Amazon PPC Strategies: What Actually Moves ACoS (With the Math)

The Core Loop: Harvest, Promote, Negate

Every profitable Amazon PPC account runs on the same loop, whether it's a $2,000/month account or a $200,000/month one: automatic campaigns find the search terms, manual campaigns take control of the ones that convert, and negative keywords stop you paying for the ones that don't. Skip any one of the three and you're paying for someone else's clicks.

Here's what it looks like on real numbers. Say you launch an automatic campaign for a cast-iron skillet and let it run for two weeks. It spends $420 and returns 14 orders. Pull the search term report and you'll usually find something like this: one term — "12 inch cast iron skillet" — accounts for $180 of that spend and 6 of those orders, a 33% ACoS. The other $240 is spread across forty low-volume terms that never converted.

The move: promote "12 inch cast iron skillet" into a manual exact-match campaign where you set the bid directly, and add it as a negative exact inside the automatic campaign so the two stop competing for the same auction. Everything that spent without converting becomes a negative keyword, full stop — not "watch it for another week."

That's the whole strategy, run on a schedule. Most sellers do it once, in month one, and then stop. The accounts that stay profitable are the ones where someone — or something — pulls the search term report every week for the life of the product, not just at launch.

Setting Bids by Keyword Intent, Not by Gut

Once a term is in manual, the bid isn't a guess — it's a formula: max CPC = target ACoS × average order value × conversion rate. If your target ACoS is 30%, your average order value is $90, and that specific term converts at 15%, your ceiling is $4.05. Bid above that and the term is structurally unprofitable no matter how well it performs.

Not every keyword deserves the same ceiling, because not every keyword does the same job. Branded and near-branded terms — your own product name, misspellings of it — convert high and should get a bid near the top of what the formula allows, because that traffic is close to a sale already. Generic head terms ("skillet," "cookware") convert lower and cost more per click; they're a discovery tool, not a profit center, and the formula will usually tell you to bid low or skip them until the listing itself converts well on its own. Competitor ASIN targeting sits in its own bucket — it's defensive or offensive share-grabbing, and it should run on its own budget cap so a bad week there doesn't eat the budget your best-converting terms need.

ACoS Tells You About the Campaign. TACoS Tells You About the Business.

ACoS is ad spend divided by ad revenue — it only ever looks at the dollars a campaign can see. TACoS (total advertising cost of sale) is ad spend divided by total revenue, organic included. A campaign can carry a 35% ACoS and still be doing its job, if the keyword it's ranking for is also lifting organic placement and organic orders on that same term.

This is where sellers cut budget on the wrong thing. A term with a high ACoS that's also driving organic rank movement is not a losing term — check the organic rank and organic order count for that keyword before you kill the ad. If organic movement isn't happening, the ACoS is just telling you the truth and the term should come down.

Track both numbers monthly at the portfolio level. If TACoS is falling while ACoS holds flat or rises, PPC is doing its job — buying rank that keeps paying after the ad is turned off. If both are rising together, the account isn't converting the way it used to, and no amount of bid tuning fixes that on its own; that's a listing or price problem, not a PPC problem.

The Mistakes That Actually Kill an Account (Including Ones We've Made)

The one we've made ourselves: promoting a search term into exact match after a single order. One order tells you almost nothing — you need enough clicks to trust the conversion rate the formula depends on, usually north of fifteen to twenty for a low-volume term. Move too early and the bid you set is based on noise, not signal.

• One target ACoS for the whole catalog. A 40% ACoS is fine on a high-margin SKU and ruinous on a low-margin one.
• Turning a campaign off after one bad week instead of just cutting the bid.
• Treating negative keywords as permanent when seasonality can bring a dead term back to life six months later.

The discipline that actually prevents most of this: every change should carry three things before it goes live — the evidence for making it, a way to measure whether it worked, and a trigger for reversing it if it didn't. Most sellers have the first and skip the other two, which is exactly how a good idea — kill the underperforming term — turns into a bad outcome: a term that was about to turn the corner, killed too early.

When the Fix Doesn't Work

Sometimes you do all of this correctly and ACoS still won't move. Before touching bids again, check three things outside the ad account: is the listing converting on organic traffic at a normal rate for the category (if organic CVR is also low, the ads aren't the problem, the listing is); has a competitor dropped price or launched a coupon in the last two weeks (check the buy box, not just your own price); and is the product actually in stock at the warehouse the ad is serving from (an ACoS spike on a listing that's low on inventory is Amazon spending your budget to sell units you don't have).

If all three check out and the number is still bad, the honest answer is that the bid ceiling the formula gives you is lower than what it costs to win the auction right now — the term isn't affordable at your current margin this week. Pause it, don't fight it, and revisit when margin or the competitive landscape changes.

Where Automation Fits

Everything above is manageable by hand on a handful of SKUs with a weekly hour of spreadsheet work. It gets harder to sustain past a few hundred keywords across a growing catalog, which is the actual argument for automating the loop rather than a claim that automation is required to do it at all.

Dr. PPC — built by Full Circle, an agency that has managed more than $500M in revenue across 100+ brands — reads the account, writes the harvest-promote-negate loop and the bid math above per product against that brand's real margins, and attaches the evidence, measurement plan, and rollback trigger to every change before it runs. The client sets how much runs automatically versus waits for a click; inventory risk, pricing, and new launches always go to a person regardless of setting. It costs $300 a month plus 3% of ad spend, capped, month-to-month, with the first 30 days free, and includes Orbit — the analytics, trackers, and Dr PPC Console — at no extra charge. Whether or not that's the right purchase for a given account, the loop itself doesn't change: harvest, promote, negate, on a schedule, forever.