Manual vs automatic Amazon Ads campaigns: what each one is for

Almost every seller who starts advertising on Amazon ends up with the same picture: two campaigns running, one automatic and one manual, both spending, and no certainty about which one is doing what. The automatic usually spends more because nobody watches it; the manual was built with the five keywords that came to mind at launch and hasn’t been touched since. When it’s time to cut, the one with the ugliest ACoS column gets cut, and that’s rarely the one that should go.

The underlying problem isn’t bidding: it’s that most sellers treat both campaigns as variants of the same instrument, one “easy” and one “laborious”. They aren’t. They are two tools with different jobs inside one system. The automatic campaign is not a lazy version of the manual, and the manual is not “the automatic done properly”. One discovers, the other exploits. Confusing them is why a reasonable budget delivers half of what it should.

The practical consequence of that confusion is twofold. On one side you scale the wrong campaign, pouring budget into the one designed to explore cheaply. On the other, both end up competing for the same search term, paying twice for the same click and pushing cost up with no chart to explain it.

There’s a third layer that rarely makes it into the conversation: neither campaign is judged well against its own ACoS. A discovery campaign with a high ACoS can be the best money you spent all month if it handed you three terms you later scaled profitably. A manual campaign with a low ACoS can be buying sales you were going to get anyway. The only honest judge is what’s left over for the product the campaign is financing.

what an automatic campaign actually does

A Sponsored Products automatic campaign doesn’t pick terms: Amazon proposes them by reading your own listing — title, bullets, description, backend search terms — plus the sales signal that ASIN has built up. You set a budget and a bid, and the system decides which searches and which product pages you show against.

That matching isn’t one single bucket. Amazon splits it into targeting groups you can bid on separately: close match and loose match for searches, and substitutes and complements for product pages. The first two put you in front of queries similar to what your listing says you are; the last two place you inside the detail pages of similar or compatible products. Those are four distinct behaviors under one name, and the first thing a seller gains by splitting them is no longer paying a close-match bid for traffic that actually came from loose match.

Worth saying plainly: the names, the groups and the available controls change from time to time, and Amazon doesn’t announce every adjustment loudly. Before you rebuild your structure, open your Amazon Ads console and confirm which groups and campaign types are enabled for your account and your marketplace. What you read in any guide — this one included — is a map, not the territory.

The operational conclusion matters more than the naming: the automatic campaign is a machine for discovering demand, not for scaling it. Its value isn’t in the sales it brings, it’s in the search term report it produces. That report holds what people actually typed before buying from you, including the variants you would never have thought of, the typos that convert and the competitor ASINs where your product holds its own.

what a manual campaign does and what it costs to keep

The manual campaign flips the relationship: you declare what you want to appear against. Keywords with their match type — exact, phrase, broad — or product and category targeting, each with its own bid. Nothing is proposed: what isn’t on the list doesn’t get bought.

That makes it the only place where you genuinely control cost. You can raise the bid on the term that converts at twice the average and lower it on the one that only brings clicks. You can split by match type and discover that exact converts three times better than broad on the very same term, then pay accordingly. You can isolate your own brand in its own campaign to see how much you’re spending on people who were already searching for you by name.

The price of that control is maintenance. An unreviewed manual campaign ages badly: bids that worked three months ago are now under or well over market, keywords that converted stopped doing so because the competition changed, and the new term that exploded in season isn’t on any list. A manual campaign isn’t switched on, it’s cultivated.

And there’s a limit worth accepting from day one: a manual campaign can only exploit what someone already discovered. Without a source of new terms it becomes a walled garden, ever more efficient over an ever smaller slice of demand. That’s why the two campaigns aren’t alternatives: they’re two halves of the same cycle.

Glossary: ACoS is ad spend divided by the sales attributed to those ads; it measures spend efficiency, not profit.

why scaling the automatic campaign is the most common mistake

The mistake looks like this: the automatic campaign brings sales, its ACoS isn’t bad, so the budget goes up. The following month spend rose far more than sales and nobody understands why.

The explanation is structural. When you give an automatic campaign more budget, the system doesn’t buy more of the term that was working for you: it reaches further out toward the periphery, which is exactly what it was designed to do. The good terms were already being served; the new money goes to queries that are further away, with worse purchase intent and worse conversion. Scaling an automatic campaign is paying to explore further, not to sell more of the same.

There’s a second, quieter effect. If a term the automatic campaign discovered also lives in your manual campaign and nobody negated it in the automatic one, your two campaigns are now bidding in the same auction. The result isn’t winning twice: it’s your own cost per click rising while your budget splits across two report rows telling the same story. Diagnosis becomes impossible, because neither campaign shows that term’s real performance.

The corollary is simple: discovery budget and exploitation budget don’t follow the same logic. The first one gets capped; the second one gets scaled for as long as margin allows.

the workflow that runs both together

The cycle mature accounts use is no mystery, but it does demand discipline. It boils down to four moves repeated weekly or biweekly, depending on your volume:

• Discover. The automatic campaign runs on a capped budget and a moderate bid. Its job isn’t to sell a lot: it’s to generate the search term report. If you split the four targeting groups, you’ll also know whether your finds come from searches or from product pages.
• Harvest. You read the report and separate the terms that converted from the ones that only spent. The criterion isn’t “which got more clicks”: it’s which produced orders at a cost the product can pay.
• Exploit. Converting terms get added to the manual campaign, usually as exact match in their own ad group, with a bid calculated from what that product can absorb. That’s where budget gets scaled.
• Separate. The same term you just scaled gets added as an exact negative keyword in the automatic campaign. From that moment each campaign owns its territory and neither buys clicks from the other.

That fourth step is the one almost nobody takes, and it’s the one that turns two noisy campaigns into a system. Without it, the cycle breaks on its second lap.

negatives are the border, not a junk filter

Negative keywords usually get assigned a single job: blocking irrelevant searches that spend without selling. True, and useful, but that’s half of what they do.

The other half is architectural. An exact negative in the automatic campaign, applied to a term that already lives in the manual one, is what keeps the two roles apart. The automatic campaign stays free to keep exploring new ground — the only thing it does well — and the manual one keeps proven demand with no internal competition. The negative stops being a filter and becomes the line that divides discovery from exploitation.

That also changes how you read reports. With the border in place, the automatic campaign’s ACoS becomes an indicator of exploration cost, and the manual one’s an indicator of efficiency over known demand. Two numbers that no longer contaminate each other, and that can finally be compared against different things: the automatic one against how many useful terms it delivered, the manual one against how much profit it left.

It’s worth treating the negative list as a catalog asset rather than a monthly cleanup: review it, document it, and inherit it when you launch a similar product.

a campaign is judged against the contribution of the product it finances

Here’s the point that almost every conversation about campaign structure skips. Neither campaign is evaluated well by its own isolated ACoS, because ACoS doesn’t know what the product cost, what the category referral fee took, what fulfillment took, what inbound shipping cost, or how much came back as returns.

An example with invented numbers that do add up. A product you sell for 1,160 pesos including VAT: the price before tax is 1,000. Subtract 420 of product cost, 130 of category referral fee, 95 of fulfillment and 25 of returns provision. You’re left with 330 of contribution per unit before advertising, that is 33% of the pre-tax price. That 33% is the absolute ceiling of what this product can pay in ads without losing money, and that’s where your break-even ACoS comes from. If your manual campaign runs at 18%, you keep 15 points of contribution; if the automatic runs at 40%, every discovered sale cost you money, and it’s only justified when the harvested terms get exploited later at that 18%.

That’s the correct reading: the automatic campaign is paid for by what the manual one earns afterward. It’s amortized research spend, not a sales campaign that went wrong. And that’s precisely why its budget is capped: research that never turns into exploitation is pure cost.

The same reasoning explains why two products with the same ACoS don’t deserve the same treatment. One with 45% contribution can take an aggressive manual campaign and a generous automatic one; another at 14% can’t even afford discovery, and its advertising should probably be limited to defending its own brand. Without the contribution number per SKU sitting next to the campaign report, that decision is made on intuition.

Glossary: real net margin is what’s left after ALL costs — product, fees, shipping, returns, tax and advertising — not just price minus cost.

how this reads in iqseller

In the panel, advertising doesn’t live alone. The Profitability module builds net margin per SKU from the COGS you load, the commissions that come from the Amazon settlement and from MercadoLibre orders, FBA and Full fees, shipping, and the VAT and withholding breakdown. That number defines how much each product can pay in ads, and it’s the floor for reading any campaign.

With that settled, ad spend stops being a loose column and becomes one more line in each SKU’s cost structure. The question shifts from “what ACoS does this campaign have?” to “what was left for this product after paying for it?”. And because the Parent → Model → SKU tree is unified, the answer can be read at whatever level you need: the full model when you decide budget, the individual SKU when you decide a bid.

The Alerts module closes the loop on the side that usually breaks it: an ACoS that spikes overnight is often not the campaign’s fault but a conversion problem, and conversion drops when real available stock runs out or when price moved. Seeing advertising next to inventory and price avoids the single most expensive wrong diagnosis there is, which is cutting bids on a healthy campaign when the problem was in the warehouse.

the weekly cycle that holds it all together

None of this works as a one-afternoon project. It works as a short, repeated routine: read the automatic campaign’s search term report, harvest what converted, add it to the manual campaign at the bid margin allows, negate it in the automatic one, and start over. Half an hour a week, well spent, beats a full restructure every six months.

Two reminders so you don’t lose the thread. First: the automatic campaign is never scaled, it’s capped — its budget is the cost of your research, and that cost needs a ceiling. Second: the manual campaign scales for as long as the product’s contribution allows, not for as long as the ACoS “looks fine”.

And one last piece of hygiene: confirm in your Amazon Ads console which campaign types, targeting groups and controls you actually have available today. It’s one of the parts of the platform that changes most, and a structure built on an old assumption degrades without warning.