ML Live Bidding
ML Live Bidding
ML Live Bidding adds a learning layer on top of the deterministic engine. Models trained on your account's own performance propose bid moves continuously through the day — sensitive to time-of-day demand — across every eligible target, not just your proven winners. The model proposes; your Guardrails still decide what is allowed to run. Available on Hosted Growth and Hosted Pro.
A learning layer, inside the envelope
Where the daily Runbook and Rally-Control apply explicit rules, ML Live Bidding learns the patterns in your own data — clicks, conversion behavior, product economics, and live Amazon signals — and proposes bids that fixed rules alone would miss. It is never a black box with your credit card: every proposal is clamped to your bid floors and ceilings, your shared daily movement cap, your confidence threshold, and per-target cooldowns before anything can execute. The model proposes within the envelope; deterministic logic decides.
Day-and-night aware
Demand isn't flat across the day. ML Live Bidding reads time-of-day and day-of-week patterns — dayparting — so a target can lean in during the hours it actually converts and ease off when it doesn't, instead of carrying one static bid around the clock. It runs across an active window you control (a sensible daytime default, extensible toward 24-hour coverage), and when you extend it the same dayparting carries your overnight bids.
Broad coverage, not just the winners
Rally-Control works a narrow, weekly-curated pool of proven converters; ML Live Bidding casts wider, continuously tuning across all of your eligible targets so the long tail gets attention too. The two are designed to run side by side — they share the same safe execution substrate, the same combined movement cap, and the same cooldowns, so running both never double-moves a target.
Live on Growth and Pro — opt-in, and reversible
ML Live Bidding is live for Hosted Growth and Hosted Pro. Switching it on is your choice — opt-in per account, on when you want it. You can watch it in preview first, every move is reconciled afterward, and the emergency stop is always one tap away. AI and ML narrate and propose — they never see raw data they don't need, and never bypass a Guardrail.