An effective Amazon PPC campaign structure separates discovery, harvesting, and brand defense wherever those jobs need independent spending or settings. Use campaigns for independent control, ad groups for coherent product-and-target groupings, and a naming register to keep ownership visible. Restructure an existing account through a documented handoff—not by copying every target and immediately switching the old campaigns off.
This guide is an account architecture and migration template for Sponsored Products. It does not prescribe your first month’s launch plan or recalculate your advertising margins. Use the beginner setup for the former and the profit-first PPC playbook for the latter.
Platform documentation checked 2026-11-04: Amazon Ads’ campaign creation instructions, ad-group documentation, and Sponsored Products targeting guide. The architecture below is a working template, not a claim of account-level testing.
Amazon PPC campaign structure: split at the control boundary
Start with a decision that the current structure makes difficult. Perhaps branded demand hides weak generic acquisition, discovery consumes money reserved for proven queries, or products with different economics share the same decision rule. A new campaign should solve that problem, not merely make the account look tidier.
Amazon’s ad-group documentation, checked 2026-11-04, says a campaign contains one or more ad groups and that manual targets in an ad group apply to the products within it. Its campaign creation instructions place the daily budget in campaign settings. Consequently, moving products into separate ad groups is not the same as giving them independent campaign budgets.
| Decision you need to make independently | Recommended boundary | Reason for the split |
|---|---|---|
| Protect discovery spending from proven-target spending | Campaign | Each job needs a visible spending decision |
| Evaluate branded searches apart from generic acquisition | Campaign | Blended returns obscure the acquisition question |
| Group similar products using a coherent target set | Ad group | The product-to-target relationship remains readable |
| Change one keyword or product target’s bid | Target | A separate campaign is unnecessary for this alone |
| Assign responsibility and document exceptions | Naming register | Labels explain ownership without multiplying entities |
Amazon also documents campaign-level placement metrics and adjustments in Adjust Sponsored Products bids, checked 2026-11-04. If two product groups need fundamentally different placement policies, that is another reason to consider separate campaigns.
Do not split solely because two ASINs have different names. Split when their margins, shopper intent, availability, or required controls demand different actions. Conversely, avoid grouping dissimilar products just because they belong to the same brand. A target that suits one product may be a poor match for another.
Give discovery, harvesting, and defense different ownership
Use three lanes as an operating map. The labels describe responsibilities; they are not additional ad types in Amazon’s console.
| Lane | What belongs here | What the owner decides | What should leave the lane |
|---|---|---|---|
| Discovery | Unproven queries or product contexts being explored | Which opportunities deserve further evaluation | Validated opportunities with a named destination |
| Harvesting | Selected queries or product targets with supporting evidence | Whether the controlled target earns continued spending | Targets that no longer meet the product’s objective |
| Brand defense | Your own branded search intent | Which branded demand to support and how much to allocate | Generic acquisition and competitor-brand experiments |
Discovery needs an exit route
Document what the campaign is learning and where a useful finding will go. A discovery campaign without a destination becomes a permanent holding area. A destination without a source record loses the evidence behind its targets.
Separate automatic and manual discovery implementations where their setup requires it. Amazon’s campaign creation guide distinguishes automatic from manual targeting, checked 2026-11-04. You do not need an additional campaign for every vocabulary label if no independent decision follows from it.
Harvesting needs a specific destination
Record the actual shopper query or product context, the source campaign and ad group, the advertised product, the evidence window, and the destination target. For a query, specify the intended keyword match type. For a product context, use an appropriate product-targeting destination rather than treating an ASIN as a shopper keyword.
Keep the method for choosing negatives in the negative-keyword workflow. Here, the architectural requirement is simpler: every proposed exclusion must have an owner, a scope, and a reason.
Defense needs its own denominator
Amazon’s targeting guide explicitly discusses branded keywords and brand defense, checked 2026-11-04. Keep your own brand intent separate from nonbranded harvesting so an easy branded conversion does not make generic acquisition appear healthier than it is.
A defense label does not establish incremental sales. It tells you what demand the campaign addresses. Review branded performance on its own objective; do not interpret its return as evidence that all the account’s prospecting works. Competitor-brand targeting belongs in a separately identified acquisition experiment, not in your own-brand defense totals.
Use names that survive exports and staff handoffs
Adopt a stable field order rather than an elaborate code nobody can decode. Use a separate register for details that change frequently.
A practical campaign-name template is:
MARKET | ADTYPE | PRODUCT-FAMILY | ROLE | TARGETING | VERSION
For example, US | SP | TRAVEL-MUG | DISCOVERY | AUTO | V1 identifies a US Sponsored Products discovery campaign for a travel-mug family. Its harvesting counterpart might use HARVEST | KW-EXACT; a branded keyword campaign might use DEFENSE | KW. These are illustrative names, not performance examples.
Use a shorter ad-group template: PRODUCT-SET | INTENT. A product-set code should point to an explicit ASIN list in the register rather than rely on a colleague remembering which variations were included.
Amazon’s ad-group documentation says ad-group names must be unique within a campaign but can repeat across campaigns, checked 2026-11-04. Therefore, never use the ad-group name alone as your migration key. Preserve the source and destination identifiers available in your exports alongside their readable names.
Keep the following register fields:
- Campaign and ad-group identifiers, current names, and marketplace.
- Advertised ASINs or SKUs and the owner responsible for them.
- Lane, targeting method, and the intended shopper or product context.
- Current budget, bidding strategy, and additional bid adjustments.
- Performance objective, evidence window, and next review trigger.
- Source-to-destination mapping, migration status, and exception notes.
Do not embed a current bid or a temporary promotion in the permanent name. Those values change, leaving names misleading. Version the structural change instead, and preserve prior names in the register so historical reports remain interpretable.
Migrate through a handoff, not a mass replacement
Begin with an export and a saved baseline. Capture the existing product set, active targets, negatives, budgets, bidding settings, placement adjustments, and any rules affecting delivery. Record inventory, price, or promotion changes that could interfere with comparison.
Use the bulk-operations guide for upload mechanics. The migration document decides what should change; the upload file merely implements that decision.
Map the old account before building destinations
Assign each existing campaign a lane or mark it mixed. For a mixed campaign, identify which targets and products belong together after the split. Retain the original identifiers so historical performance remains traceable even if you rename an entity.
Do not promise to move historical results into a new campaign. Maintain a reporting bridge between the old and new entities. A clean destination starts a new observation stream; the source export supplies its context.
Prepare the destination and its spending envelope
Create the intended product-and-target combination and verify its settings before routing traffic away from the source. Confirm the advertised product is available and eligible. Amazon’s campaign creation instructions say only active, in-stock ASINs can be advertised, checked 2026-11-04.
Assign the destination’s budget deliberately. Running a source and destination together can expose both budgets to spending; copying the source budget is not a neutral bookkeeping action. Record the combined amount you intend to make available during the handoff.
Keep bids and other settings stable where possible. If the migration also introduces a new bidding strategy, major price change, and placement policy, you will struggle to interpret the result.
Route only after checking the destination
For a harvested query, first verify that the destination target is enabled, relevant to the advertised product, and capable of delivery. Then decide whether to leave source coverage in place or apply a narrowly scoped exclusion. Do not assume adding the destination automatically removes source eligibility.
Use exclusions to manage routing and reporting—not as a blanket claim that duplicate eligibility makes you bid against yourself. Amazon’s targeting guide says using different match types on the same keyword does not make you bid against yourself, checked 2026-11-04.
A routing negative changes eligibility; it is not a guaranteed traffic-transfer switch. If the destination stalls, inspect its budget, bids, product eligibility, and targeting before expanding exclusions. Keep a record of the exact source exclusion so it can be reversed without rebuilding the whole account.
Accept the migration against a written checklist
An account with prettier names has not necessarily migrated successfully. Accept each source-to-destination handoff separately.
| Acceptance gate | Evidence to save | If it fails |
|---|---|---|
| Product mapping is correct | Advertised ASIN list beside the intended product set | Fix the mapping before excluding source traffic |
| Targets have a destination | Query or product-context mapping and destination identifier | Leave the handoff open |
| Settings match the plan | Budget, strategy, bids, and additional-adjustment snapshot | Correct the unintended difference |
| Routing is intentional | Source negatives, scope, and reason | Remove unintended exclusions |
| Delivery is possible | Eligibility checks and observed impressions or clicks | Diagnose the destination before retiring coverage |
| Economics remain readable | Separate lane totals and combined source/destination view | Repair reporting before judging the split |
| Reversal is possible | Saved source settings and a change log | Reconstruct the missing baseline first |
Choose the review point using traffic volume and the applicable attribution window, rather than accepting a migration merely because a calendar week passed. Early delivery checks and later performance checks answer different questions. The former establishes that the handoff functions; the latter evaluates whether it deserves continued spending.
Keep old and new results separate in the raw data, then combine them in a clearly labeled transition view. Otherwise, a drop in source sales can look like lost demand when the destination now receives those sales. Also inspect the combined spend: maintaining volume by spending much more is not automatically a successful restructure.
If delivery or economics deteriorate, reverse the specific routing change or setting responsible. Keep the register intact. Recreating the old account from memory defeats the purpose of a controlled migration.
Frequently Asked Questions
How many Amazon PPC campaigns should one product have?
Use the number needed for independent decisions, not a fixed campaign-per-ASIN formula. Discovery, harvesting, and defense can justify different budgets and policies. Additional splits need a similarly clear control or reporting benefit.
Does every keyword need a separate campaign?
No. A separate target bid does not itself require a campaign. Isolate a keyword when it genuinely needs independent campaign-level control and generates enough activity to make that separation useful.
Should I immediately negate every harvested query in discovery?
No. Check the destination and decide the intended routing first. A source exclusion can remove useful coverage before the destination delivers. Document its scope and keep the reversal simple.
Should I rebuild everything at once?
Migrate a coherent product group first. Accept its mapping, settings, routing, and reporting before repeating the pattern. That gives you a reusable architecture and a bounded troubleshooting problem rather than an account-wide guessing exercise.