Amazon Campaign Optimization: What to Change Weekly, Monthly, and Never

Most Amazon accounts we inherit aren't underperforming because the campaigns are built wrong. They're underperforming because somebody optimizes them every day.

Daily bid changes on a campaign with eleven clicks a week isn't optimization. It's noise amplification. You're reading variance, calling it signal, and changing the bid before the data could have told you anything. The account looks busy and goes nowhere, and the person doing it is working hard, which is the part that makes it difficult to stop.

So the useful question isn't how to optimize Amazon campaigns. It's what to change, how often, and what to leave alone. Below is the routine we run across the accounts we manage and the reasoning behind each interval. If you'd rather hand the whole cadence to a team that does only this, that's what our Amazon PPC management engagement is.

Why change frequency matters more than change quality

Every optimization decision has a minimum amount of data underneath it before it stops being a guess. A bid change is a bet on a conversion rate. If the keyword has produced nine clicks and one order, you don't know its conversion rate. You know one outcome.

The practical consequence is that the right interval for a decision is however long it takes that part of the account to accumulate enough events to decide. That's different for a search term than it is for a placement modifier, and different again for a whole campaign. Which is why a single daily routine applied to everything is wrong in both directions at once: too fast for the keyword bids, too slow for the budget caps.

Interval	What you're deciding	What the data needs to support it
Daily	Nothing, except budget caps that are hard-stopping a proven campaign	Spend pacing only, not performance
Weekly	Search term harvesting, negatives, budget shifts between proven campaigns	Enough clicks to see obvious waste, which accumulates fast
Monthly	Keyword bids, placement modifiers, dayparting	Enough orders per target to estimate a conversion rate
Quarterly	Campaign structure, targeting expansion, creative, ASIN-level allocation	A full purchase cycle, including the lagging attribution window

What to change weekly

Search term harvesting and negatives

This is the one job that genuinely rewards weekly attention, because waste accumulates in clicks rather than orders and clicks arrive quickly. A search term that has taken 40 clicks and produced nothing is a decision you can make with confidence after a week. The same 40 clicks spread over six weeks is the same decision made five weeks late.

Pull the search term report, sort by spend, and work down until the rows stop mattering. Terms with spend and no orders become negatives. Terms with orders and good efficiency get promoted into exact match in a campaign where you control the bid directly. Nothing else on this list is as reliably profitable per minute spent.

Budget caps on campaigns that are already working

A campaign hitting its daily cap by noon is a campaign telling you it would spend more at the efficiency it's already producing. That's not a performance judgment, it's a pacing fact, and you can act on it immediately. The reverse is also true and more commonly missed: a campaign that hasn't come close to its cap in a month doesn't need a budget change, it needs a bid or a targeting change, and that's a monthly decision.

What to change monthly

Keyword and target bids

Bids are where most accounts do the most damage, because a bid is a conversion-rate bet and conversion rate is the slowest number in the account to stabilise. We work bids once a month, on targets that have accumulated enough orders to say something, and we leave the rest alone rather than nudging them for the sake of activity.

The honest version of this is that a large share of targets in any account will never accumulate enough data to justify a bid decision at all. They sit in a long tail of one and two click targets. Those don't get optimized. They get aggregated into a structure where the campaign-level numbers are decidable even when the keyword-level ones aren't.

Placements and dayparting

Placement modifiers and dayparting are both pattern claims, and a pattern needs several weeks to separate from a seasonal blip. Top-of-search usually carries a higher conversion rate and a higher cost, and whether the premium is worth paying is an account-specific answer you can only read off a month of data. Dayparting is the one we're most sceptical of in general. Across our client portfolio it's produced less than the time it absorbs, except on accounts with a genuinely concentrated buying window.

What to change quarterly

Structure, targeting expansion, creative, and how budget is split across the catalog. These are the changes with the longest feedback loop and the largest effect, and they're the ones most accounts never get to because the daily routine consumes the attention.

The biggest single lever we've pulled on an account wasn't a bid change at all. On an established home and kitchen brand, the existing strategy was conservative: a limited daily ad window, keyword-only targeting, no category or competitor campaigns. Widening the window, expanding into category and competitor targeting, and allocating budget by ASIN performance history produced a 109% sales increase in one month, with 190% more attributed sales and 197% more orders. None of that was reachable through a bid adjustment, at any frequency.

The four things not to touch

Bids on targets with fewer than roughly ten orders. You're adjusting toward a number you can't see yet.
Campaign structure, more than quarterly. Restructuring resets the learning on everything inside it, and the cost of that reset is usually larger than the structural gain.
Anything, on the basis of a single bad week. Amazon's attribution window is long enough that last week's campaign report is still filling in.
Two variables at once on the same campaign. If you change the bid and the placement modifier in the same session, you've spent the data and learned nothing transferable.

How do you know an optimization worked?

You compare the same period length before and after, on the same campaigns, with the attribution window fully closed, and you accept that some of what you see is still noise. That's the whole answer, and it's less satisfying than a dashboard.

The two failure modes are symmetrical. Judging a change after three days credits it with variance it didn't cause. Judging it after a quarter folds in six other changes and a season. We settle on 28 days for bid and negative work and a full quarter for structural work, which matches the intervals the changes were made on.

Where this routine doesn't apply

Two cases, and both are common enough to name.

If your account is small, the intervals stretch rather than shrink. A brand spending a few thousand a month doesn't have a weekly search term decision, it has a monthly one, and running the weekly routine on it is the noise-amplification problem from the top of this post. The cadence scales with data volume, not with ambition.

And if you're inside a launch or a peak season, the cadence breaks on purpose. A new ASIN has no history to be patient with, and Q4 compresses a quarter of buying into six weeks. Both are the exceptions that prove the rule, which is that the interval should follow the data rate rather than the calendar.

What we'd do first on a new account

Search terms and negatives, weekly, before anything else. It's the highest return per hour in Amazon advertising and it's the job that gets skipped when the daily bid routine owns the calendar. Then one month of untouched bids to get a clean read, which is harder to agree to than it sounds, and is usually the moment a brand discovers how much of its past year of optimization was motion.

If you want to know which interval your account is actually on, send us the account and we'll tell you what we'd change and what we'd stop changing. The structural quarter is where the money is, and most accounts have never had one. Our Amazon advertising team starts every engagement there.