Managed Amazon Advertising vs Software: How to Choose, and When a Hybrid Wins

Most brands make this decision under pressure. A contract is up for renewal, a channel manager just resigned, ad spend crossed the point where nobody can review it carefully any more, or the quarterly number came in soft and someone upstairs asked why. So the question gets framed as a shopping question: agency or software? And it gets answered by whichever vendor got the meeting first.

The cost of getting it wrong is not the fee. It is a year of your account history living inside someone else's system, or a seat licence nobody logs into, or a service relationship that quietly became your only record of why the account is structured the way it is. All three are expensive to unwind, and none of them show up on the invoice.

This is a Trellis page, so we have a stake in the answer. We sell software, and we also run accounts for brands who want that. We have written this to be useful even if you pick neither, because the decision turns on two things about your team that have nothing to do with which vendor you are talking to.

Quick answer

Two questions decide this, and neither is about vendors: who owns the standard for your Amazon account, and who has the hours to work it. Answer those two and the model picks itself.

• Managed Amazon advertising. A team outside your company does the execution: campaign builds, bid changes, negations, budget moves, and the reporting that explains them. Right when nobody internally owns the channel and nobody is going to.
• Self-service software. You keep the execution in-house and the platform does the volume work under rules you set. Right when someone owns the channel, has a view on what good looks like, and is out of hours rather than out of judgment.
• Hybrid. You license the platform and buy execution hours on top of it, so the logic, the data, and the change history stay in an account you own while someone else works the weekly pass. Right when you have the owner but not the hours, which is the most common situation and the one both categories are worst at describing.
• If you are already in the wrong one, the move is smaller than it feels. Your campaigns and history live in Amazon rather than in a vendor's tool, so they stay put when you change who works them.

Managed, self-service, and hybrid: three models, three different questions

The category names describe how the work is bought, not what you get. Underneath, each model answers a different question.

Self-service software answers "how do I apply my standard across more decisions than I can review?"

You define the rules, the platform executes them at volume. This works when there is a standard to encode. It fails quietly when there is not: the account gets defaults instead of judgment, and the defaults are the vendor's, not yours.

One thing to check carefully in this category, because vendors are loose with it. Nearly every platform here describes itself as AI. In most cases that means machine-learning bid optimization inside the engine: a model adjusting bids toward a target, which you configure and then leave to run. That is a different product from a copilot you can converse with about the account, ask a question of, and get a reasoned answer back from. Both are useful and they answer different questions. Make the demo show you which one you are being sold, because the two get marketed with the same two letters.

Managed service answers "who is going to do this work?"

An external team takes the hours. The good ones bring pattern recognition across accounts you will never see, which is a real advantage and worth paying for. The structural tradeoff is that you are buying attention out of a shared pool, and you cannot see how that pool is allocated. Most agencies price as a retainer or a percentage of ad spend, which means the fee rises with spend whether or not the return does.

The hybrid answers "how do I keep the standard and rent the hours?"

You hold the platform contract. An execution team, either your agency working inside your instance or the vendor's own team, does the weekly work in it. The account, the logic, the logs, and the history stay yours. When the arrangement ends, you keep the machine.

Amazon itself sells along the same split, which is a useful reference point because the terms are published. Display ads bought through the Ads Console are self-service with no minimum campaign spend and no management fees. The managed-service option through Amazon DSP, where Amazon's own team runs the buys, typically requires a minimum spend of $50,000 and varies by country. Same inventory, same auction, two very different commitments. If you are weighing that specific decision, we broke the economics down in how much Amazon DSP costs.

Two variables decide it: who owns the standard, and who has the hours

Call it the owner-hours grid. Every other input, spend level, catalog size, category, marketplace count, matters less than these two.

Owner means a specific named person who has a view on what a good outcome looks like for this account and the authority to hold to it. Not someone who receives the report. Someone who would notice if the report was wrong.

Hours means recurring, protected time to work the account: reviewing search terms, moving budgets, reading what changed and deciding whether it should have. Not the hours someone theoretically has. The hours that survive contact with a Tuesday.

Software without an owner is a subscription. An agency without an owner on your side is a black box with a monthly call. The grid below is how the four combinations usually resolve.

When a managed service is genuinely the better answer

We would rather say this plainly than have you find out six months in.

When there is no owner and hiring one is not on the roadmap. This is the clearest case. Software applies a standard. If nobody is going to author that standard, an outside team who will is worth more than any feature list.

When the work you need is not advertising execution. Creative production, listing photography, A+ and brand store design, catalog cleanup, international expansion paperwork, Vendor Central terms, escalation of a suppressed ASIN. These are service problems. A platform can flag them. It cannot do them.

When your ad program is small relative to a platform's fixed cost. A subscription priced to your account scale is straightforward, but if your total program spend is modest, a fractional specialist or a small agency retainer can be the cheaper path to competent management. Run that arithmetic before you run a demo.

When you are rebuilding from a mess. An account with years of accumulated campaign sprawl needs a human pass first: consolidation, naming, a structure that makes sense. Automation applied to a broken structure gets you to the wrong place faster. Our own PPC account audit workflow is written to be run by a person before anything gets scaled, for exactly this reason.

When you want a partner who carries the pattern. Good agencies see fifty accounts in your category. That cross-account pattern recognition is real, and it is not something you get from a tool that only sees you.

When Amazon advertising software alone is enough

The opposite case is just as specific.

You have someone who owns the channel and has a view. Your problem is that there are more decisions in a week than that person can work through carefully, so most of them get a default. That is the classic software case, and it is the one where a platform pays for itself fastest: bid moves across hundreds of targets, budget pacing that does not wait for Monday, search term harvesting and negation on a schedule rather than when someone remembers.

Software is also right when you want the record. Every change logged, every rule inspectable, the account history sitting in a system you hold the contract to. When performance dips, the question "what changed, and who changed it" has an answer you can look up rather than ask for.

And it is right when you are consolidating. Running Amazon through one agency, Walmart through another, and pricing through a third means you are the integration layer between them. That is the job the tools were supposed to remove.

If you are at the shortlist stage, we keep a running comparison of Amazon PPC automation software that includes tools we compete with.

The hybrid is the answer most teams are looking for without a name for it

The reason "agency or software" is a bad framing is that the most common real situation sits between them: there is an owner, that owner has a clear view, and that owner has about four hours a week for a channel that needs fifteen.

A hybrid puts the platform contract in your name and the execution hours on top of it. Done properly, three things stay with you that a pure service arrangement does not leave behind:

• The account. Campaign structure, targeting history, and performance data live in an instance you control, not in a vendor's internal tooling.
• The logic. The rules being applied are visible and editable by you, so the standard is documented rather than resident in one person's head.
• The change record. When the number moves, the log of what changed and when is yours to read without filing a request.

The risk in a hybrid is the handoff line. If it is not written down, both sides assume the other is watching budget pacing. Fix that on day one: name who reviews what, on what cadence, and who has approval authority for changes above a threshold you set. If you have never written a handoff document like that before, our budget pacing SOP is a template you can copy and edit rather than draft from a blank page.

What each model keeps and what it gives up

What it costs to stay in the wrong model

Most people reading this are not choosing from scratch. They are three quarters into an arrangement that is not working, and the reason they are still in it is that changing feels like a project and staying feels free. Staying is not free, and the bill arrives in a specific order.

Month one to three, you lose decisions. The weekly pass does not happen, or it happens without judgment. Search terms accumulate, budgets sit where they were set, and a handful of campaigns quietly become the account's whole spend profile. Nothing breaks. The account just stops being steered.

Month four to six, you lose the explanation. Someone asks why a number moved and there is no record to consult, only a call to schedule. This is the point at which most teams realise the thing they were buying was not execution, it was an account they could answer questions about.

Month seven onward, you lose the option. The knowledge of why the account is structured the way it is now lives entirely outside your company, and the longer it stays there the more a change looks like a rebuild. That perception is the real lock-in, and it is mostly a perception, which the next section is about.

What moving between models actually looks like

The reason people stay put longer than they should is that the move feels like starting over. It generally is not, and the reason is structural: your campaigns, targeting, performance history and search term data are records held in your own Amazon advertising account. An agency or a platform works in that account through the API or the console. Change who works it and the account does not move. Its history comes with it because it never left.

Leaving an agency does not mean starting the account over

This is the fear worth naming directly, because it keeps people in relationships they have outgrown. The campaigns are yours. The targeting history is yours. The search term data, the impression history, the conversion record: all of it sits in your advertising account and stays there.

What can genuinely leave is the part nobody wrote down. Why the account is structured the way it is. Which SKUs get protected and why. What the thresholds were, and what the last three seasonal patterns looked like. Collect that before you give notice rather than after, and ask for it as documentation rather than as a conversation. Any professional agency will provide it, and how readily they do is itself a useful signal.

The practical sequence for going self-service is shorter than most people expect: connect the account, then spend the first two weeks reading rather than changing anything, then encode the two workflows that consume the most of your week, then widen from there. The implementation is mostly a connection and a conversation about your standard. The longer part is trust, and trust comes from watching it run against decisions you would have made yourself.

Going from software to hybrid is the lightest move of the three

Because nothing migrates. The instance is the same, the logic is the same, the change log continues uninterrupted. You are adding hours to a setup that already exists rather than replacing anything. If you bought a platform and it is under-used, this is almost always cheaper and faster than swapping it for a different platform that will also go unused.

If you are staying put for now, the work that pays off either way

Four things are worth doing whether or not you move, and doing them is what makes a future move small:

• Get the account structure documented, in your own file, in plain language.
• Get historical data exported on a schedule rather than on request.
• Write the standard down: what good looks like, which SKUs are protected, what the thresholds are.
• Name the owner, even if that person only has four hours a week.

None of that requires a vendor conversation, and all of it turns a rebuild into a connection.

The five things that stop people, and what they are actually worth

These come up in nearly every one of these conversations. All five are reasonable. None is as expensive as it feels from the inside.

"We will lose our account history." You will not, for the reason above: it lives in Amazon rather than in your vendor's tool. What is worth protecting is the undocumented reasoning, and a week of asking questions before you give notice covers it.

"Nobody has time to learn a new tool." This one is real, and the answer is to refuse the full rollout. Pick the two workflows eating the most hours, run only those for a month, and widen once they are saving time rather than costing it. Teams that try to adopt everything in week one are the teams whose platforms go unused, and that is a scoping decision rather than a product problem.

"We are mid-contract." Then use the time rather than waiting it out. Documentation, exports, and writing down the standard all happen in parallel with an existing arrangement and cost nothing. Teams that arrive at renewal having done that work choose in a week. Teams that arrive having done none of it renew by default, which is how most wrong models get a second year.

"We tried a platform before and nobody used it." Worth diagnosing rather than repeating. Almost always it was one of three things: no named owner, too much scope at once, or no protected time. A different platform fixes none of those. A hybrid fixes the third directly, which is why it is the usual answer for teams who have been burned once.

"What if we pick wrong again." The two variables that decide this, an owner and the hours, move slowly, so the answer is stable for longer than a contract term. And the cost of being wrong drops sharply once the account and the logic are in your name, because the next change is a change of hands rather than a change of systems. Ask any vendor what the path looks like in both directions before you sign. A clean answer to that question tells you whether they expect to keep your business by being useful or by being difficult to leave.

Questions to ask before you sign either one

The answers to these separate vendors more reliably than a feature demo.

1. If we leave, what do we keep? Campaign structure, historical data exports, the rules themselves. Get it in writing.
2. Who does the day-to-day work, and how many other accounts do they carry? Ask for the name and the number. A reasonable answer exists; a defensive answer is information too.
3. What is the notice period, and what triggers renewal? Auto-renew with a long notice window is where the real lock-in sits.
4. Show me the change log for a real account. Redacted is fine. If there is not one, ask how they would answer "what changed last Tuesday" six months from now.
5. When you say AI, do you mean bid optimization or a copilot? Machine learning tuning bids toward a target and an assistant you can ask questions of are different products. Ask which one is on the screen.
6. What happens when we want to bring this in house? A vendor that has a clean path for this is telling you something about how they expect to keep your business.
7. What is explicitly out of scope? Creative, listing copy, catalog fixes, Vendor Central, international. Get the exclusions listed, not implied.

If you are already in the wrong model

Four situations, four different first moves.

If you are in an agency relationship that has gone quiet, do not open with a termination. Ask for the change log and the account structure documentation first. What comes back tells you whether the problem is effort or fit, and you need that documentation either way before you move.

If you bought software nobody uses, the fix is usually not different software. Name an owner, cut the scope to the two workflows that matter most, and set a standing weekly slot. If you cannot protect the slot, you have learned something useful: you need execution hours, not features, and the hybrid is a smaller step than a replatform.

If you are paying a percentage of spend and spend has grown, run the number both ways. A retainer or a fixed platform fee at your current scale may be materially different. It is a reasonable renegotiation to open, and most agencies would rather have that conversation than a departure.

If you are mid-contract and stuck, use the time. Get your account structure documented, get your historical data exported on a schedule, and get the standard written down in plain English. Whatever you move to next, that document is the asset.

Where Trellis fits

The break point we see most often: a brand has a channel owner with a clear view, buys a platform to apply it, and then discovers that nobody has the fifteen hours a week it takes to work the platform properly. The standard exists, the capacity to run it does not. Buying a second agency on top of the tool recreates the split you were trying to close.

Trellis sells both sides of that, deliberately. Self-service is the platform on its own: Qinetix, our execution layer, runs continuous ads automation with selectable optimization logic per campaign, bidding floors and ceilings, change logs, and dynamic pricing where you set the thresholds and the system moves price inside them. Ads and pricing run together in one platform on shared account data, which means one place to set the conditions for both, one change log covering both, and inventory and margin signals available to each. The artifact to ask for in a demo is that change log: a real one, with the rule that fired and the timestamp, is a better test of a platform than any feature list.

Strategic Management is the hybrid. For an additional monthly flat fee, our in-house team takes on the execution: campaign builds and cleanup, bid optimization using the same platform logic you can see, promotions, pricing, and listing work done with your guidance rather than over your head. You get a dedicated Customer Success Manager, regular strategy calls, and extra sessions around tentpole events. The platform contract stays yours, which means the account, the logs, and the history stay yours too. Brands blend the two, and some start fully managed and move to self-service once the team is comfortable in the software. That path is normal here, not an exception we tolerate, and the move itself is a change of hands rather than a migration.

On the fee: it is set once at contract, calculated from either revenue or ad spend depending on the service, and then locked for the duration of the contract. It does not recalculate as you scale, so growth stays with you. The specifics belong in a conversation rather than a blog post, and the question worth asking us, or anyone, is what the number is calculated from and whether it is locked.

If the deeper problem is that your operating standard lives in one person's head, Qore is the part of the platform where a review you currently run by hand becomes a workflow that runs on a schedule and returns the same output from the same inputs. It is an approval log with an automation component: the skill proposes the change, you approve it, and the record stays with the run. Worth knowing if you are keeping your agency: Qore reads, analyzes, and makes an approved change on any account regardless of what runs the bids, so it sits alongside your existing bid manager, or works with no automation platform at all. That also makes it the lowest-commitment way to test whether your problem is judgment or hours, because it does not require you to change who runs anything. It is in open beta and you can sign up here.

The honest limits. Trellis does not produce your creative, write your brand strategy, or negotiate your Vendor Central terms, and our managed team will not decide what a good outcome looks like for your business on your behalf. Competitor product research is another one: Helium 10 and Jungle Scout lead that category and we are not trying to displace them, though keyword discovery from your own running campaigns is a core part of what the platform does. If you want a free starting point, we publish the Trellis Chrome extension. Channel coverage across Amazon, Walmart, Shopify, and TikTok is real, but depth differs by channel, so ask about the specific one you care about. And a managed arrangement with us has the same structural feature as any other: someone on your side still has to read the reporting and push back when it does not match what you expected.

For the narrower version of this question, whether a platform displaces the person or the agency currently running your account, we wrote a separate answer: does Trellis replace my agency or in-house PPC manager. Agencies evaluating the platform side can start at Trellis for agencies.

The decision in one line

Ask who owns the standard and who has the hours, and the model picks itself. No owner means buy a team. Owner plus hours means buy software. Owner without hours, which is most brands, means buy a platform you hold the contract to and rent the execution on top of it. Every other input is a detail you can change later; those two are not.

And if you are already in the wrong one, the move is smaller than it looks from inside it. Document the account, write the standard down, and the change becomes a connection rather than a rebuild. If you want to walk your own situation through the grid with someone who will tell you when the answer is an agency, book a walkthrough and bring your current setup.