Common questions.

First week: we connect to your Amazon accounts via official SP-API, ingest 90 days of historical sales, PPC, and inventory data, and run a baseline analysis. Most brands see their first cost-saving actions within 14 days (waste keywords, structural bid issues), and meaningful TACoS movement within 30–60 days as the organic flywheel responds.

You own everything. Your Amazon data never leaves your encrypted account, and you can export your full performance history, recommendations log, and baseline reports as CSV at any time. No 12-month contracts, no data lock-in.

Most of the day-to-day work — blocking wasteful search terms, trimming bids on losing keywords, reallocating budgets — runs autonomously once the system has proven those decisions reliably improve performance. Higher-stakes moves (pausing a major campaign, changing a list price, restructuring an ad group) get reviewed by a human and discussed with you before they ship. Over time, more decision types graduate to fully autonomous as the system learns from each outcome.

Most tools optimize one lever at a time and leave you to operate the dashboard. Most agencies do the same work manually, with junior analysts on spreadsheets. Circes runs all five levers as one connected system that learns from every outcome — routine optimizations execute autonomously, while the human team focuses on strategy, structural changes, and your business relationship. The result: faster decisions than a manual agency can match, and better-judgment decisions than a tool can make on its own. And because the system measures whether each action actually improved performance, it gets sharper every week.

No. We inherit your existing structure, learn from its history, and improve incrementally. For structural changes — pausing a major campaign, restructuring an ad group — we walk you through the reasoning before we make the change. Usually it's because the math shows the campaign is structurally unprofitable, not because we'd do it differently.

This is one of the most common reasons PPC tools make bad calls. We exclude the last 3 days of data from every analysis because Amazon's attribution isn't settled yet — spend is recorded immediately but sales take days to attribute. Most tools cut bids prematurely on apparent ACoS spikes that resolve themselves once data settles. We don't.

The system tracks your FBA + AWD inventory in real time and computes days-of-cover per SKU. When a SKU goes "defensive" (low cover), Circes automatically blocks all growth-oriented actions on that SKU — no scaling ads into a stockout. When you restock, it shifts into "recovery" mode and pushes top-of-search aggressively to rebuild rank velocity.

All API credentials are encrypted with AES-256-GCM at rest, never logged, and never exposed in our application. We connect through Amazon's official Selling Partner API authorization flow — the same security model Amazon uses for its own internal tools. You can revoke our access in one click from your Amazon Seller Central account.

Every recommendation Circes makes is tracked against the actual outcome 7–14 days later — did ACoS improve, did sales hold up, did the change measurably move the needle? Action types that consistently improve outcomes graduate to fully autonomous; ones that don't get revised or retired. To date, recommendations on our founder's brand are applied at an 82%+ rate, with several action types (bid trims, keyword harvests, budget increases on winners) already running fully autonomous. Your account makes the system smarter, and the system gets smarter on your behalf as time goes on.