The Iceberg Principle: Why the Most Important Amazon Work Is the Work You Don’t See

As it happens, the first several months of a Troveres engagement may look (from the outside) like not much is happening. No campaign overhauls, weekly change log full of bid adjustments, or frantic listing rewrites sent over for approval. A founder who has worked with two previous agencies might glance back at month two and wonder what they’re paying for.

That lack of razzle dazzle is the visible part of the strategy.

When our work began with FilterBaby, they had zero Amazon presence, zero organic rank, and zero review history. Eight months later, after keyword landscape research, competitive gap analysis, review mining, and an advertising strategy calibrated to build organic rank rather than buy temporary visibility, their first holiday season on Amazon generated more than $200,000 in sales. For a lot of that time, you couldn’t see much of the work going on. But when Q4 arrived, it was impossible to miss.

This is what we think of as The Iceberg Principle. What’s visible in most of our work is a small percentage of what’s actually going on. The invisible is what makes the visible, well, visible.

The Amazon agency industry has a structural problem. The dominant model is to diagnose fast, act faster, and show results in 30 to 60 days. It creates pressure to optimize whatever is most visible, typically advertising, without understanding the underlying situation. The result is a cycle of surface-level fixes that never address root causes. Brands churn through agencies every six to twelve months. Each new partner makes the same adjustments. No one diagnoses why performance is actually declining, maybe because the engagement doesn’t give them the space to. 

This article explains what’s happening under the surface during a research-first engagement. We’ll look at what it produced for FilterBaby, what it cost one brand when they left before the payoff arrived, and why our 12 to 18 month model exists. Hint: it’s not as a convenience for us.

The Quick-Fix Trap: Why 30-Day Turnarounds Don’t Work

The standard agency onboarding often looks like: audit the ad campaigns, identify obvious inefficiencies, adjust bids, restructure campaigns, rewrite some listing titles. Show the client an improvement in ACoS or ROAS within the first month, which often feels like progress, but it isn’t.

These are surface-level optimizations that any competent practitioner can make. They don’t address why the account was underperforming in the first place. The issues that drive sustained underperformance on Amazon like wrong-customer acquisition, operational margin erosion, algorithmic misalignment, and compliance vulnerabilities, are invisible from an ad dashboard. An agency that starts optimizing on Day 1 is treating symptoms while root causes compound.

As if there wasn’t enough going on, now there is something new that a percentage of agencies don’t even know to look for yet: AI discoverability gaps. While brands are running campaigns, shoppers are asking ChatGPT and Amazon’s own Alexa which brand to buy. The answer those AI systems return is drawn from off-platform content, review specificity, and listing copy written in natural language, not keyword density. An agency with no view of a brand’s AI discoverability footprint is running an incomplete diagnosis from Day 1.

This is why brands cycle through agencies. Agency 1 optimizes ads for six months. Results plateau. Brand hires Agency 2. Same adjustments, slightly different. Results plateau again. The problem isn’t agency competence. The model itself (short timelines, quick results, surface-level scope) makes it near impossible to do the kind of diagnostic work that produces lasting change.

Expectations, Expectations

Part of what makes this so difficult to communicate is the expectation gap Amazon has created. When brands work with an SEO agency, they arrive pre-trained to expect a six-to-twelve month process. The SEO industry has done the patient work of setting that expectation. Amazon has done the opposite. The dominance of Sponsored Products, where you can run ads today and see clicks tomorrow, has trained an entire generation of sellers to expect instant results from everything on the platform. It isn’t accurate, but it’s deeply embedded. 

The marketing intelligence work of collecting and organizing data, running A/B tests, building AMC lookalike audiences, and identifying who the actual customer is, doesn’t produce revenue in week two. For some brands, the real results don’t surface for six to twelve months. For one successful brand we work with extensively, it took close to a year and a half before the foundational work began producing meaningful, measurable returns. None of that work was invisible because it was unimportant, it’s like that because that’s how foundational work looks from the outside. A lot like building with good old brick and mortar.

There is a subtler version of this trap: what happens when an agency does the right foundational work, the client can see it building, gets impatient before the payoff arrives, and leaves. Unfortunately that is not a hypothetical, it is a documented case study, and we will get to it.

What’s Actually Happening Below the Surface

Early on, we are 100% devoted to understanding the brand and business before recommending any action. The account is not being ignored. It is being studied.

It’s worth being specific about what “below the surface” actually means in practice. When a full stack client works with us, they see two things immediately: catalog work (listing copy, imagery, A+ content) handled by our sister agency Azama, and advertising (Sponsored Products, Sponsored Brands, campaign structure) handled by Troveres. Both are visible relatively quickly. What doesn’t surface for months, and what many of our clients don’t initially realize they’re getting, is the marketing intelligence layer: collecting and organizing competitive data, running A/B tests, building AMC lookalike audiences, researching who the customer actually is at the demographic and behavioral level, and testing hypotheses about what’s driving or suppressing conversion. It’s a ton of effort, and while it’s a solid foundation to keep the business stable and to build upon, none of it announces itself on a dashboard. It shapes every decision that does show up on a dashboard, but it operates out of sight, as foundations often do.

For Example

FilterBaby is the clearest illustration of what this looks like in practice. Before a word of their listing copy was written, Azama spent weeks on customer research: mapping the full keyword landscape for the water filtration category, mining competitor reviews to understand exact buyer language and unmet needs, and identifying competitive positioning gaps.

Three things came out of that research phase that every downstream decision was built on: a search term map, a buyer sentiment profile from review mining, and a competitive gap analysis. Not assumptions, real data. Azama could also see brand search traffic on Amazon before launch: people were already looking for FilterBaby and coming up empty. That was the first diagnostic finding.

The diagnostic engagement covers six areas. Each one is examined before any implementation begins.

Advertising analysis looks not just at campaign metrics but at who the ads are actually reaching, what behavioral signals those customers generate, and whether the current strategy is acquiring the right customers or the wrong ones.

Operational forensics means a full P&L reconciliation: every cost layer, from FBA fees to AWD to storage to routing to chargebacks, to identify where margin is leaking and quantify the gap between perceived and actual profitability.

Competitive and market analysis establishes who’s winning in the category, what strategies they’re using, and how Amazon’s algorithm is currently positioning the brand relative to competitors.

Customer and persona mapping identifies who the brand’s actual Amazon customer is, which may differ significantly from their retail or DTC customer. The FilterBaby avatar, built from keyword research and review mining before a word of copy was written, is a precise example of what this produces.

Compliance review assesses existing vulnerabilities, category-specific requirements, and protections against unauthorized resellers. For FilterBaby, compliance review surfaced a genuine competitive advantage, NSF certification, that most competitors couldn’t match. Compliance review is not just risk mitigation. It can reveal strategic opportunities.

The off-platform AI discoverability audit

This is the newest diagnostic layer. This means running the brand’s top buyer-intent queries through ChatGPT, Perplexity, and Google AI Overviews to establish an AI share-of-voice baseline, and evaluating listing copy for Alexa-legibility. The FilterBaby listing was built in the customer’s own language from the start, which is not coincidentally how Alexa reads and recommends products when a buyer asks a question before touching search.

At the end of the research phase, a comprehensive diagnostic is delivered: what’s actually happening with the brand’s Amazon business, what the root causes of underperformance are, and a prioritized roadmap for the corrective strategy. This is what the first phase is building toward. Not an impressive dashboard but a real diagnosis.

A note on how Troveres uses technology: AI tools are used extensively in the research phase, in order to build and run programs that apply consistent, time-tested analytical procedures, not to produce generic recommendations. Every ASIN is treated as its own input with its own data intact. The diagnostic says “here is what your data shows.” Not “the AI decided this.”

The Compounding Return: Why Months 7–18 Are Where the Value Lives

Months 1 through 6 are diagnostic and foundational. Visible results may be modest. The work done in this phase creates the conditions for accelerating returns in months 7 through 18.

Results compound because each element of the strategy reinforces the others. Correct advertising feeds better algorithmic data. Better organic rank reduces ad dependency. Operational efficiencies improve margin. Compliance infrastructure prevents disruptions. Off-platform content and AI citation authority compound on the same curve: a brand that begins building AI discoverability early will see that investment reflected in Alexa recommendations and external AI citations by months 12 to 18… an asset that grows independently of ad spend.

The most instructive illustration of this curve is what happens when a brand doesn’t stay on it.

The OmniSense story is not an argument that every brand should stay in every engagement indefinitely. It is a precise illustration of what the patience curve looks like from the inside, and why the months that feel like they’re “just building” are the ones the payoff depends on.

A 6-month engagement captures only the diagnostic and early implementation phases. The client sees the investment but not the compounding returns. They leave just before the strategy begins to pay off, and the next agency starts the cycle over. The 12 to 18 month model exists because that’s how long it takes for the full strategic flywheel to engage.

The Foundation Your Competitors Are Too Impatient to Build

The reason some Amazon agencies fail to produce lasting results is structural. The model (short timelines, quick actions, surface-level scope) makes it impossible to diagnose and address root causes. Troveres’ research-first approach exists because doing the work right takes time, and doing it wrong is what brands have already experienced.

FilterBaby had no Amazon presence, no reviews, and no algorithmic history. Eight months of methodical research-driven work produced $200,000 in a single holiday season. That is what the front-loaded phase builds toward.

OmniSense had 11 months of strategic foundation completed and metrics surging. They left one month into the payoff phase. Three months later, organic rank had collapsed across 255 keywords, sales velocity had halved, and every sale was being bought at full advertising cost. That is the cost of leaving before the flywheel engages.

Now’s the time to build the strategic and AI-discoverability foundation that your competitors are too impatient to.

Frequently Asked Questions

What actually happens in the first few months of an Amazon agency engagement?

At most agencies: campaign audits, bid adjustments, listing tweaks, and an improved ACoS on a slide deck. At Troveres: research. The process begins with a full keyword landscape map, competitive review mining, customer avatar construction, P&L forensics, compliance review, and an AI discoverability audit, before a single campaign change is made. FilterBaby’s first four weeks were entirely customer research. That research became the blueprint for a listing that generated $200K+ in its first holiday season.

How long does it take to see real results from an Amazon agency?

Surface metrics like ACoS improvement and restructured campaigns can appear within 30 to 60 days. Lasting compounding results typically emerge in months 7 to 18, after a diagnostic research phase has identified and addressed the root causes of underperformance. FilterBaby’s $200K+ holiday season was the result of roughly eight months of methodical work. It wasn’t a lucky break. It was the predictable outcome of a process designed to build momentum systematically.

What happens if I leave an agency engagement before the 12–18 months are complete?

A brand Azama worked with terminated their engagement at month 11, just as the strategic work was producing its first major performance surge. Three months later, organic rank had declined across 255 keywords and Best Seller Rank had deteriorated from the 25–30K range to 45–55K. Every sale that had been coming from organic discovery was now being bought through paid advertising at full margin cost. The full story is at azamabrands.com/the-breakthrough-that-never-was/

Can a brand with no Amazon presence really compete with established sellers?

FilterBaby proved it. They launched with zero reviews, zero organic rank, and no advertising history. What they had was a listing built on deep customer research: keyword landscape mapping, review mining in adjacent categories, and competitive gap analysis that identified angles established sellers were failing to address. Conversion is driven by relevance. Relevance comes from research. FilterBaby’s conversion rate on one variant reached 5.47% after Premium A+. These are numbers that rival listings with far longer track records.

I’ve already been through two agencies. Why would this be different?

Most agencies optimize around the edges of what’s visible, like ad campaigns, listing titles, and bid structure, without diagnosing why the account is underperforming. If two agencies have made the same category of improvements without producing lasting change, the problem isn’t their competence. It’s the model. Troveres spends the first several months doing mostly research, and delivers a comprehensive diagnostic before making a single change.

Do you use AI in your process, and should I be concerned about that?

Troveres uses AI tools extensively, but to build and run programs that apply consistent, time-tested analytical procedures, not to substitute for judgment or produce generic recommendations. Every ASIN is treated as its own input with its own data intact. You won’t receive a recommendation that says the AI decided this. You’ll receive an analysis that shows exactly what the data found and why the recommended action follows from it.

What should I look for in an Amazon agency?

Start with how they spend the first 90 days. If the answer involves immediate campaign changes and quick-win optimizations, ask what diagnosis those changes are based on. The structural issues that drive sustained underperformance (wrong-customer acquisition, margin erosion, algorithmic misalignment, compliance vulnerabilities, AI discoverability gaps) are all invisible from an ad dashboard. An agency that starts optimizing before diagnosing is treating symptoms. The right partner starts with research.