Dynamic Bidding

Dynamic Bidding is the Amazon mechanism that adjusts your submitted bid in real time at the moment of each auction, based on the model's prediction of conversion probability for that specific impression. It is one of three bidding strategies — the other two being Fixed Bids and Rule-Based Bidding.

Two sub-modes exist.

Dynamic Bids — Down Only

Amazon may reduce your bid by up to 100% (all the way to zero) when its model predicts low conversion probability for the impression. It will never bid above your submitted number.

• Use case. Default for any new campaign, any unproven keyword, any campaign with a tight Target ACOS, any auto or broad discovery campaign. It cannot blow up the budget — that's its primary virtue.
• Trade-off. You leave the high-converting tail of impressions to competitors who run Up and Down. Branded competitors will systematically outbid you for the most valuable clicks.
• What you give up. Typically 5–15% of incremental conversions versus Up and Down on the same campaign, in exchange for tightly bounded variance.

Dynamic Bids — Up and Down

Amazon may raise your bid by up to +100% for placements where it predicts high conversion probability, and reduce it for low-probability placements.

• Use case. Mature campaigns with proven CVR data (≥30 conversions on the keyword), branded defence, retargeting-equivalent keywords, peak-event campaigns (Prime Day, Q4) where missing the high-intent click costs more than overpaying.
• Trade-off. Spend can run 1.5–2× your nominal daily target on a high-CVR day. ACOS variance increases. Requires statistically significant data to justify.

The +100% lever

The +100% upper bound is not a typo for +10% or +20%. A €0.50 bid under Up and Down can be submitted into the auction at €1.00 — and then a +50% top-of-search modifier stacks on top, taking the auction bid to €1.50. With both levers maxed, the effective bid can reach 3–4× the headline number.

This is why Up and Down is appropriate for branded defence (the +100% lands on already-high-CVR clicks) and dangerous for cold prospecting (the +100% lands on clicks that may not convert at all).

Choosing between Down Only and Up and Down

A defensible rule: promote a keyword from Down Only to Up and Down only after all three are true:

1. ≥30 conversions on the keyword in the last 60 days.
2. Keyword CVR is within ±20% of the campaign average (no anomaly).
3. The margin headroom in Target ACOS can absorb a 30–50% ACOS variance on the keyword.

If any of the three fails, stay on Down Only.

Dynamic Bidding and Amazon Marketing Stream

Dynamic Bidding makes its decisions on Amazon's internal CVR predictions, which are updated more frequently than the data in the standard reports. Amazon Marketing Stream gives advertisers hourly visibility into how Up and Down is actually shading bids — which is the difference between "I trust the algorithm" and "I can verify the algorithm."

Common mistakes

• Treating Up and Down as "free optimisation." The +100% lift is paid for in spend variance and ACOS variance. Free is the wrong mental model.
• Running Up and Down on broad-match keywords. Broad already has lower CVR variance; the algorithm's +100% lift amplifies the lower CVR rather than fixing it.
• Switching from Down Only to Up and Down during a launch. Launch keywords have unstable CVR data — the algorithm has nothing reliable to shade against. Stay Down Only for the first 30 days.
• Stacking aggressive placement modifiers on Up and Down without modelling the final bid. A +100% modifier under Up and Down can deliver an effective 4× bid in moments of high predicted conversion. Acceptable for branded defence; ruinous for unproven keywords.