Foundational Principles
Every structural decision you make is a trade on the same spectrum: control versus simplicity. More granular campaigns give you more levers to pull. But more levers also means more campaigns to manage, more budgets to monitor, and more data spread thinly across more places.
The question isn't "what's the best structure?" — it's "what's the right level of granularity for this account, this catalog size, and my capacity to manage it?"
The Mental Framework: Campaigns, Ad Groups, Products
Before you build anything, understand what each level controls. Mixing up what belongs where is one of the most common structural mistakes.
Settings
Budget controls, placement modifiers (ToS, product pages, rest of search), dynamic bidding strategy
Targeting
Product-to-keyword relationships, individual keyword bids, ad group negative keywords
Performance
Conversion rate, revenue per click, relevancy signals — what informs your bid strategy
Products should be grouped together when they share: similar budgets, similar placement performance, compatible bidding strategies, and aligned ACOS goals. The moment any of these diverge significantly, those products need separate campaigns.
Amazon's Two Jobs
Amazon's advertising system has two objectives that are in constant tension:
Deliver Great Shopping Experiences
- Show relevant products at competitive prices
- Surface the products most likely to convert
- Customer satisfaction drives repeat shopping
Maximize Ad Revenue
- Advertising is one of Amazon's top 3 revenue sources
- Amazon will not voluntarily reduce your impressions
- More keywords = more opportunities for Amazon to charge
Campaign Naming Conventions
This is the most boring topic in PPC and one of the most impactful. Standardized naming is the difference between instantly knowing what a campaign does and spending 20 minutes pulling reports to figure it out.
Reporting
Filter performance by product, tactic, or match type instantly using search — no custom report builds.
Organization
Understand any campaign's purpose, targets, and goals from the name alone — no clicking into it.
Management
Troubleshoot and optimize with zero guesswork. Search "NB | Exact" to see all non-brand exact campaigns.
The Naming Template
Brand vs. Non-Brand: The Sacred Separation
This is non-negotiable. Always separate brand and non-brand campaigns. If you don't, branded search terms (which naturally have higher CVR and lower ACOS) will leak into your non-brand campaigns through broad and phrase match — inflating your non-brand metrics and hiding real performance.
In one account I audited, 26% of total ad sales in "non-brand" campaigns were actually from branded search terms. The account owner thought non-brand performance was strong — it wasn't. The branded traffic was masking poor non-brand conversion rates. Negate your brand name and all variations as negative phrase matches in every non-brand campaign from day one.
The Four Campaign Structures
Every Amazon PPC account uses one of these four approaches — or a hybrid. The right one depends on your catalog size, budget, and management capacity.
| Structure | Best For | Control | Complexity |
|---|---|---|---|
| MPAG | Simplest accounts, initial testing | Low | Minimal |
| SPAG (Baseline) | 80-90% of all accounts | High | Moderate |
| Single Product Campaign | Fewer products, high spend | Very High | High |
| Single Keyword Campaign | VIP ranking keywords only | Maximum | Very High |
1. MPAGs: Multiple Products, One Ad Group
One campaign, one ad group, multiple products all sharing the same keywords and bids. This is the structure most sellers start with — and the one that causes the most damage at scale.
Catch-all campaign with $13,000 in spend across 63 products in a single ad group. ACOS ranged from 15% to over 100%. No ability to fix underperformers without damaging the profitable ones. Only one ad per ad group shows per query — so 63 products competing for one slot.
The exception: for massive catalogs (hundreds or thousands of SKUs), catch-all campaigns can work if products are intentionally grouped by similar ACOS targets. But for most accounts, this structure destroys visibility into what's actually working.
2. SPAGs: The Baseline for 80-90% of Accounts
One campaign organized by parent ASIN or product line, with each child variation in its own ad group — each with its own keywords and bids. Moving from MPAGs to SPAGs is the single biggest performance jump most accounts will ever make.
Multiple Products, One Ad Group
- No product-keyword conversion visibility
- Can't set different bids per product on same keyword
- One ad per ad group per query — one placement total
- Can't negate for one product without affecting others
Single Product Ad Groups
- Full product-to-keyword conversion visibility
- Individual bids per product-keyword combination
- Multiple products can win placements on same query
- Ad group-level negatives target specific products
3. Single Product Campaigns
One product per campaign. Everything SPAGs offer, plus campaign-level budget caps and placement modifiers at the individual product level. Best when you have fewer than 50 products, require product-level budget control, or products have significantly different placement performance.
With 400,000 products, five campaigns per product equals two million campaigns. Auto campaigns would each need negative ASINs for all other owned products. At scale, single product campaigns become operationally impossible. SPAGs with smart grouping handle scale far better.
4. Single Keyword Campaigns
The most granular approach — one keyword per campaign. Maximum control. Covered in depth in Part 8.
Not Sure Which Structure Fits Your Account?
I'll audit your current campaigns and recommend the right architecture for your catalog size and goals.
Match Types: The Targeting Dial
Match types control how broadly or narrowly Amazon interprets your keywords. Think of them as a dial — from wide-open research to surgical precision.
| Match Type | Function | Example Keyword | Sample Triggers |
|---|---|---|---|
| Broad | Widest net — synonyms, misspellings, reordered words | tennis racket | "racquet for tennis," "best tennis rackets 2025" |
| Phrase | Keeps keyword intact and ordered — allows words before/after | "tennis racket" | "best tennis racket for beginners," "tennis racket sale" |
| Exact | Highest CVR, lowest volume — only the keyword + minor variations | [tennis racket] | "tennis racket," "tennis rackets" |
| PAT | Targets specific ASINs — shows on detail pages AND in search | ASIN: B07XYZ123 | Competitor product page, search results where ASIN ranks |
Should You Segment by Match Type?
Yes. Match type segmentation is one of the highest-leverage structural decisions you can make. Each match type behaves differently, generates different search terms, and requires different bid strategies. Mixing them in one campaign means losing control over all three.
Keep broad, phrase, and exact in separate campaigns. This gives you different budgets for research (broad/phrase) versus performance (exact), per-match-type placement modifiers, and independent optimization without tradeoffs.
Auto Campaign Match Types
Auto campaigns have four targeting segments: Loose Match, Close Match, Substitutes, and Complements. Loose match and substitutes typically perform strongest for discovering converting terms. Complements often produce mixed results.
Should you split autos by segment type? Usually not. Splitting creates four campaigns per product, distributes data thinly, and adds budget complexity for marginal gains. Only split if you need different placement settings or budgets per segment.
The Standard Campaign Stack
For each product (or product line in SPAGs), you need a set of campaigns that cover the full spectrum from discovery to performance. Here's the standard stack.
| # | Campaign | Bucket | Purpose |
|---|---|---|---|
| 1 | Auto | Research | Wide-net keyword and ASIN discovery |
| 2 | Manual Broad/Phrase | Research | Explore long-tail variations beyond auto |
| 3 | Manual Exact | Performance | Proven keywords with precise bid control |
| 4 | Product Targeting | Performance | Exact ASIN targets — competitor conquest + cross-sell |
| 5 | Brand Defense | Defense | Protect branded search terms (optional) |
Source Campaigns vs. Destination Campaigns
Your campaigns function as a funnel. Source campaigns discover. Destination campaigns optimize.
Create a campaign mapping document listing every source → destination relationship: source campaign name, source ad group, destination campaign, destination ad group, match type. This makes harvesting fast, repeatable, and error-free.
All Variations Should Bid on the Same Keywords
If your product has five color variations, all five should bid on your main keywords — not just the top seller. Amazon tailors search results based on shopper demographics, purchase history, and preferences. The premium buyer sees your Pro version. The bargain hunter sees the Basic. Bidding only the top performer makes you invisible to other customer segments.
Keyword Harvesting Done Right
Harvesting is the process of moving converting search terms from research campaigns (auto, broad/phrase) into performance campaigns (exact). Most sellers harvest wrong — they grab everything with a single order and stuff it into manual campaigns.
The #1 Harvesting Mistake: Over-Harvesting
Low-volume bloat
One-click, one-sale terms are rarely searched again. Harvesting them creates thousands of keywords averaging 0-1 monthly clicks, bloating your account with dead weight that bidding systems then try to optimize around.
Coincidental orders
A customer searches "bananas," sees your peanut butter ad, and buys it. That doesn't make "bananas" a good keyword for peanut butter. One order doesn't confirm relevance.
Brand halo sales
Customer clicks your Product A ad but buys Product B instead. The search term report shows an order — but the keyword isn't relevant to the product that was advertised. Always verify the right SKU got the order.
Death by 1,000 cuts
Bidding systems increase bids on low-impression keywords to win placements. Each individual increase is small, but across thousands of harvested terms, collective CPCs creep up invisibly.
The Right Harvesting Criteria
| Criterion | Why It Matters | How to Check |
|---|---|---|
| 2+ orders minimum | Dramatically reduces coincidental sales, confirms real search volume | Search term report, 30-60 day lookback |
| Same-SKU orders | Verify the advertised product actually received the order, not a halo sale | Search term report: purchased ASIN column |
| Relevancy check | Search the term on Amazon — are the top results similar to your product? | Manual search on Amazon.com |
| Search volume exists | Confirm the term is actually searched regularly, not a one-off | SQP, Helium 10, Brand Analytics |
Should You Negate Harvested Terms from Source Campaigns?
Default answer: No. This is a common debate. Most of the time, negating harvested terms from source campaigns hurts more than it helps.
"Prevent Cannibalization"
- Fear of bidding against yourself on same term
- Want all spend concentrated in exact campaign
- Believe it wastes money in source campaigns
Overlap Creates Opportunity
- Auto/broad often get lower CPCs for same terms
- Multiple placements = double brand presence on page
- Amazon runs account-level auctions — no self-bidding war
- Auto discovery capacity shrinks with every negative added
Amazon uses account-based auctions. Your highest bid across all campaigns represents your account's top bid. You're not bidding against yourself — you're creating additional opportunities to show up in different positions on the same page. The exception: extremely tight budgets where every dollar of overlap matters.
Negative Keywords: Cutting the Fat
Over-negating is more destructive than over-harvesting. I've seen accounts kill their own traffic by negating everything with high ACOS or zero sales. Here are the three mistakes everyone makes.
Mistake #1: Negating High ACOS Keywords
Any keyword with ACOS has sales. The problem is the bid, not the keyword. Lower your bid instead of negating a converting term.
Mistake #2: Negating High Spend, No Sales
Your "high spend" threshold is often wrong. For a $20 product at 50% target ACOS, you need $10 in spend per sale. Three clicks isn't enough data.
Mistake #3: Bulk Negating 60-Day Zero-Sale Terms
If your product needs 10 clicks per sale, a 3-click keyword with no sale hasn't failed — it hasn't had enough attempts. You just killed future traffic.
When to Actually Negate
Negate when the search term is clearly irrelevant to your product. Wrong material, wrong gender, wrong category, wrong use case. Focus negation on high-volume irrelevant terms — one-click terms that are never searched again don't need negation and just clog your negative keyword lists.
Negative Exact vs. Negative Phrase
| Type | What It Blocks | When to Use |
|---|---|---|
| Negative Exact | Only the specific search term | Individual irrelevant terms that share words with relevant ones |
| Negative Phrase | Any search containing the phrase | When an entire theme is irrelevant (e.g., "steel" for a silicone product) |
N-Gram Analysis: Finding Hidden Waste
N-gram analysis breaks search terms into individual words and aggregates performance across every appearance. It reveals patterns invisible at the individual search term level.
A client selling silicone products: individual "steel" search terms spent pennies each — none flagged individually. But n-gram analysis showed every search containing "steel" collectively wasted hundreds of dollars. One negative phrase match lowered account ACOS by 2 percentage points. Run n-grams at onboarding, monthly for the first few months, then quarterly.
Where to Apply Negatives
Applies to All Ad Groups
- Use when a theme is irrelevant across all products in the campaign
- Example: "steel" negated at campaign level for an all-silicone product line
Applies to One Product Only
- Use when something is irrelevant to one product but relevant to another
- Example: "large" negated for small-only product, kept for others
Want Me to Audit Your Campaign Structure?
I'll identify structural issues, wasted spend, and missing opportunities — then fix them.
Single Keyword Campaigns & The Limits Myth
The Myth: Fewer Keywords = More Impressions
There's a persistent theory that Amazon limits impressions when you have too many keywords per campaign. People split campaigns and see more impressions, then conclude the split was the cause.
The real explanation: splitting campaigns usually resets budgets. Each new campaign gets its own daily budget ($10-$20), increasing total potential spend. Or bid and placement settings get adjusted in the process. Correlation isn't causation.
25 keywords with a $100 daily budget means "spend $100 on these 25 keywords." Adding 75 more means "spend $100 on any of these 100 keywords." Why would Amazon give you fewer impressions when you've given them more opportunities to charge you?
Why Keywords Actually Get Zero Impressions
| Reason | What's Happening | Fix |
|---|---|---|
| Bid too low | Not competitive in the auction | Raise bid or check suggested bids |
| Budget exhaustion | Campaign runs out of budget — lower-priority keywords never serve | Increase budget or reduce keyword count |
| Low search volume | Keyword simply doesn't get searched enough | Verify volume via SQP or Helium 10 |
| Low relevancy | Amazon's algorithm doesn't think your product matches | Check listing — is the keyword in your title/bullets? |
When SKCs Actually Make Sense
Single keyword campaigns are surgical ranking tools, not default structure. Use them when:
- High-volume, high-importance keyword needs its own budget cap
- Unique placement performance — keyword needs its own ToS modifier
- Aggressive ranking campaign requiring precise daily spend control
- Software infrastructure exists to manage hundreds of campaigns at scale
Yes 5 Products, $50K+/mo
Hyper-competitive niche, high spend — SKCs for top 10 keywords make sense.
Test 20 Products, $15K/mo
SKCs for top 3-5 ranking keywords per product. SPAGs for everything else.
Wait 100 Products, $10K/mo
Budget is spread too thin. Get SPAGs right first. SKCs only for your #1 product.
No 5,000+ Products
Operationally impossible at this catalog size. SPAGs with smart grouping is the answer.
The Budget Cap Advantage
The most powerful SKC use case: maintain a high bid for premium placement while capping daily budget to limit total spend. You keep order velocity for ranking while controlling ACOS through spending constraints — impossible to replicate inside a shared campaign.
The Complete Weekly Playbook
Monday: Bid Optimization
Review 7-30 day data. Adjust bids using RPC methodology. Use aggregate benchmarks for low-data keywords.
Tuesday: Search Term Review
Download search term reports. Identify harvesting candidates (2+ orders, relevant, has volume). Flag negation candidates (irrelevant, high volume).
Wednesday: Harvesting & Negation
Build bulk sheets. Move harvested keywords to exact campaigns with performance-based bids. Add negative keywords where appropriate.
Thursday: Budget & Placement Review
Check for budget-limited campaigns. Review placement performance. Adjust ToS modifiers based on data.
Friday: Reporting & Strategy
Pull TACOS trends. Compare ACOS by tactic (brand vs. non-brand). Identify products needing structural changes.
The Bulking & Cutting Rhythm
Amazon PPC scaling follows a natural oscillation — like strength training.
Grow & Discover
- Add keywords, launch campaigns, increase bids
- Expand targeting, invest in discovery
- Expect higher ACOS — you're collecting data
- Goal: find what converts
Optimize & Tighten
- Optimize bids, negate irrelevant terms
- Pause underperformers, tighten budgets
- ACOS decreases as efficiency improves
- Goal: convert investment into profit
You can't stay in bulking forever — money bleeds. You can't stay in cutting forever — growth stagnates. Oscillating between phases is how sustainable scaling works.
Play in the Gray
There is no single correct campaign structure. Single keyword campaigns benefit some accounts and hurt others. Over-harvesting destroys some, under-harvesting starves others. Source campaign negation makes sense in specific scenarios and damages most accounts.
Campaign structure isn't a recipe you follow — it's a system you learn to read. The best operators aren't the ones with the "right" structure. They're the ones who can diagnose an account, understand its constraints, and deploy the right structure for that situation.— Mohsin Raza, Adaptoid E-Commerce
- Single product ad groups at minimum — non-negotiable. Moving from MPAGs to SPAGs is the biggest single performance jump most accounts will ever make.
- Separate brand from non-brand from competitor. Data integrity depends on it. Branded traffic masking non-brand performance is the most common audit finding.
- Harvest with criteria, not impulse. Two orders minimum, same-SKU verification, relevancy check. Over-harvesting bloats your account with dead-weight keywords.
- Negate irrelevancy, not poor performance. If a keyword has sales but high ACOS, fix the bid. Only negate terms that are genuinely irrelevant to your product.
- Proper bidding covers a multitude of sins. Before restructuring your entire account, check your bids first. The problem is usually simpler than you think.
- Name everything consistently. Your future self (and anyone else who touches the account) will thank you. Instant identification beats report-building every time.
- Match complexity to your capacity. An unmanageable structure is worse than a simple one executed well. Build what you can actually maintain.