The core problem: why so many campaigns get shut down too early

A recurring pattern among inexperienced Amazon advertisers looks like this: a campaign launches, the ACoS sits at a multiple of the target value in the first week, and out of concern for the ad budget the campaign gets paused or drastically scaled back. At first glance, this looks like responsible cost management. In practice, this step often destroys exactly the data foundation that would have made the campaign profitable in the following weeks.

The reason: a new campaign starts with no history at all. Amazon does not yet know which search terms match your product, which placements work, and which users actually buy rather than just click. This information only comes from impressions, clicks and purchases — in other words, from exactly the spend that looks high in the early phase.

The expensive fallacy

Anyone who shuts down a campaign in week one because of a high ACoS is judging it by a standard for which there isn't nearly enough data yet. It's like aborting an experiment after the first measurement because that first result doesn't match the target.

What actually happens, technically, during the learning phase

Amazon's advertising algorithm optimizes campaigns based on data. It learns from every impression, every click and every conversion how likely a given search term, placement or audience is to lead to a purchase. At the start, this data simply doesn't exist yet in sufficient volume.

In practice, this means: in the first days, an automatic or broadly targeted campaign serves ads for a relatively wide range of search terms, some relevant, some less so. Some of these search terms convert well, others generate only clicks without a purchase. Without this testing phase, there is no way to find out which search terms fall into the second category.

Without clicks and impressions, there is no data. Without data, there is no optimization, neither for the algorithm nor for you. This applies to Amazon's automatic bidding mechanism just as much as it applies to your own search term harvesting and manual bid adjustments. From experience, this learning process typically takes about four to six weeks in practice before enough reliable data exists for the first major structural and bidding decisions. This is not an official, Amazon-guaranteed timeframe, but a typical value we see across many campaigns, one that we at Sellantica also regularly observe in customer accounts.

Buying data, not losing money

The most important shift in perspective: the ad spend of the first weeks is not lost money, it's the price you pay for usable data. Every click that doesn't lead to a purchase tells you something about which search terms, placements or audiences don't work for your product. Every purchase tells you which combination does work. Both are valuable, even if the "unsuccessful" clicks initially drag down the ACoS.

For newly listed products, a second effect comes into play: without a review history, without an established conversion rate and without organic ranking, a listing simply converts worse than an established product with a hundred reviews. That pushes the ACoS up further, independent of the campaign's own technical learning phase. Both effects, the algorithm's technical learning phase and the missing product history, overlap in the first weeks, which is why the ACoS of new campaigns is often especially high.

Practical tip

Before launch, define a learning-phase budget you're prepared to invest even if the ACoS sits well above your target. This takes the emotional pressure off the first weeks and prevents rushed decisions.

Evaluate the learning phase automatically instead of guessing manually

Sellantica continuously evaluates search term reports, automatically harvests converting keywords and suggests negative keywords as soon as enough data is available. This shortens the learning phase without you having to comb through reports yourself every day.

Try free for 30 days

The typical progression: what to expect in the first weeks

How quickly the ACoS normalizes depends heavily on margin, competitive density, product price, and how consistently you evaluate search term reports. The table below shows a typical progression as we frequently observe it in practice. Treat these values as a rough frame of reference, not a guarantee or an official Amazon figure.

Timeframe Typical ACoS progression What's happening in this phase
Week 1 to 2 Often well above the target, sometimes over 100% Broad data collection, little room for optimization
Week 3 to 6 Usually starts to normalize First rounds of keyword harvesting, first negative keywords
Week 6 to 8 Often approaches a sustainable level, depending on margin frequently around 20% ACoS Bids have been adjusted several times, campaign structure has solidified

Regular bid adjustments and consistent search term harvesting noticeably speed up this process. Negative keywords that clearly exclude irrelevant search terms, and moving well-converting search terms into their own, tightly controlled campaigns, are the most effective levers for shortening the learning phase instead of simply waiting it out.

ACoS vs. TACoS: the wrong headline metric in the early phase

In the first 14 to 30 days, ACoS is the wrong metric to judge a campaign's success by. The reason lies in the interplay between advertising and organic ranking: advertising drives purchases, purchases improve a product's organic ranking position (the so-called sales velocity or flywheel effect), and a better organic ranking in turn lowers TACoS, meaning ad spend relative to total revenue.

Anyone who looks exclusively at isolated ad performance during this early phase systematically underestimates the campaign's actual effect. An ACoS of 80 percent in week two looks dramatic, but it says nothing about how many of the purchases generated through advertising are currently pulling your product's organic ranking upward, and thereby generating free revenue in the future.

We explain in detail how ACoS and TACoS are calculated, and why looking at both metrics together is more meaningful than either alone, in our article on calculating and optimizing ACoS and in our article on TACoS. For the learning phase, the main takeaway is: watch the development of total revenue and your organic sales during this time, rather than the isolated ACoS of individual days.

Important

ACoS remains an important control metric in the long run. During the learning phase, it's simply not the right yardstick temporarily, because it doesn't yet reflect the organic effect of the advertising.

When intervening early is still the right call

Patience during the learning phase is not a free pass to burn unlimited budget. There are concrete situations where actively intervening, even in the first weeks, is right and necessary.

  • Negate obviously irrelevant search terms: if a search term clearly has nothing to do with your product (wrong gender, wrong category, a clearly different product type), you don't need to collect weeks of data before excluding it. Obvious mismatches are visible immediately in the search term report.
  • Concentrate budget on campaigns with a clear signal: if one campaign is already, early on, clearly outperforming another (more conversions at comparable traffic), there's nothing wrong with shifting budget toward the stronger campaign.
  • Hold back products without listing quality or reviews: a listing with weak images, an unclear title or no reviews at all shouldn't be advertised at full budget. Here the problem isn't the campaign's learning phase, it's the listing's own lack of purchase readiness. Improve the listing first, then ramp up PPC.
  • Pause clearly unprofitable individual targets once enough data exists: a single keyword or product target that, after enough clicks (a common rule of thumb is 20 to 30 clicks without a single conversion), still hasn't generated a single purchase, can and should be paused or have its bid cut significantly. That is something different from blanket-shutting down the entire campaign after a few days.

The difference between sensible intervention and premature abandonment lies in the level of the decision: you correct individual, clearly identifiable mismatches immediately. You judge the fundamental structure and overall budget of a new campaign only once enough data exists for a reliable verdict.

Conclusion: collect data first, then judge

A high ACoS in the first weeks of an Amazon PPC campaign is, in the vast majority of cases, not a sign that something is going wrong, but the expected price for the fact that the algorithm, and you yourself, are still learning which search terms, placements and audiences work for your product. Anyone who pulls the emergency brake during this phase prevents exactly the data collection that would have made the campaign profitable.

Give your campaigns typically four to six weeks before making fundamental structural or budget decisions, watch TACoS and total revenue during this time rather than the isolated daily ACoS, and intervene selectively on obvious mismatches instead of shutting down the entire campaign outright. With this balance of patience and targeted intervention, most campaigns reach a sustainable level within six to eight weeks.

Sources and further reading