One account, eight jobs.

There is a particular kind of Amazon advertising account that resists every fix. Bids come down and sales come down with them. Bids go up and ACOS goes up faster. Negatives get added, and a month later the spend has simply moved somewhere else. Every individual decision looks reasonable and the account does not improve.

Almost always, that account has campaigns doing more than one job. And a campaign doing two jobs cannot be optimised for either, because the two jobs disagree about what a good outcome looks like.

Why mixing jobs breaks optimisation

Consider a single campaign that contains your brand name in exact match and a set of broad keywords aimed at people who have never heard of you. Both are legitimate things to spend money on. They have nothing else in common.

The brand terms will convert at a rate no other traffic reaches, because the person typing your name has already decided. They will show an advertising cost of sale in single digits. The broad discovery terms will convert far less often, cost more per order, and show an ACOS three or four times higher — and that is them working correctly, because their job is to find people, not to close them.

Put both in one campaign and you have one budget, one bidding strategy and one number to judge them by. The blended ACOS describes neither. If you optimise toward it, you either strangle the discovery that feeds the account or you overpay for brand terms you were going to win anyway. There is no bid that is correct for both, which is why no amount of bid work fixes this.

The eight jobs

Our budget planner splits advertising into eight objectives rather than three ad types, because ad type is a delivery mechanism and objective is a decision. A ranking push runs on Sponsored Products exact, Sponsored Brands exact and Sponsored Brands Video simultaneously — asking "what share should go to Sponsored Brands?" has no strategic answer, while asking "what share should go to ranking?" does.

The important column is the third one. Four of these eight should never be judged against your target ACOS, and two of them look better than they are. An account-wide efficiency target applied across all eight will kill discovery, protect brand defence that needed no protecting, and tell you nothing about ranking.

The second symptom: budget in the wrong tier

Once campaigns have one job each, a second question becomes answerable that was not before: how is the money split across the jobs?

This is where a lot of accounts turn out to be upside down. Discovery is cheap to start and easy to leave running, so it accumulates. Exact performance campaigns — the ones actually converting — hit their daily cap by mid-morning and stop. Nobody decided this; it is simply where the account drifted over eighteen months of small changes.

Two failure modes show up in the campaign report, they are opposites, and they are routinely confused:

• A campaign spending its full daily budget every day is constrained by budget. If it is also under break-even, it is the cheapest growth available to you and it is being throttled.
• A campaign spending a third of its budget is constrained by bid or by targeting. Raising its budget will change nothing at all, and the fact that this is the most common response to a campaign underperforming is why so many accounts have large budgets and small spend.

Both look identical in a spend column sorted high to low. Only spend read against the daily cap separates them, which is what the Budget Allocation Audit does across the whole account at once.

The third symptom: your campaigns bidding against each other

The last one is the quietest, because it produces no bad number anywhere. Nothing looks broken. You are simply paying more than you need to.

It happens when the same search query can be matched by more than one of your own campaigns — a broad keyword in a discovery campaign and the same term sitting in exact match in a performance campaign. Amazon does not run both; it picks one. But the auction it picks has one fewer competitor than you think, and the price you pay reflects a bid you set for a different purpose.

The usual cause is harvesting done halfway. A term is found in an auto campaign, promoted to exact match — and never negated in the campaign that found it. The promotion was the interesting half of the job, so the negation gets skipped, and now two of your campaigns want the same query.

Working out which of the three you have

All three are visible in reports you can export this afternoon, and none of them needs a tool to see — though a tool makes it take ten minutes instead of an afternoon.

• Mixed objectives: open a campaign and read its keyword list. If brand terms and discovery terms share it, that is the problem, and no bid change will fix it.
• Budget in the wrong tier: pull the campaign report and put spend next to daily budget. Anything at 95% or more of cap is throttled; anything under 40% has a bid problem being treated as a budget problem.
• Self-competition: pull the search term report and look for the same customer search term appearing under two different campaigns. Every instance is a term you are bidding against yourself on.

Fix structure before you touch bids

The order matters more than the individual actions. Bid changes made on top of a broken structure produce noise: the number moves, you cannot attribute the move to the change, and the account is no easier to reason about than before.

Separate the jobs first, even roughly. Negate the terms you are competing with yourself on. Then look at where budget sits against caps. Only after that does a bid change tell you something, because only then does the campaign it applies to have a single definition of success.

None of this is fast. A structural separation on a real account takes weeks, not an afternoon, and it should be done alongside the existing campaigns rather than instead of them — cut everything over at once and you have no way to tell whether the new structure is working or whether you simply broke something.

But it is the work that makes every subsequent decision cheaper. An account where each campaign has one job is an account where a bad number tells you what to do about it.