MER, blended CAC, new customers, total revenue |
| Platform diagnostics | What changed inside the channel? | Spend, conversions, ROAS, CPA, CPC, CTR, impression share |
How do you measure profitability?
Revenue is not profit. Use contribution economics where possible.
For ecommerce: Contribution profit = revenue - cost of goods - shipping - discounts - payment fees - variable fulfillment costs - ad spend
For lead generation: Expected lead profit = leads × qualified rate × close rate × gross profit per sale - ad spend
Illustrative lead-gen example: 500 leads × 30% qualified × 20% close rate × $3,000 gross profit = 30 customers × $3,000 = $90,000 gross profit - $60,000 ad spend = $30,000 contribution.
What is blended CAC?
Blended customer acquisition cost = total acquisition spend ÷ total new customers
This metric is useful when channel attribution shifts but total customer growth can be verified independently.
What is marketing efficiency ratio?
MER = total revenue ÷ total marketing spend. Use contribution-margin MER when revenue quality varies. MER is a business-level guardrail, not a replacement for channel diagnostics.
What is the difference between attributed and incremental conversions?
| Metric | Question answered | Limitation |
|---|---|---|
| Attributed conversions | Which ads receive credit under the attribution rules? | Can include conversions that may have happened anyway |
| Incremental conversions | How many additional conversions occurred because of advertising? | Requires an experiment/model and sufficient data |
A campaign can have strong attributed ROAS and weak incrementality, especially when it captures brand demand, repeat customers, or remarketing activity.
How should lead quality be measured?
Build a stage-quality report comparing front-end leads to qualified pipeline and revenue.
| Campaign | Leads | Qualified | Opportunities | Customers | Cost/lead | Cost/qualified | CAC |
|---|---|---|---|---|---|---|---|
| Campaign A | 500 | 100 | 30 | 10 | $40 | $200 | $2,000 |
| Campaign B | 300 | 120 | 45 | 18 | $60 | $150 | $1,000 |
Campaign B has a worse front-end CPL but a better CAC. Optimizing to CPL alone would choose the wrong campaign.
How do you communicate results when causality is uncertain?
Use calibrated language. Avoid: "Performance Max generated $500,000 in revenue." Use: "Google Ads attributed $500,000 in revenue to Performance Max under the current attribution settings. Brand demand, repeat customers, and cross-channel effects may contribute. Incrementality testing and blended business metrics are used to estimate causal impact."
Common mistakes
- –Reporting platform ROAS as proven incrementality
- –Optimizing to the cheapest lead
- –Ignoring margin
- –Mixing new and repeat customers
- –Using last-click only
- –Removing platform diagnostics entirely
- –Demanding perfect attribution before making any decision
- –Changing multiple variables without a test plan
- –Ignoring data quality
Related guides: Performance Max: Profitable Not Incremental, Incrementality Testing for PPC, GA4 vs Google Ads Conversion Mismatch, Google Ads Benchmarks 2026, Google Ads Management
Source notes:Search Engine Journal, "How To Measure PPC Performance When AI Controls The Auction," April 13, 2026. Google Ads Help: Conversion Lift documentation and attribution reports. Google Ads Help: Smart Bidding and Performance Max documentation.