Amazon Bid Optimization: How to Set the Right Bid (With the Math)

What Bid Optimization Actually Means

Bid optimization is the ongoing adjustment of what you pay per click on a keyword, product target, or placement, based on how that click actually performed — not how you guessed it would perform when you launched the campaign. It isn't a setting you configure once. It's a cycle: pull the data, check it against a target, move the bid, then measure what moved.

The mistake most sellers make is treating ACOS as the only input. ACOS tells you the ratio of spend to sales on clicks that already happened. It doesn't tell you whether tomorrow's bid is too high, too low, or exactly right — because tomorrow's auction and tomorrow's conversion rate haven't happened yet. Good bid optimization treats ACOS as a lagging signal and pairs it with a target set in advance from your margin.

The Bid Formula, With Real Numbers

The core formula: max bid = target ACOS × conversion rate × average sale price. If your product sells for $40, your target ACOS is 25%, and a keyword converts at 10%, your maximum defensible bid is 0.25 × 0.10 × $40 = $1.00.

Now compare that to what's actually happening. Say the keyword is bid at $1.30 and running a 32% ACOS. The formula says you're overpaying by roughly 30 cents a click. The instinct is to cut straight to $1.00. Don't. Cut in increments of 10–15% (to roughly $1.15), then hold long enough to trust the new number before cutting again. A single jump from $1.30 to $1.00 on a keyword doing real volume can crater impression share overnight, and you won't know afterward whether the drop in sales came from the bid or from something else that happened the same week.

The same math runs in reverse. A keyword converting at 15% with a $40 sale price and a 25% target ACOS supports a $1.50 bid — if it's currently bid at $0.90 and still winning impressions, it's underpriced and you're leaving impression share, and probably rank, on the table.

Reading the Signals Before You Touch a Bid

Before moving a bid, check what's actually driving the number. A rising ACOS can mean the bid is too high — or it can mean CTR dropped because a competitor changed their main image, CVR dropped because you went out of stock in one size, or the sale price moved and nobody touched the bid. The table below is the same triage sequence to run before changing anything.

When the Change Doesn't Work

Sometimes you cut the bid and ACOS doesn't move, or you raise it and sessions don't follow. Check three things before deciding bid optimization has failed:

• Not enough data. Under roughly 15–20 clicks in the window you're judging, the ACOS number is noise, not signal. Extend the window before drawing a conclusion.
• The bid wasn't the constraint. If impressions didn't move after a bid increase, you weren't losing the auction on bid — check for a budget cap, a suppressed listing, or a low relevance score on that keyword.
• Something else moved at the same time. Price changes, stock-outs, review count, and seasonality all move ACOS independently of anything done in campaign manager. If you didn't isolate the bid change, you can't credit — or blame — it.

This is also why a change needs a plan before it runs, not just after it. Every proposed change carries three things before it runs: the evidence behind it, a measurement plan, and a rollback trigger. Written down in advance, that third piece is what stops a bad optimization from sitting live for three weeks because nobody was watching.

The Mistakes That Actually Cost Money

The most expensive mistake is optimizing at the account level when margin varies by product. A 25% target ACOS might be profitable on a $60 item with high margin and ruinous on a $15 item with a thin one. One rule across a whole catalog is faster to set up and wrong for at least half of it.

The second is reacting to a single bad day. A one-day ACOS spike on low volume is usually noise. We've cut a top-of-search bid the same week a competitor ran a lightning deal, watched sessions fall, and initially credited the bid cut — when the real cause was a temporary shift in the auction itself. The fix wasn't a bigger cut; it was waiting a week and watching the placement recover on its own.

The third is cutting bids on a keyword that also drives organic rank. A keyword with a mediocre direct ACOS can still be worth an aggressive bid if it's the term customers use to find you organically. Judge it on total contribution, not the PPC line alone.

Where This Fits With Automation

Amazon's own tools — Dynamic Bidding, Bid+, rule-based automation — apply a version of this formula automatically, and they're a reasonable starting point for accounts without the time to run the cycle weekly. Third-party bidding tools go further, applying it per keyword at scale with more granular data than the console shows.

Where those tools stop is the judgment layer above the formula: knowing a click drop was a competitor's lightning deal and not your bid, or that a keyword is worth protecting for organic rank despite a weak direct ACOS. That's the part that still needs a person, or a system built to read the whole account rather than one campaign at a time. Dr. PPC — built by Full Circle, which has managed over $500M in revenue across 100+ brands — runs that fuller read on every account it manages, proposes each bid change with the evidence and rollback trigger attached, and includes Orbit, its analytics and tracking suite, at no extra cost on top of $300 a month plus 3% of spend, capped, month-to-month. Whether or not that structure suits you, the formula above works the same regardless — it's the judgment layer that's the actual product.