SellerLeak — PPC Profit Optimizer

Profit-focused Sponsored Products review

Upload a Sponsored Products Targeting or Search Term CSV/TSV and confirm your SKU costs and report context. SellerLeak calculates observed ACoS, contribution-based break-even ACoS, and a profit-target threshold that preserves your chosen contribution per unit.

What you receive

• Up to three prioritized, evidence-linked next steps.
• Small manual bid-test candidates when margins, attribution and current bids are confirmed.
• Zero-order target reviews with a sourced conversion benchmark.
• Negative-exact candidates for search terms you confirm are irrelevant.
• A text summary, Markdown report, standalone HTML review and action-review CSV.

Example — synthetic data

A product sells for €20 and has €10 contribution before Ads. To retain €3 per unit, its target ACoS is 35%; break-even is 50%. A target with €80 spend and €200 attributed sales has 40% ACoS and remains profitable, but falls below the retained-contribution goal. With sufficient mature data and a confirmed €0.80 bid, the Agent proposes a bounded test at €0.70. This is an experiment, not a savings guarantee.

Inputs

One SP Targeting or Search Term CSV/TSV, report dates and extraction date, attribution window, marketplace/currency, product mapping, confirmed costs, current bids and campaign goal. Normalized JSON is also supported. Base CSV adapters support documented Italian and English headers. Other schemas require explicit mapping; XLSX can be converted only when the runtime has a trusted spreadsheet reader.

Boundaries

Sponsored Products only. No automatic Amazon changes or Amazon-uploadable bulk file. Sparse or incomplete data produces monitoring or data-review requests. Search-term evidence alone does not justify changing the parent keyword bid.

Contribution estimates exclude fixed costs and income taxes and do not measure advertising incrementality. Historical excess spend is not guaranteed recoverable money. Whole-account profit and TACoS are outside this version's engine output.

Data handling

The bundled scripts make no network calls and require no Amazon credentials. Uploaded inputs are processed in the platform runtime by its configured AI service, as listed in the Agent's Data Sharing Declaration. All bundled examples are synthetic.

Expertise

Built from hands-on Amazon seller operations and contribution-margin analysis, using deterministic calculations, source-row traceability and explicit missing-data gates. Independent product; no claim of Amazon endorsement.