Let me start by arguing against my own agency’s default.
Amazon’s published guidance on keyword targeting says: “Add at least 25 keywords in your campaigns and ad group.” A single keyword ad group has one. Amazon’s own product documentation is telling you not to build SKAGs, and any post that doesn’t mention that is selling you something.
Now the other half. Most of the “SKAGs are dead” writing you’ll find is about Google, and it doesn’t transfer. Understanding why is the whole decision.
Why SKAGs died on Google, and why that argument doesn’t carry
SKAGs made sense when exact match meant exact. Google spent a decade dismantling that:
| Year | What Google did |
|---|---|
| 2014 | Close variants: misspellings, plurals, stemming |
| 2017 | Exact match expanded to reordered words and function words |
| 2018 | Synonyms and intent added to exact match |
| 2019 | Extended to phrase and modified broad |
| 2021 | Modified broad match eliminated |
| 2023 | Exact match expanded again to “more synonyms and closely related terms” |
By the end of that, a Google “exact match” keyword matched a cloud of related queries. The one-keyword-one-adgroup structure had nothing left to isolate, because the platform had dissolved the isolation.
Amazon has not done this. Amazon’s documented close-variant scope is still only “plurals, misspellings, and translations”, applied to all three match types. Amazon exact match still means “matched word for word (same words and in the same order)” plus those three variations. It is a genuinely tighter control than Google exact match has been since 2018.
So the popular argument: SKAGs are obsolete because close variants killed them — is reasoning from the wrong platform. On Amazon, exact match still isolates.
That doesn’t make SKAGs right. It just means the case against them has to be made on Amazon’s own terms.
The real argument against SKAGs: conversion density
The constraint isn’t platform limits. Amazon allows 10,000 campaigns per account, 20,000 ad groups per campaign and 1,000 keywords per ad group — far more structure than any DTC brand needs. You will never hit the ceiling.
The binding constraint is that Amazon’s automation needs conversion volume per campaign, and fragmentation starves it.
Rule-based bidding is the clearest example. To be eligible, a campaign must have run at least 30 days, have at least 30 conversions in the preceding 30 days, and carry a daily budget of at least $10. If the campaign misses its guardrail across a 21-day evaluation period, Amazon reverts to the previous bidding strategy. (These figures come from two independent secondary sources that agree; I could not retrieve Amazon’s own help article for them, so treat them as well-corroborated rather than official.)
Now the arithmetic. At a typical CPC around $1 and a conversion rate of 10–12%, 30 conversions takes roughly 250–300 clicks, or something like $260–$310 of spend per campaign per month.
A brand running 50 granular campaigns on 200 total monthly conversions averages four conversions per campaign. Every one sits below the threshold. Nothing qualifies for rule-based bidding, dynamic bidding has almost no signal to work with, and you have converted a data problem into fifty smaller data problems.
That’s the honest case for grouping. It isn’t about laziness. It’s that below a certain volume, granularity actively degrades what the platform can do for you.
The real argument for SKAGs: control, where control is worth paying for
Three things a SKAG gives you that a themed ad group cannot:
Bid precision. In a fifteen-keyword ad group, one bid serves terms with different conversion rates and different values. You’re averaging. On a thin-margin product where the gap between your best and worst term is the gap between profit and loss, that average is expensive.
Clean attribution. You know exactly which keyword produced which search term, without inference. In a themed group you’re reading a search term report and guessing at the parent.
Budget ring-fencing. A strategic term gets its own budget that nothing else can consume. On a launch, or on the one keyword that carries a third of your organic revenue, that matters.
The published practitioner position that survives scrutiny: reserve SKAGs for the top three to five highest-volume terms, and only where the keyword alone clears 30 monthly conversions on its own. Above that threshold, the SKAG has enough data to feed the algorithm and gives you the control. Below it, you get the control and lose the automation.
Amazon is building in the opposite direction
Worth knowing, because it tells you where the platform is heading.
Keyword Groups is an Amazon targeting control that “improves campaign performance by dynamically updating the keywords within a group through the campaign lifecycle and eliminates the need for advertisers to constantly curate new keywords.” One bid per group. Amazon recommends pairing it with dynamic up-and-down bidding.
That is the structural opposite of a SKAG, and it’s Amazon’s own product.
Alongside it: target promotion APIs that automate the harvesting workflow — identifying terms to move from automatic targeting into focused manual targeting with optimised bids, which is the job the search term report used to do by hand. And Performance+, an AI-driven tactic that went generally available in late 2025 and automates targeting, bidding and budget allocation across Sponsored Products and Sponsored Display together.
Sponsored Brands Collections launched in the US on 16 February 2026, using AI-powered automatic collections to curate up to ten related products per ad unit. SB placement bid adjustments were raised to 900%, and negative bid adjustments for SB placement groups were deprecated on 15 June 2026 — positive only from that date.
None of this is a reason to abandon manual control. It is a reason to know that manual control is increasingly something you maintain against the current rather than with it.
The comparison, honestly
| SKAG | STAG | |
|---|---|---|
| Bid control | Per keyword, exact | Averaged across the group |
| Attribution clarity | Direct | Inferred from the search term report |
| Conversion density per campaign | Low — fragments the data | High — feeds the algorithm |
| Rule-based bidding eligibility | Rarely reached | Usually reached |
| Negative keyword maintenance | Heavy; cross-negation between groups | Lighter |
| Self-competition risk | Higher, if terms overlap | Lower |
| Amazon’s own guidance | Contradicts it | “At least 25 keywords” |
| Best fit | Top branded and head terms with 30+ monthly conversions each | Everything else |
Two costs of SKAGs that don’t get mentioned enough. Cross-negation: with N single-keyword groups, you need negatives between them or they compete against each other in the same auction, and that relationship grows faster than the number of groups. Budget fragmentation: fifty daily budgets means fifty chances to cap out on a good day while another campaign sits unspent.
What we actually do, and why
Ecom Enable runs strict SKAG architecture for thin-margin DTC brands, and the reason is narrower than “SKAGs are better.”
It’s that when gross margin is 20% rather than 45%, the cost of a wrong bid is asymmetric. A themed ad group averaging across fifteen terms will overbid on some of them, and on a thin-margin product every overbid order is a loss rather than a smaller profit. The manual labour of maintaining SKAGs is cheaper than the margin lost to averaging.
Reverse the margin and the calculation reverses with it. On a 45% margin product with high volume, the automation is worth more than the control, and I’d run themed groups and let rule-based bidding do its job.
The structure follows the margin, not the ideology. Anyone telling you one architecture is correct for all accounts hasn’t looked at your P&L.
How to decide for your account
Three questions, in order:
1. What’s your break-even ACoS? Under about 30%, drift is expensive enough that control wins. Above 40%, you can afford the averaging and should take the automation. Calculate it properly first — most sellers overstate theirs.
2. How many conversions does your best keyword generate monthly? Thirty or more on its own, and it can be a SKAG without starving. Under that, it belongs in a group with related terms.
3. Who maintains it? SKAG architecture requires weekly search term work and continuous cross-negation — worth costing honestly when you compare an agency against a hire. If nobody has that time reliably, a SKAG structure will decay into an unmaintained SKAG structure, which is worse than a well-run themed one.
Whichever you choose, the containment routine for broad match drift is the same. The hybrid that usually wins: SKAGs for the three to five head terms carrying your revenue, themed groups for the long tail, automatic campaigns for discovery, and a monthly harvesting routine moving proven terms from one to the other.
FAQ
What is the difference between a SKAG and a STAG? A single keyword ad group contains one keyword, giving exact bid control at the cost of conversion data density. A single themed ad group contains roughly 5–15 related keywords, giving the bidding algorithm more data at the cost of bid precision.
Are SKAGs dead on Amazon? No, though the argument that they are usually comes from Google, where close-variant expansion dissolved exact match. Amazon’s close variants still cover only plurals, misspellings and translations, so exact match still isolates. Amazon’s own guidance nonetheless recommends at least 25 keywords per ad group.
How many conversions does an Amazon campaign need? For rule-based bidding, reportedly at least 30 conversions in the preceding 30 days, with the campaign running at least 30 days and a daily budget of at least $10. At a $1 CPC and 10% conversion rate, that’s roughly 250–300 clicks a month per campaign.
How many keywords can I put in an Amazon ad group? 1,000 for manual targeting, excluding negatives. Ad group and campaign negative keyword limits are 10,000 each.
Should I use SKAGs for a thin-margin product? Usually yes. The cost of a bid averaged across terms with different values is asymmetric when margin is thin — overbids become losses rather than smaller profits. On higher-margin products, the automation gained from grouping is generally worth more than the control lost.
Sami Sultan runs Ecom Enable, an Amazon growth partner for DTC brands. Campaign limits, match type definitions and Amazon’s keyword guidance verified against Amazon Ads documentation on 27 August 2026. Rule-based bidding thresholds are corroborated by independent practitioner sources rather than Amazon’s own help page.
Sources: Set up keyword targeting; Sponsored ads campaign limits; Amazon Ads targeting guide; Amazon Ads API release notes and limits reference; Store Growers on the Google SKAG timeline.