Auto vs. Manual Amazon Campaigns: The Harvesting Loop That Scales Sponsored Products

CH.01How auto and manual targeting differ, and why it is not a binary

Automatic targeting does not use keywords you choose. When you create an auto ad group, Amazon creates exactly four targeting groups for you, each with its own settable bid. Two are query-based: close match and loose match. Two are product-based: substitutes and complements. You never build these manually; they appear with the ad group and you simply tune their bids.

Manual targeting is the opposite. You choose every target. Keyword targeting offers three match types of escalating precision, and product targeting lets you go after specific ASINs or whole categories refined by brand, price range, ratings, and Prime eligibility.

The targeting surfaces, side by side

Read that table as a flow, not a fork. Auto discovers what converts. Manual exact locks in the winners at a controlled bid. Product targeting captures the ASINs the substitutes and complements themes surfaced. The choice between auto and manual was never an either-or. It is a loop where each side feeds the other.

CH.02Stand up the discovery engine: launch the automatic campaign correctly

Build the auto campaign first so it begins gathering signal while you set up the rest. The most common mistake here is leaving a single blended default bid across all four themes. If you set no per-group bid, Amazon applies the ad group's default bid to close match, loose match, substitutes, and complements alike, which buries the differences you most need to read.

• Set a distinct bid per targeting group rather than one default, so each theme's economics stay legible.
• Start with a conservative bidding strategy, typically dynamic bids down only, so a hot launch query cannot overspend before you have data.
• Give the campaign an always-on configuration with no end date, in line with Amazon's published best practice.
• Plan to monitor metrics at least twice a week during the first week, then optimize within the first 30 days.

CH.03Read the search-term report and define your graduation rule

The automatic search-term report is where discovery turns into decisions. It shows the actual customer queries the four themes matched your product to, with their clicks, orders, and ACOS. Your job is to set one evidence-based rule that separates proven winners from noise, then apply it the same way every week.

A graduation rule answers a single question: at what point has a search term earned a controlled manual bid? Anchor it to conversions at or below your target ACOS over a defined window rather than to a single lucky order.

1. Pull the auto search-term reportFilter to the automatic campaign for a rolling window, typically the last 30 days, so each term has had time to accumulate clicks.
2. Apply a conversion-and-ACOS thresholdFlag terms that converted at or below your target ACOS across the window. Common practitioner heuristics require something like two or more orders before promoting, but treat any specific number as your own house rule, not Amazon guidance.
3. Separate the dead weightNote terms with many clicks and no orders. These are candidates to negate, not to promote. Some practitioners use a rule of thumb such as 10 to 15 clicks with no conversion, again as a heuristic rather than an Amazon rule.

CH.04Harvest winners into manual exact-match ad groups

Once a term clears your graduation rule, move it into a manual ad group where you can bid it precisely. Exact match is the default destination because it is the most restrictive match type, matching the query word for word, which gives you the tightest control over spend on a term you already know converts.

• Add the graduated term as an exact-match keyword in the manual campaign at a higher, controlled bid based on its observed performance in the auto report.
• Use phrase match selectively for terms where you want to capture close variations in the same word order, since phrase is more restrictive than broad but wider than exact.
• Use product targeting to capture the specific ASINs that the substitutes and complements themes surfaced, targeting those detail pages directly.

This is where the loop starts paying back. Spend on these proven terms is now governed by a deliberate manual bid instead of Amazon's automatic matching, which is exactly the control the old "graduate into manual" advice was reaching for. The difference is that you are not abandoning auto to get it.

CH.05Close the loop: negate harvested terms in the auto campaign

Harvesting a term into exact is only half the move. If you leave that same term active in the auto campaign, the two campaigns compete against each other for the same query, and you lose the controlled-bid advantage you just built. Closing the loop means adding each graduated term as a negative in the auto campaign so its spend routes to your deliberate manual bid.

1. Add the harvested term as a negative exactNegative exact is the conservative first move: it blocks only the precise term you promoted, leaving the rest of the theme free to keep discovering.
2. Escalate to negative phrase only after several cyclesNegative phrase is broader and easier to over-apply. Reach for it only when multiple report cycles show a whole family of variants belongs in manual.
3. Stay inside the ceilingsYou have generous headroom, but it is finite, so track how many negatives each campaign and ad group carries as the loop runs for months.

CH.06Scale the winners with bid-by-placement and dynamic bidding

With proven terms isolated in manual exact, you can now pour fuel on them without the risk that comes from doing this on unproven auto matches. Two levers stack: bid-by-placement multipliers and dynamic bidding. Understanding how they compound is the difference between scaling profitably and surprising yourself with a blown daily budget.

Bid-by-placement lets you add up to a 900 percent, or 10x, increase to the base bid for placements such as product pages and Amazon Business. That adjustment stacks on top of your chosen bidding strategy. Dynamic bidding up and down can then raise the bid further, by up to 100 percent in real time, when Amazon judges a conversion likely.

A one dollar base bid with a 50 percent product-pages placement adjustment is one dollar fifty, and dynamic bidding can push that to three dollars when a conversion looks likely. Always work the stacking math before you turn the dials. Amazon Ads, bid-by-placement help

• Apply placement multipliers only to proven manual exact terms, never to raw auto discovery.
• Mirror the current help-page wording, which lists product pages and Amazon Business placements, rather than legacy top-of-search splits.
• Use dynamic bidding up and down on winners where you want Amazon to lean in on high-intent impressions.
• Recompute your worst-case effective bid: base, times the placement multiplier, times the dynamic adjustment.

CH.07Why you never turn the auto campaign off in 2026: the AI-placement dividend

Here is the part the old playbook could not have anticipated. As of March 25, 2026, Amazon Sponsored Products prompts moved from open beta to general availability in the United States. These are AI-conversational placements tied to Rufus, Amazon's shopping assistant, and they match products to natural-language questions rather than to exact keywords.

The decisive detail for campaign structure is enrollment. Existing Sponsored Products and Sponsored Brands campaigns are automatically enrolled in prompts and leverage existing campaign parameters, such as campaign targeting, with no additional setup, billed under your existing cost-per-click bidding. Your auto campaign's targeting is the surface that feeds this.

Treat Rufus as the assistant interface and "Sponsored Products prompts" as the durable name of the placement. That keeps the post accurate regardless of any interface renaming, and it keeps the structural takeaway clean: the auto campaign is no longer a temporary research tool. It is permanent AI-coverage infrastructure that happens to also discover keywords. For the organic side of this same shift, see how to earn ASIN visibility inside Rufus at https://www.capconvert.com/learn/blog/amazon-rufus-asin-visibility, and the broader marketplace GEO play at https://www.capconvert.com/learn/blog/geo-for-marketplaces-getting-etsy-ebay-and-amazon-listings-surfaced-by-ai-shoppe.

CH.08Operating cadence and common mistakes

The loop is a rhythm, not a project. The accounts that compound are the ones that run the same harvest-and-negate motion every week, then review bids monthly. Sharpen the discovery side by pairing the auto search-term report with Search Query Performance data, covered at https://www.capconvert.com/learn/blog/amazon-sqp-2026, and tie the negation mechanic to a fuller strategy at https://www.capconvert.com/learn/blog/negative-keyword-strategy-for-2026-lists-scripts-and-automation-tips.

The recurring mistakes we fix on inherited accounts

• Negating a term before it has enough data, which throws away a query that was about to prove itself.
• Over-negating with negative phrase too early, suppressing a whole family of profitable long-tail queries.
• Leaving a single blended default bid across all four auto themes, which makes their economics unreadable.
• Treating prompts coverage as a reason to ignore manual control, when the loop needs both.
• Turning the auto campaign off after harvesting, which severs both discovery and AI-shopper visibility.

If a brand is still deciding whether Amazon Ads belongs in the mix at all, that decision comes before the loop; the upstream context lives at https://www.capconvert.com/learn/blog/amazon-ads-for-e-commerce-brands-when-to-invest-beyond-google-shopping. Once you have committed, the loop is what keeps the program both controlled and visible: discover with auto, prove in exact, negate to stop self-competition, scale the winners, and never switch off the engine that now also carries you into AI search.