Deep contractGoogle Ads
How to find which products earn their ad spend
SKU-level truth across Shopping, Meta catalogue ads and Pinterest: which products convert the spend they take, and which quietly burn it.
Contract v1.0.0last verified reviewed
IntentRank products by advertising efficiency with names, not item IDs, so the merchandising decision is readable.
- 5
- Steps
- 6
- Tools
- 3
- Rules
- 3
- Failure modes
When to use it
Use it when
- Monthly on any catalogue-driven account.
- Before excluding or boosting product groups.
- When blended ROAS is fine but someone suspects it hides losers.
Do not use it when
- Catalogues under ~50 products: read the campaigns directly.
- Windows shorter than the purchase cycle: slow considered purchases need longer reads.
What it needs first
Required
Preconditions
- Catalogue campaigns with product-level reporting live.
- One currency, or explicit conversion notes.
Inputs
window- Analysis window.Default: last 30 days.
spend_floor- Spend below which a product is not judged.Default: 10x target CPA equivalent.
Procedure
In order. Every tool named is one the gateway ships, and links to its reference.
Pull product-level Shopping performance with spend and conversions.
On Meta, join product-level insights to the catalogue so rows carry names and brands.
On Pinterest, run the conversion product report.
Merge by item ID, then bucket: earners, burners above the spend floor, and unjudged.
If GA4 ecommerce is healthy, sanity-check the top rows against site revenue.
Decision rules
The observable condition, what it lets you conclude, and what takes the conclusion back.
When
A product spent above the floor with zero conversions on every platform
Conclude
Burner confirmed; propose exclusion with the exact spend as evidence.
Unless
It is a new launch inside its ramp window; date it instead.
When
A product earns on one platform and burns on another
Conclude
Platform fit finding, not a product finding; propose moving budget, not delisting.
When
The top 20% of products carry over 80% of catalogue spend
Conclude
Report concentration; the long tail's problems are secondary by construction.
Evidence every conclusion must carry
- Every row carries product name, item ID, platform, spend, conversions, window.
- Exclusion proposals quote total spend saved over the window.
Where the agent stops
- Exclusions and product-group changes are merchandising decisions applied by humans in platform UIs.
- The gateway writes none of this; the deliverable is the ranked list.
How it goes wrong quietly
The cases where the analysis is wrong and still looks right. Read them before trusting a number.
- Item IDs differ across platforms for the same SKU; merge failures silently double-count. State the join rate.
- Returns are invisible to ad platforms; high-return products look like earners. Flag categories with known return rates.
- Price changes inside the window shift ROAS without any media cause; note repricing dates when known.
What the answer contains
- 01Earners
- Top products by efficiency with volumes.
- 02Burners
- Spend, zero-or-poor conversion products above the floor.
- 03Platform fit
- Products whose efficiency differs sharply by platform.