Every Amazon ad menu looks like a wishlist: Sponsored Products, Sponsored Brands, Sponsored Display, video, DSP, live streaming. Most brands try three of them at once, on a budget that can't feed one.
The right set of ad types depends on how much you are spending, and each addition belongs on the budget that can support it. Here is the ladder we use, and the account-structure habits that matter more than which rung you're on.
What makes a good Amazon PPC strategy?
A good Amazon PPC strategy decides which ad types your budget can support, gives each campaign one job, and caps what a click may cost against your margin. Most of what goes wrong in an account comes from skipping one of those three decisions.
Amazon's sponsored ads are sold per click in an auction. You set the most you'll pay, higher bids are more likely to win the placement, and Amazon says there's no minimum spend. Nothing in the platform stops you from opening every format on day one, so the limit has to come from you.
A note on names before the ladder, because Amazon has changed them. Its ad platform is now called Amazon Ads Agent. Amazon says Sponsored Display is now known as display ads, and that Amazon DSP's programmatic buying now sits inside Display, Video and Audio campaigns, which it shortens to DVA+. Existing campaigns keep running. We've kept the older words where the people reading this will still search for them.
First: a budget too small to learn from
There's a floor below which ads can't teach you anything. A listing that spends a couple of hundred dollars a month and records a handful of attributed sales has produced numbers that are mostly noise. One order more or less swings the ACoS by tens of points. We see this often on small accounts, and it's easy to conclude the ads aren't working when the honest conclusion is that there's too little data to read.
Take illustrative numbers. At $1.25 a click, $200 buys 160 clicks. Spread across twenty keywords, that's eight clicks each, and Amazon's own targeting guide says to wait for at least 20 clicks before judging a keyword. The same $200 on two or three tightly chosen terms can tell you something.
At that spend, the job is to fix the listing before buying traffic to it. Listing, then ads is a sequence for a reason: ads send shoppers to a page, and a page that doesn't convert wastes every click. Run a small number of tightly chosen campaigns to confirm the listing converts, and save the rest.
How much should you spend on Amazon PPC?
Set the monthly budget by multiplying monthly sales by your target TACoS, then use the table at the end of this piece to see which ad types that budget can carry.
An illustrative brand selling $40,000 a month with a 12% TACoS target has $4,800 to spend, or about $160 a day. That puts it in the mid four figures, so it can run search ads and start retargeting, and it should leave DSP alone for now. The target comes from the job: our TACoS piece sets out single digits for protecting an established catalog and 15% to 25% while winning new customers, with launches running above both for a window you decide in advance.
Our typical starting ad budget for a brand is in the low four figures a month, built up rather than blasted. A new account starts with enough to gather data on a few campaigns and grows as the numbers clear break-even. The pace is set by what the data supports, and we don't let a month-three hope set it.
Amazon's guide suggests starting at $10 a day, which is fine for testing that a campaign serves and too small to fund this ladder. Two budget mechanics are worth knowing. Budgets sit at the campaign level, and Amazon says monthly spend can reach 30 times the daily budget, so a $100 daily budget can bill up to $3,000 in a calendar month. And when you divide the money, keep tests small and on their own: Amazon's advice in its Ad Intel series is to hold experimentation to no more than 10% to 15% of budget.
Stage 1: Sponsored Products and Sponsored Brands
These two are the minimum for any serious program.
Sponsored Products place a single listing in search results and on product pages. They're where most of the learning happens, because they're tied to keywords and ASINs and report cleanly.
Sponsored Brands put your logo, a headline and several products at the top of search results. They carry the brand, and Amazon's own guide says to judge them on new-to-brand metrics and not to treat a high ROAS as the main measure of success. For sellers, they require enrollment in Brand Registry, so sort that out first.
At this stage, build campaigns in this order:
- Branded defense. Your brand name and product names, exact match. It protects your own searches from competitors bidding on you, and it should show a high ROAS.
- Exact match on the terms you rank or want to rank for. These are your proven and your target keywords.
- Automatic and phrase match to discover variants, then harvest the converting search terms back into exact.
- Product targeting against competitor ASINs and your own, once the listing is converting.
Stage 2: display ads (Sponsored Display)
Display ads show your product on and off Amazon, to audiences that include shoppers who have viewed your products. Amazon calls this views remarketing, and Amazon's guide lists Brand Registry enrollment for sellers here too. The best early use is retargeting, bringing back people who looked and didn't buy. That is a cheaper audience than a cold search, and it makes sense once there's enough traffic to retarget.
Adding display too early splits a small budget across more campaigns than it can fund. Wait until Stage 1 is stable and above break-even.
Stage 3: video and everything creative-heavy
Video ads and richer Sponsored Brands formats ask more of the creative side. They help with products that need explaining and with categories where shoppers compare on look and story. Amazon's Sponsored Brands video runs 6 to 45 seconds, autoplays in shopping results and has its own placement and auction. The test before adding them is whether you have a video worth watching. A weak video costs the same as a good one.
Stage 4: DSP
DSP buying, which Amazon now places inside Display, Video and Audio campaigns, reaches display and video placements across Amazon and its network. We don't discuss it until the pay-per-click program is running at around five figures a month in spend. Below that, DSP competes for budget the search ads need, and the audiences it reaches are expensive to measure. At scale, it fills the top of the funnel and reaches shoppers the search programs can't.
Amazon's FAQ says managed-service options typically require a minimum spend of $50,000 in the US. Our threshold comes from what the search program needs, and it applies to self-service too.
The mistake is treating DSP as the next step after a bad month. It's a scaling tool for an account that already makes money.
How many campaigns does a food brand need?
A food brand needs one campaign for each job on each product group, and that count should stay small enough for the budget to fund every campaign through the day.
The jobs we separate are branded defense, non-branded exact, discovery, competitor terms, competitor product pages and defense of your own product pages. Take an illustrative brand with three product lines: granola, bars and a mix. Branded defense can be one campaign. Non-branded exact, discovery and product targeting for each line make nine more. That's ten campaigns, and at $1,500 a month each one gets about $5 a day, or four clicks at $1.25. So on that budget you'd start with one or two lines, or fewer jobs per line, and add the rest as the money comes in.
Amazon's best-practice guide says to keep ASINs from the same category together and to group closely related products, because keywords and bids apply to everything in an ad group. In food that usually means flavors and pack sizes of one product can share an ad group, and a granola shouldn't share one with a baking mix, since the searches differ.
Name campaigns so anyone can read them without opening them, for example SP | Granola | Non-brand | Exact. It looks fussy until you inherit an account, or need to total spend by job in a spreadsheet.
Structure rules that matter at every stage
Separate match types into their own campaigns
Budget is set at the campaign level. If exact, phrase and broad match share a campaign, a loose broad term can eat the budget your exact terms need, and you can't fund the good one without funding the bad. Give each match type its own campaign, and a clear name to go with it.
Negate broad match early and often
Broad match finds queries you didn't plan for: combinations, misspellings and phrasing no one person would think of. That's its purpose and its risk. On audits we often find that much of the non-branded spend is losing money, with broad match running without negative keywords as the largest leak. Review search terms weekly, and add the terms that spend without selling as negatives.
Split brand from non-brand in both naming and reporting
Branded campaigns show a ROAS that has little to do with how the rest of the account is doing. Name them so they can be read apart, and report them apart. Why ROAS is the wrong scoreboard goes deeper, and our piece on choosing an agency covers the question to ask about that split.
Remember that campaigns have memory
An ad campaign builds relevance history over time: which terms it has converted on, what the bids have done. When a campaign is copied and the original is paused or deleted, that history doesn't move with it, and the duplicate rarely performs the way the original did. We've found accounts carrying several versions of the same sponsored product campaign, launched and relaunched at inconsistent bids. If you want to test a change, create a new campaign alongside the old one, leave the old one running and compare. If you need to change a campaign's structure, change it in place where you can.
Accounts with a lot of near-duplicates are a sign of churn, and churn is the thing to ask about when you inherit an account.
How do you set an Amazon PPC bid?
Your break-even bid is what a unit keeps before ads multiplied by your conversion rate, and that number is the ceiling for a campaign that has to pay for itself.
Take an illustrative $30 SKU with $8 of landed COGS and $10 of Amazon fees. Pull your own fee numbers rather than borrowing these. The unit keeps $12 before advertising, a 40% pre-ad gross margin, so break-even ACoS is 40% and break-even ROAS is 2.5. If the listing converts 10% of clicks, you get one order per ten clicks, and ten clicks can cost $12 in total. That's $1.20 a click to break even. Bid less if the campaign is meant to leave profit behind: at a 30% ACoS target, the ceiling on the same SKU is $0.90.
For a sense of the market, Ad Badger reports an average cost per click of $1.22 and an average ACoS of 30.67% from its own app data for January through August 2026. That's one software vendor's user base across all categories, and your category and terms will differ. Hold it against your ceiling, and if the going rate on your terms sits above what your SKU can pay, look at the SKU before you look at the bid.
This is where food brands get caught. An illustrative $8 single unit with a 30% margin keeps $2.40, and at the same 10% conversion rate its break-even bid is $0.24. On a contested term that bid is unlikely to win much. No campaign setting fixes that SKU, and it's usually the argument for a different pack on Amazon. Repeat purchases can change the sum, and we come to that below.
Amazon's targeting guide suggests setting the highest bid on exact match, lower on phrase and lowest on broad. Beyond the bid itself, you can choose fixed bids or let Amazon move them with dynamic bids, down only or up and down, and you can add placement adjustments of up to 900%. One use we recommend is on branded terms, where a top-of-search modifier can hold the first position without raising bids across the account. Any automated setting should still answer to the ceiling you worked out above.
When should a search term become a negative keyword?
Make a term negative once it has had enough clicks that zero orders can't be blamed on chance: roughly three divided by your conversion rate.
Amazon's guide says to evaluate a term after at least 20 clicks. That's a sensible minimum, and the conversion rate decides how far it gets you. At a true 10% conversion rate, a term has about a 12% chance of going 20 clicks with no order, so one in eight good terms would look bad by luck. At 30 clicks that falls to about 4%. At a 5% conversion rate, 20 clicks with no order happens about 36% of the time, and you need around 60 clicks to get under 5%.
So at 10% conversion, the rule is 30 clicks. At $1.25 a click that's $37.50 spent on a term, more than three times the $12 the SKU keeps per order, so don't wait for the full count on a term that's plainly wrong for the product. "Sugar free" searches on a product with added sugar can go on day one. Keep the click count for terms that are plausible, and when a plausible term has converted but too expensively, cut the bid before you negate it, since it may still be feeding rank.
Amazon recommends phrase or exact negatives, and suggests leaving them for two weeks or more before changing strategy. A wrong call costs little, because negatives can be reversed.
How do you move a search term from discovery to exact match?
When a term in a discovery campaign has earned orders at an ACoS under your break-even, add it as a negative in that campaign and move it into an exact-match campaign with its own daily budget.
Automatic campaigns are the cheapest way we know to do this research, because they surface converting terms that keyword tools miss, and testing on exact match from the start is an inefficient way to find them. Amazon's automatic targeting runs four groups, close match, loose match, substitutes and complements, and you can bid on each separately.
The negative is the step people skip. Without it, the discovery campaign and the exact campaign both keep bidding on the term, and you lose control of which one pays for it. Amazon's Ad Intel series makes the same recommendation on segmenting match types and moving proven terms into tighter campaigns. Once the term is moved, it has a budget you set on purpose, and discovery goes back to looking for the next one.
Set your bar before you open the report, in orders and in ACoS. One lucky sale shouldn't promote a term.
Should you bid on head terms like "granola"?
Bid on head terms only after your winnable long-tail terms are funded and converting, and give them their own campaign so one term can't take the whole budget.
Head terms carry the most traffic and the most competition, they cover several sub-niches, and what you have to bid on them moves through the year. We regularly find them running at an ACoS several times break-even on accounts that have margin to spare. The instinct to keep them comes from wanting to be seen on the big search. Our advice is usually to pull those bids down and put the money on the longer terms the brand can win, using the data the account already has.
There's a case for head terms too, and it's rank. In our experience Amazon won't rank you for a term you don't sell on, and an ad sale on a term also helps your organic rank for it. So a head term can be a planned investment. Treat it as one, in a separate campaign with its own budget and a date when you'll review it, and keep it out of the ACoS you report for the rest of the account.
Competitor product pages deserve a look before you spend more on head terms. We've seen product targeting be one of the smallest lines in an account and one of its better converters, because the shopper is already looking at something like yours.
How often should you change Amazon PPC campaigns?
Review search terms weekly and bids as the data supports it, and leave campaign structure alone unless you have a reason you can write down.
Amazon suggests checking metrics at least twice a week in the first week of a campaign. After that, this is the rhythm we'd suggest for a brand running its own ads:
- Weekly: search terms, new negatives, terms ready to move to exact, and any campaign running out of budget before the day ends.
- Every two to four weeks: bids against your ceiling, and placement performance, once there are enough clicks to read.
- Monthly: spend by job, new-to-brand orders and TACoS per ASIN.
- Quarterly, or at a real change in the business: whether you've earned the next rung.
Campaigns need time to run, particularly when you're migrating keywords between them, and frequent relaunching is what produces the near-duplicates described above. Check the Buy Box before you read a bad week as an ad problem. A reseller holding the Buy Box makes your ad reports undercount what the campaigns did, and bids cut on that evidence can take out keywords that were working.
Match the stage to the margin
Whatever the stage, the ad has to clear the break-even for the SKU it sells. A staged plan on a 15% margin product still loses money at every stage. If the pre-ad gross margin is under 20%, fix the margin before adding rungs, using the levers in the pack size piece and the 30% floor.
Can an Amazon ad campaign run below break-even?
Yes, when you chose it, you've capped it, and it's buying new customers you can show reorder.
Everything above treats break-even as the ceiling, and for most campaigns it is. The exception is a product people buy again. If the first order loses a little and the second and third make it back, a campaign below break-even ROAS can be the right call. You need your repeat rate to prove that, and new-to-brand reporting to show the campaign is reaching first-time buyers, though as we read it that reporting doesn't cover Sponsored Products. The ROAS piece covers that calculation.
What we won't accept is a loss with no plan attached. Our position with brands is that even allowing for experiments and for campaigns that are still maturing, a brand with healthy margin shouldn't be running the channel at a loss. Josh Grillo, who runs brand management here, puts the other half of it well: an account may need more spend, and in the same breath you can point at the worst ad groups and say not there.
A quick self-check
| Monthly ad spend | What to run | What to hold off on |
|---|---|---|
| Low hundreds | A few exact-match Sponsored Products and branded defense | Everything else; fix the listing |
| Low four figures | Sponsored Products and Sponsored Brands, match types separated | Display, video, DSP |
| Mid four figures | Add Sponsored Display retargeting | DSP |
| Around five figures and above | Video, richer creative and a DSP test | Nothing, if each tier is above break-even |
These are guides for a food brand with healthy margins, not rules. A larger catalog can justify more campaigns sooner, and a thin margin argues for fewer.
The bottom line
Ads work in stages: fix the listing, run search ads on exact terms, add retargeting, then creative, then DSP once the account can carry it. At every stage, keep match types apart, negate early, report brand and non-brand separately and leave your campaigns alone long enough to learn from them.
Work out the break-even bid on your top three SKUs this week, because that one number tells you which campaigns can pay for themselves. If you'd like a second read, send us your ASINs and we'll look at your spend by job, your search terms and which rung your budget supports, and tell you what we see.






