Most Amazon PPC accounts we look at in 2026 have the same problem. The account-level ACoS looks fine, sitting somewhere around 14%, and the account owner assumes everything is healthy. It isn’t. A blended number like that usually means one campaign is quietly funding another campaign’s losses, and nobody can see it because the structure was never built to separate the two.

Before You Restructure

You’ll want a few things in place before you touch a live campaign:

• Advertising console access with edit permissions in Seller Central.
• Brand Registry: if branded defense includes Sponsored Brands or Sponsored Brands Video.
• At least 30 to 60 days of Search Term Report history. You need to see what your current campaigns are actually doing before you decide what to change.
• A working grasp of ACoS and TACoS. This guide won’t re-explain them.
• Two to three weeks set aside for the migration. This isn’t a one-afternoon project.

Why Your Old Amazon PPC Structure Is Bleeding Money

Run a branded campaign, a non-branded campaign, and a competitor-conquest campaign as one line item, and the account-level ACoS becomes an average of three completely different businesses. Branded search often converts above 20% and costs pennies per click, because the shopper already typed your name. Non-branded search converts in the high single digits and costs real money, because you’re introducing yourself to a stranger.

Here’s what that blending actually does to the math. Say an account spends $20,000 a month at a blended 14% ACoS, which looks perfectly healthy on a dashboard. If branded search is a third of that spend running at 8% ACoS, and non-branded is the other two-thirds running at 22% ACoS, the blend still reads as 14%. Nobody looking at the top-line number would guess that the non-branded piece, the part actually responsible for new customers, is already past most sellers’ break-even point. The account owner keeps funding it because the dashboard never told them to stop.

The same blending problem shows up inside a single campaign. Mix five keywords across broad, phrase, and exact match in one ad group, and Amazon’s delivery engine will funnel most of the impressions toward whichever term has the most volume, usually the broadest and least precise one. Your best-converting exact match keyword sits there starved of traffic while the account spends on vague queries it should have negated months ago.

Add in rising cost per click, and the math gets worse fast. Several independent ad-benchmarking trackers put the 2026 average Sponsored Products CPC somewhere between $1.15 and $1.25, up from around a dollar just two years ago. These are aggregate industry estimates, not an Amazon-published figure, but the direction is consistent across every tracker we checked. At that price, a structure that can’t tell you which segment is actually profitable isn’t a filing problem. It’s a direct hit to margin, every day the account runs.

What Changed: Amazon Is Matching Customers, Not Just Keywords

On May 13, 2026, Amazon retired the standalone Rufus chatbot and folded it into a new assistant called Alexa for Shopping, according to CNBC’s reporting on the launch. The rebrand puts a conversational shopping agent directly into Amazon’s main search bar rather than a separate chat window, and it’s available to any US shopper without an Echo device, a Prime membership, or the Alexa app.

That matters for campaign structure because the assistant doesn’t work off literal keyword strings. A shopper who types “something that keeps my coffee hot for a full workday” isn’t searching your keyword “insulated travel mug,” but that’s exactly the product they need. An exact-match-only campaign has no path to that shopper. Broad match and automatic targeting do, because they’re built to catch related and synonymous queries, not just the literal term you bid on.

This doesn’t mean exact match is obsolete. It means the old hierarchy, where broad and auto were treated as a temporary discovery phase before “graduating” everything to exact, undersells what broad and auto are actually for in 2026. They’re a permanent discovery layer now, not a phase you age out of.

Amazon has also started giving sellers a way to check their own numbers against the market instead of guessing. On May 18, 2026, Amazon Ads made its benchmarks reporting feature generally available worldwide, after limiting it to a US-only beta. The feature compares your ACoS, CPC, CTR, and new-to-brand purchase rate directly against category peers, pulled through the same Reporting API most sellers already use for campaign data. That matters for a restructure, because it means you no longer have to rely on someone else’s benchmark table to know whether your new 4-tier structure is actually performing; you can check it against real accounts in your own category.

The 2026 Amazon PPC Architecture: Four Intent Tiers

The fix for both problems, blended reporting and missed conversational traffic, is the same. Stop building campaigns around keywords and start building them around intent. We organize accounts into four tiers, each in its own campaign and its own budget.

Branded defense. This tier exists to own your own name. When someone searches your brand, you want to appear above any competitor trying to intercept that traffic. Target ACoS here is low, typically 5% to 12%, because the shopper already decided to look for you. Keep this tier isolated. Its rock-bottom ACoS will quietly flatter your account-wide average if you let it blend with anything else.

Non-branded harvest. This is where new revenue comes from. Shoppers here are searching the category, not your brand, and they don’t know you exist yet. Target ACoS sits closer to your break-even margin, since this tier is the actual growth engine. Segment by match type inside this tier. Proven exact-match keywords get their own tightly-bid campaigns, and broad and phrase match feed new candidates up to them.

Discovery. Auto and broad match campaigns whose job is to surface search terms you didn’t know to target, including the conversational queries Alexa for Shopping is now generating. Set a deliberate, capped budget here. You’re paying for data, not immediate profit. Once a search term converts a handful of times, add it to your exact-match harvest campaign and immediately negate it inside the discovery campaign, so you stop paying for the same click twice.

Competitor conquest. ASIN and category targeting aimed at shoppers considering a competitor’s product. This is the highest-risk tier and the one most likely to bleed budget if you don’t cap it. Only run it against competitors where you have a real, verifiable advantage in price, rating, or reviews. Otherwise you’re paying a premium to lose the comparison.

The one-campaign-one-job rule sounds obvious once you say it out loud. Most accounts we audit are still running two or three of these tiers inside a single campaign, which is exactly why nobody can answer a simple question like “how much are we spending to defend the brand versus grow it.”

Match Types in 2026: What Still Earns Its Keep

Broad, phrase, and exact still have distinct jobs. What changed is how much weight each one carries. Broad match now covers a meaningfully wider set of related and synonymous queries than it did three years ago, per Amazon’s own targeting guide, which makes it a real discovery tool rather than a blunt instrument you tolerate while waiting to move everything to exact.

Phrase match sits in the middle. It’s less restrictive than exact and less noisy than broad, which makes it useful for capturing longer, more specific queries that still contain your core term. If you sell an insulated travel mug, phrase match on “insulated travel mug” will pick up “insulated travel mug for truckers” without pulling in the unrelated queries broad match sometimes does.

Exact match is still where you concentrate bid control and budget on keywords with a proven conversion history. The mechanism that ties all three together is the harvest-and-negate loop: a search term proves itself in discovery, gets promoted into exact match with a real bid, and gets negated in the campaign it came from so the account isn’t paying for the same click in two places. A term that converts three times out of forty clicks in a discovery campaign is a strong candidate for promotion. A term that racks up fifteen clicks with zero conversions is a strong candidate for a negative, not a bid increase. Skip the negation step and your own campaigns start bidding against each other, which is one of the fastest ways to inflate CPC without adding a single new customer.

Amazon PPC Bidding Strategy and Budget by Product Stage

Bid strategy should change with the product’s stage, not stay fixed at whatever setting you picked at launch. A campaign three months into ranking should not be running the same bid logic as a campaign that launched last week.

Treat these ranges as a starting point, not a fixed rule. They reflect a general consensus across the ACoS benchmarks we see published industry-wide, and your actual targets should shift based on category, margin, and how competitive your specific niche is.

Dynamic bids, down only, is the safer default for discovery campaigns. Amazon lowers your bid when a click looks unlikely to convert and never raises it above what you set, which limits downside while you’re still gathering data. Dynamic bids, up and down, works for exact-match campaigns where you already trust the conversion data and are comfortable letting Amazon bid more aggressively on your best terms. Fixed bids make sense for branded defense and tightly controlled conquest campaigns, where predictable spend matters more than Amazon’s real-time optimization.

Governing a Multi-Product Portfolio

Everything above assumes one hero product and a single Amazon PPC account structure. Once you’re running 20 or more SKUs across a few categories, the four-tier structure by itself won’t tell you anything useful unless you also group campaigns into portfolios with their own budget caps.

Portfolios are where you enforce the role of each tier at the account level. Cap the branded portfolio low. It should never consume a large share of spend. If it starts ballooning, a competitor is probably bidding on your brand term, and you want to know immediately. Give non-branded the largest allocation, since it’s the growth engine across the whole catalog. Treat discovery as a fixed research budget you’re comfortable spending to learn. Hard-cap conquest, since it’s the tier most likely to quietly run away from you.

Picture a catalog with three product lines and a combined $30,000 monthly ad budget. Instead of one blended number across everything, you’d cap branded defense at roughly $2,000 across all three lines, put $20,000 into non-branded harvest split by each line’s revenue share, keep discovery capped around $5,000 as a fixed research cost, and hard-cap conquest at $3,000. Now a single portfolio view answers the question that a flat structure never could: how much of the $30,000 is defending revenue, growing it, funding research, or attacking competitors.

Granularity should follow revenue, not a rulebook. Your top keywords, the ones actually driving spend and sales, deserve their own exact-match campaigns so you can control bid and budget at the term level. The long tail doesn’t. Group low-volume keywords into themed campaigns instead. Dozens of single-keyword campaigns each pulling two or three clicks a week never accumulate enough data to optimize, and the management overhead becomes its own problem.

The Migration Checklist: Moving From Old to New Without Losing Sales

Don’t pause the old campaigns and flip a switch. Amazon’s delivery algorithm needs consistent data to keep optimizing, and a hard reset throws that data away right when you need it most. Build the new structure in parallel and shift budget gradually.

Common Mistakes When Restructuring

Blending intent inside one campaign. This is the mistake that started the whole problem, and it’s easy to reintroduce by accident during a rebuild if you’re not disciplined about which campaign each keyword belongs to. A single “let’s just test this together” campaign is how a clean structure turns messy again six months later.

Skipping the negative-keyword bridge. A discovery campaign without negatives keeps bidding on terms you’ve already promoted to exact match, so you end up paying twice for the same click and cannibalizing your own reporting.

Restructuring everything overnight. Amazon’s optimization needs time and consistent signals. A same-day teardown resets that and usually costs a week or two of performance you didn’t need to lose.

Timing the migration around a major sales event. Moving budget the week before Prime Day or a big promotion adds volatility exactly when you can’t afford it. Migrate in a quiet period instead, and treat the weeks around any major event as a freeze, not a rebuild window.

Copying an ACoS benchmark that doesn’t match your margin. A 25% ACoS target is meaningless without knowing your product’s actual margin. Set targets from your own numbers and your own category benchmark data, not a number you saw in someone else’s case study.