An Amazon auto campaign is a Sponsored Products campaign where Amazon picks the targets for you. Instead of naming keywords, you hand Amazon your product listing, and it matches your ad to the search terms and product pages it reads out of your title, bullets, and category. Every auto campaign targets through four groups, close match, loose match, substitutes, and complements, and each group takes its own bid. The real job of automatic targeting is discovery: surfacing the search terms and competitor ASINs you then graduate into manual exact campaigns, where the precise bidding happens. This guide covers what the four groups actually buy, how I bid them, the harvest loop that turns auto's findings into performance, and the auto-share ceiling that separates a healthy account from one paying Amazon to rediscover what it already knows.
What automatic targeting actually is
Automatic targeting means Amazon selects where your Sponsored Products ad shows, using your listing content as the targeting input: title, bullet points, description, and category. You control the budget and the bids; Amazon controls the matching. The campaign buys placements in two arenas, search results and product detail pages, and reports every match in the search term report.
That inversion of control is the whole character of the campaign type. In a manual campaign, your keyword list is your statement about what the product is. In an auto campaign, the listing itself is the statement, and Amazon's matching shows you how that statement was heard. I read the first weeks of a new auto campaign partly as targeting output and partly as a diagnostic: when the matches come back strange, the listing is telling Amazon the wrong story about the product, and no bid adjustment fixes a briefing problem.
The output lands in the search term report, and it arrives in two flavors. Search matches show up as the words shoppers typed. Detail-page matches show up as ASINs, the rows that read like B0 codes instead of words, which is how you find out which competitor shelves your ad has been standing on. Every row of that report is the product an auto campaign exists to manufacture: evidence about where your product sells that you did not have to guess in advance.
The four auto targeting groups
Every auto campaign splits its targeting into four groups. Close match and loose match place ads on search terms closely or loosely related to your listing. Substitutes and complements place ads on product detail pages: products similar to yours, and products bought alongside yours. Each group takes its own bid, which is the control most sellers never touch.
| Targeting group | Where the ad shows | What I typically see |
|---|---|---|
| Close match | Search results for terms tightly related to your listing | The tightest relevance of the four, and usually the strongest conversion rate |
| Loose match | Search results for broader, loosely related terms | Wilder matches at cheaper clicks; a steady source of terms nobody would have guessed |
| Substitutes | Detail pages of products similar to yours | Conquest traffic and a rich vein of competitor ASINs, worth watching closely |
| Complements | Detail pages of products bought alongside yours | The most mixed results of the four; the cross-sell logic only fits some catalogs |
The deeper split is search versus shelf. Close and loose match compete in search results, so they behave like keywords you never had to write. Substitutes and complements buy space on other products' detail pages, so they behave like product targeting: the shopper is further along, already looking at a specific alternative, and your ad is an interruption on someone else's listing. Those are different economics, and lumping the four together under one blended read hides which half is earning.
Behavior follows that split. Close match converts best for the obvious reason: the shopper typed something close to what the listing says the product is. The two richest discovery veins are usually loose match and substitutes, precisely because they wander furthest from anything you would have thought to target yourself. Complements is the group I watch closest, because its logic only holds when the adjacency is real. A lens ad on a camera body's page makes sense; the same logic applied to a loosely related kitchen gadget is a guess wearing a strategy's clothes.
Auto's real job: discovery, not training wheels
The auto campaign is campaign number one in my per-product stack, and it never retires. Its job is to keep feeding new search terms and competitor ASINs into the manual structure, where exact match runs the proven ones at precise bids. Auto and manual are not rivals, and they are not life stages. They are two ends of one conveyor.
The training-wheels myth goes like this: launch on auto, copy the first winners into manual campaigns, then switch auto off as a sign of maturity. What actually gets switched off is the account's supply of new information. Shopper search language moves constantly: competitors launch and fold, phrasing drifts, seasonal qualifiers appear and vanish. An account with no discovery running is optimizing a photograph of a market that kept walking.
In the campaign structure I build per product, auto is the first of four or five campaigns, each with its own budget and its own job: auto discovers, a broad and phrase research campaign extends discovery into the long tail, exact match runs the proven terms, product targeting works chosen ASINs, and brand defense stays walled off. Keeping discovery money and performance money in separate campaigns is what lets you throttle research in a tight month without touching a single proven keyword. The auto campaign runs for the life of the product; what changes is how much of the account's spend it deserves, which is the ceiling covered below.
How to bid an auto campaign
Start with one auto campaign per product and four separate bids, one per targeting group. Price them from your own economics: revenue per click times target ACOS is the fair price for an average click, and unproven traffic starts below it. Split a group into its own campaign only when it needs its own budget or placement settings.
The arithmetic first. On the dataset I grade across these guides, $111,058 in ad sales from 45,672 clicks makes revenue per click $2.43, so a 30% target ACOS prices the average click at $0.73 against an actual average CPC of $0.79. An auto campaign has not earned the average: its whole population is unproven. So each targeting group starts under that fair price, and raises get argued for by the group's own rows, not by Amazon's suggestion box, which knows your budget and nothing about your margins.
Dynamic bidding compounds whatever you set, so I run discovery on dynamic bids, down only. Up-and-down lets Amazon raise a bid by up to 100% at top of search when it predicts a conversion, and on proven exact terms that can be a reasonable trade. On a campaign defined by uncertainty, it is a premium charged on guesses; down only lets Amazon cheapen the doubtful click but never inflate it. The compounding math, and what it does at each placement, is walked through in my Top of Search guide.
Mechanically, the console shows the four groups inside the auto ad group, each with an editable bid. In the bulk file, they arrive as targeting rows under the auto ad group with their own Bid cells, which turns a per-group bid pass across the whole account into a filter and an Update instead of a campaign-by-campaign crawl.
When a targeting group earns its own campaign
The default is one auto campaign with four bids, because the controls a group usually needs, its bid and its negatives, already live at that level. A group graduates to its own campaign only when it needs a control that lives at the campaign level: its own budget cap, its own placement strategy, or its own bidding strategy.
Substitutes and complements are the usual candidates. When detail-page traffic starts drinking a disproportionate share of the campaign's budget at a conversion rate the search groups would never tolerate, a bid cut is the first answer; a separate campaign with a hard budget cap is the second. What I do not do is split all four groups into four campaigns by default. That quadruples the budgets to manage and quarters the data each campaign accumulates, and most accounts get nothing back for the complexity.
The harvest-then-negate loop
An auto campaign's output is raw search term data, and the loop that processes it has two motions. Promote a proven term into manual exact at a bid of revenue per click times target ACOS. Then add the same term as a negative exact in the auto campaign, so discovery stops re-buying what performance now owns.
- Launch one auto campaign per product. All four targeting groups enabled, each with its own bid, on a moderate discovery budget. It runs for the life of the product.
- Price starting bids below the fair average click. Revenue per click times target ACOS prices the average; unproven traffic starts under it, on down-only bidding.
- Give every search term a fair trial. One divided by your conversion rate is the clicks an average order takes: roughly eleven clicks at 9.1%, about $8.77 of spend per verdict on this account's numbers. Judging earlier punishes terms that never got their chance.
- Harvest what proves itself into manual exact. My bar: two or more orders, sales on the advertised product itself rather than a sibling, and a relevancy check against the live search results. The term enters exact at revenue per click times target ACOS.
- Negate the harvested term in the auto campaign. Negative exact, at the source, the same day. This is the half of the loop that most accounts skip.
- Watch auto share against the phase ceiling. A loop that is working shows up as a falling auto share; a share that will not fall is the loop telling you it is broken.
Cadence matters less than consistency: monthly is enough for a stable account, tighter when discovery is running hot, as my search term report guide covers. The loop also runs in reverse for losers: terms that spend well past their fair trial with nothing to show become negatives, on the click thresholds from my negative keywords guide. And because auto search terms fail in patterns more often than one by one, an n-gram analysis of the file shows the losing words hiding across hundreds of rows; the free N-Gram Analyzer runs that read in the browser, and the Negative Keyword Finder turns the verdicts into an upload-ready list.
The auto-share ceiling: one number that grades the whole loop
Auto campaigns should take 40% or less of ad spend in the awareness phase, 25% in market share, 20% in growth, and 15% or less in profitability. The ceiling tightens because a maturing account should already own its converting terms at exact match. An auto share stuck high is the signature of a broken harvest loop.
| Phase | Auto campaign % of spend | What the ceiling assumes |
|---|---|---|
| Awareness | ≤ 40% | The account is new; discovery deserves its largest share |
| Market Share | ≤ 25% | Early winners exist and should be carrying spend at exact match |
| Growth | ≤ 20% | The exact structure is now the engine; auto prospects at the margin |
| Profitability | ≤ 15% | The account owns its converting terms; auto is pure R&D |
Put dollars on it. On the dataset I grade across these guides, $36,303 in monthly spend, the awareness ceiling allows $14,521 of auto spend and the profitability ceiling allows $5,445. The roughly $9,076 between those columns is what the harvest loop is supposed to move: not money that stops being spent, but money that migrates into exact match campaigns where every term carries its own precise bid. An account still running 40% through auto years after launch is paying discovery prices, month after month, for traffic it proved long ago.
This is why auto-campaign share is one of the nine metrics the free Account Health Snapshot grades against your phase, alongside ACOS, wasted spend, and the rest of the benchmark table. The free Audit Dashboard computes it from the same bulk file, and the demo account shows what the read looks like on real anonymized rows.
The five auto campaign mistakes I keep finding
Nearly every broken auto setup I audit fits one of five patterns: discovery with no harvest, harvest with no negation, detail-page groups left unwatched, auto running as the only campaign type, or auto switched off once manual exists. The first two pay Amazon twice, the third leaks quietly, and the last two each amputate half the funnel.
- Running auto with no harvest. The campaign finds winners and nobody moves them, so the account keeps buying its best terms at discovery bids under Amazon's control. This is the stuck-high auto share in its natural habitat.
- Harvesting with no negation. The term gets promoted, the negative exact never gets added, and now two campaigns buy the same search and split its conversion data. That is search term bleed, self-inflicted.
- Letting substitutes and complements run unmonitored. Detail-page traffic does not announce itself; it accumulates as ASIN rows most sellers never read. A pass through the Wasted Spend Finder shows what those placements produced, which for complements especially is sometimes nothing at all.
- Treating auto as the only campaign type. With no exact campaigns, every proven term keeps being bought at whatever the group bid happens to be, in whatever placement Amazon chooses. Discovery without performance is a research department with no factory.
- Killing auto once manual exists. The account stops learning while the market keeps moving. Six months later, the search term report reads like a reunion: the same terms, older, with nothing new in the room.
Frequently asked questions
Should I turn off my Amazon auto campaign once manual campaigns are running?
No. The auto campaign runs for the life of the product, because its job never finishes: shopper search language shifts, new competitors list, and seasons change what people type. Manual exact holds the terms you have already proven; auto keeps finding the next ones. What should change over time is auto's share of spend, from 40% or less at launch down to 15% or less in a mature account, as winners keep graduating.
What are close match, loose match, substitutes, and complements?
They are the four targeting groups inside every Amazon auto campaign, and each takes its own bid. Close match and loose match place your ad on search terms tightly or loosely related to your listing. Substitutes and complements place it on product detail pages: substitutes on products similar to yours, complements on products bought alongside yours. Two groups buy searches, two buy shelf space on other products' pages.
What is a good starting bid for an Amazon auto campaign?
Price it from your own economics, not Amazon's suggestion. Revenue per click times target ACOS gives the fair price for an average click; on the dataset I use across these guides, $2.43 times a 30% target prices it at $0.73. Auto traffic is unproven, so I start each targeting group below that fair average and let performance argue bids up. The four groups rarely deserve the same number for long.
How much of my budget should auto campaigns take?
At or under 40% of ad spend in the awareness phase, 25% in market share, 20% in growth, and 15% or less in profitability. The ceiling tightens because a maturing account should own its converting terms at exact match, where bids are precise. An auto share stuck above the bar for your phase usually means proven terms are not graduating, and the free Account Health Snapshot grades this exact metric from your bulk file.
Do Amazon auto campaigns stop working over time?
They change jobs rather than stop working. As you harvest winners into exact match and negate them at the source, the auto campaign is left holding mostly unproven traffic, so its ACOS will usually look worse than your exact campaigns. That is the cost of exploration, not decay. I judge a mature auto campaign on what it discovers and what its share of spend is, not on its ACOS alone.
Auto vs manual targeting on Amazon: which is better?
Neither, because they are not competing for the same job. Auto targeting is the widest net Amazon offers, built to discover search terms and ASINs you would not have guessed. Manual exact is built to run proven terms at precisely controlled bids. An account running only auto has handed Amazon the wheel; an account running only manual has stopped learning. The structure that works runs both, permanently, connected by the harvest loop.
Find out what your auto share is telling you
Your auto-campaign share is already computable from a bulk file you can download in five minutes. Drop that file into the free Account Health Snapshot and it comes back as one of nine graded metrics, scored against the phase ceilings in this guide: about 60 seconds, parsed in your browser, no email, no account. If the share comes back stuck above your phase's bar, the question becomes which half of the loop is broken, the harvesting or the negating, and that diagnosis is exactly what the free 30-minute diagnosis call is for, with your search term data on the screen.