Most Amazon advertising performs a very simple transaction: spend a dollar, get back roughly the same dollar plus or minus a margin. The campaign runs while spend runs. The moment spend pauses, traffic collapses. The brand owns nothing it didn't already own at the start of the campaign. This is renting traffic. It works as a tactic. It doesn't build an asset.
The alternative — what experienced Amazon operators actually do — treats advertising as a tool to engineer permanent organic rank improvement. The mechanism that converts paid spend into permanent position is the relationship between three levers: TACOS, deals, and BSR. Wire them together correctly and the flywheel turns. Run them in isolation and you stay in the rental market.
The two ways to read a TACOS number.
TACOS — Total Advertising Cost of Sales — divides total ad spend by total sales (paid plus organic). A typical reading: "our TACOS is 25%, that's roughly category-average, the business is working." That reading is incomplete.
The more important question is the direction TACOS is moving. A stable TACOS at scale means paid and organic are scaling together — the brand is renting the rank it has. A declining TACOS at scale means organic share is increasing faster than paid spend — the brand is building permanent position. A rising TACOS at scale is the opposite signal: ad spend is being asked to do work that organic should be doing.
For most growing Amazon brands, the TACOS trajectory over a 12-month period tells more about the business than the absolute number. The flywheel is turning when TACOS is going down and revenue is going up — simultaneously, sustainably, across multiple SKUs.
How deals actually build rank.
The mechanism most Amazon agencies under-use is the deal — specifically Best Deals and Lightning Deals, sequenced as a flywheel rather than treated as discrete promotional events.
A correctly-sized deal does three things at once. It produces a velocity spike — the unit-throughput rate during the deal window is high enough that Amazon's ranking algorithm registers the SKU as a category leader. It produces a glance-views spike — the SKU appears on the Today's Deals page, which is one of the highest-traffic surfaces on Amazon. It produces a conversion-rate spike — discounted price plus high traffic plus algorithmic surfacing converts at a much higher rate than the same listing would on a normal day.
Each of those spikes, on its own, would fade when the deal ends. Together, they teach the algorithm that the SKU deserves higher organic rank — and the rank gain persists after the deal closes. The deal is the discount. The rank gain is the asset.
The flywheel mechanics that follow: higher organic rank → more impressions on category search pages → more organic sales → more velocity → algorithm reinforces the rank → cycle repeats. Each turn of the wheel adds a small amount of permanent position. Twelve consecutive months of well-sequenced deals can shift a SKU from page two to top-of-page-one in its primary category — without a corresponding increase in paid spend.
The wiring between ads and deals.
Ads and deals don't work as separate workstreams. They support each other when sequenced correctly.
During a deal window, paid advertising should accelerate, not pause. Higher bids on the SKU during the deal capture the surge of traffic Amazon is sending. The combination of deal pricing + paid amplification + algorithmic surfacing produces conversion volumes that bake into the SKU's permanent rank. Most agencies cut ad spend during deals to "protect margin" — exactly backwards. The deal is when paid ads do their highest-leverage work.
Between deals, paid spend should optimise against the converters revealed during the deal window. Search Term Reports from the deal period surface the keywords that convert at the highest rate when the SKU is discounted. Those same keywords convert respectably on normal pricing too — and they should be promoted to exact-match targets and bid against, not just on deal weeks.
The deal teaches you which terms convert. The ads exploit that knowledge year-round. The result: an advertising program where every dollar is informed by what actually works, instead of bidding on category-typical keywords that look good in a planning document.
A concrete case: Compostic.
Marketplace OS engaged Compostic — the leading compostable bag brand on Amazon US by parent-level revenue — at a point when conventional Amazon advertising was producing diminishing returns. Monthly sales had grown 41% year-on-year in March 2026, but net profit had fallen 16% over the same period. TACOS was rising. The brand was buying growth.
The intervention was a coordinated flywheel restructure. Deals were sequenced for BSR — each event sized against the cooling period and inventory cover, priced at Amazon's max-eligible (not deeper). Ad spend during deals accelerated rather than retreated. Converting terms from STRs were systematically harvested and promoted. Bleeders were paused weekly. Multi-unit promotions lifted AOV at every checkout.
The result in one month: sales up 25.8%, net profit up 15.4%. TACOS held flat despite a lower average selling price across the catalog. Organic rank improved. Sales velocity material. The same product. The same marketplace. A different operating discipline.
The structural prerequisites.
The flywheel can't be retrofitted onto every Amazon account. Three structural conditions need to be true for it to turn:
- Margin floor — every deal must clear the per-unit contribution margin floor net of Amazon fees, deal fees, advertising, and shipping. Deals priced below floor compound the wrong direction. Most agencies don't have visibility on this.
- Inventory cover — a Best Deal needs 21+ days of stock at 2.5× normal velocity. A stockout mid-deal destroys the rank gain it would have produced. Inventory discipline isn't separate from deals — it's a deal precondition.
- Ad structure — campaigns must be organised by intent and conversion rate, not by category convention. Bid changes must be authorised and executed weekly, not monthly. The Search Term Report has to be read.
None of these conditions are difficult. They're just disciplines. The agencies that don't run them aren't bad — they're optimised for hours billed, not for compounding rank. The fractional executive model exists for exactly this kind of multi-lever, weekly-rhythm work.
The flywheel is the operating thesis. The TACOS number tells you whether it's turning. The deal is the event that produces the permanent gain. None of them work alone.