Amazon has its own bidding controls built into Seller Central. Dynamic bidding adjusts bids in real time based on Amazon's prediction of conversion likelihood. Placement multipliers let you push bids higher at top-of-search positions. These are both called "bid adjustments" inside the console, and they are useful. But they are not what sellers usually mean when they talk about bid adjustments as a rule-based optimization strategy.
Rule-based bid adjustments work differently. Instead of Amazon making decisions based on real-time auction signals, a rule-based approach steps bids up or down on a defined schedule based on each target's own performance history. The controlling signal is trailing ROAS: what has this keyword or product target actually returned over a defined look-back window? If it is above your target, the bid can step up. If it is below, the bid steps down. You set the ceiling and the step size. Every move is logged. Every move is reversible.
Understanding how each piece of this works is what makes the difference between a system that self-corrects and one that compounds errors.
How the trailing ROAS window works
The look-back window determines how much data the system uses before making a bid decision for each target.
A longer window (14 to 21 days) smooths out noise. A keyword that had a bad Tuesday because of a competitor promotion will not get penalized if the surrounding two weeks were strong. The window absorbs short-term variance. The tradeoff is responsiveness: a genuine shift in performance takes longer to register and trigger an adjustment.
A shorter window (7 days) responds faster to real changes. If a product receives a wave of negative reviews and conversion rates drop, a 7-day window catches it sooner than a 14-day one. The tradeoff is noise sensitivity. Short windows react to random variation rather than true performance shifts, which can produce bid moves that do more harm than good.
The right window length depends on keyword volume. High-volume keywords generate enough data quickly that a shorter window is usable. Low-volume targets need more time to accumulate meaningful signal. For most accounts, 14 days is a reasonable starting point. If you are not sure what window to start with, the what to automate guide covers how to match automation settings to account volume and structure.
Step size and cooldown
Step size is how much the bid changes on each evaluation. A 5 percent step means a keyword bidding $1.00 becomes $1.05 on an up-move or $0.95 on a down-move. A 15 percent step is a more aggressive shift.
Small steps are self-correcting. If the first step up moves the bid into a range that performs worse, the next evaluation catches the decline and steps it back down. The system converges toward the efficient bid level through iteration rather than guessing in one move.
Large steps compound errors. A 20 percent jump up on a keyword that was already near its efficient ceiling can push ROAS below target on the next evaluation period. That triggers a correction step down, which may overshoot in the other direction. Large step sizes introduce oscillation that small ones avoid. Start with 5 to 10 percent and adjust if the account warrants it.
Cooldown prevents the system from chasing its own moves. After a bid changes, the rule waits a defined period before evaluating again. This gives the new bid time to accumulate enough data to assess fairly. Without cooldown, a system can re-evaluate before the previous change has produced measurable results and start moving again based on insufficient data. A 14-day cooldown pairs logically with a 14-day trailing window: the evaluation window and the cooldown period match, so every re-evaluation is based on data from after the previous adjustment landed.
The ceiling
Without a hard cap on how high a bid can go, automation can bid a keyword to break-even or below during a strong run. A keyword with consistently high ROAS gets stepped up repeatedly. At some point the bid is high enough that incremental spend on that keyword is marginal. Without a cap, the system keeps stepping.
The ceiling is a seller-controlled value, set per campaign or per target. It is not set-and-forget: revisit it when product margins change or when competitive dynamics shift enough that the previous ceiling is no longer accurate. But it is the safety valve. A rule that runs without one can work against you in a good week, then take time to step back down when performance normalizes. Set a ceiling before you activate bid adjustments.
When to use bid adjustments vs. performance rules
These serve different purposes and are designed to work together, not to replace each other.
Bid adjustments are continuous and keyword-level. They run on a rolling window for each individual target and make small, frequent moves to optimize each bid independently. The frame of reference is the target's own performance history.
Performance rules are campaign-level and react to the prior day's account data. A campaign that exceeded target ACoS on a daily roll-up triggers the rule. The action is typically a budget reduction or a pause, not a bid change. Performance rules are coarser by design: they respond to meaningful deviations from targets at the campaign level, not to granular per-keyword trends.
The complement is clear: bid adjustments tune individual keywords toward their efficient price over time. Performance rules protect the account from a bad day at the campaign level before it compounds into a bad week. The four rule types serve layered purposes across an account, and bid adjustments fit into that stack as the most granular level of ongoing optimization. The automation rules guide explains how these layers interact and what to configure first.
Undoability: what auditable means in practice
Every bid adjustment that fires is logged: which target, which direction, what the previous bid was, what the new bid is, and what performance data drove the decision. That log is the audit trail.
Auditable in practice means two things. First, if a bid move produces unexpectedly bad results, you can look at the log and understand exactly what happened and why. There is no inference required. Second, every move is reversible. One action restores the previous bid. If a data anomaly caused a bad evaluation, the damage is contained to a single step and correctable immediately.
This is a meaningful distinction from manual bid management. Manual changes are rarely logged in a way that makes cause-and-effect clear, and correcting them requires manual work proportional to how many campaigns were affected. A logged, reversible, per-target system is auditable in a way that ad-hoc management is not.
Before activating bid adjustments, a structured account audit is worth completing first. Bid adjustments work best when the underlying campaign structure is clean: naming conventions are consistent, match types are intentional, and there are no structural issues that would produce misleading ROAS signals at the keyword level.
If you are new to automation, Spend Alerts are included free on every plan and require no configuration. They are a useful starting point for understanding what automated monitoring looks like before layering in bid-level optimization.
Frequently asked questions
What is the difference between Amazon bid adjustments and dynamic bidding? Amazon dynamic bidding is Amazon's own real-time system that raises or lowers bids at auction time based on its prediction of conversion likelihood. It operates inside Amazon's own system and gives sellers no visibility into how individual bids are being modified. Rule-based bid adjustments are seller-configured rules that step bids up or down based on trailing performance data the seller can see and verify. Dynamic bidding is always on. Rule-based adjustments are optional, configurable, and every change is logged.
How often should Amazon bid adjustments run? The right frequency depends on the window length. A 14-day trailing window logically pairs with a 14-day cooldown between evaluations. At that cadence, bid adjustments run roughly twice a month per keyword, which gives each evaluation period enough data to assess. Shorter windows with shorter cooldowns can run more frequently but are more sensitive to noise and short-term variance.
What ceiling should I set for bid adjustments? Start by working backward from your target ACoS or ROAS and your product's conversion rate. If a keyword converts at 10 percent and your target ACoS is 25 percent on a $30 product, the efficient maximum bid is $0.75 (30 times 0.25 times 0.10). That is your economic ceiling. Set the cap slightly below that figure to leave margin for variance. Revisit the cap when product price or conversion rate changes significantly.
Off Hours Bid Adjustments steps keyword and target bids on their own trailing ROAS — small moves, a cooldown between them, a cap you control, every change undoable. Start a free 14-day trial.