Advertising economics
How Amazon Works · C3
PPC as a system: campaign structure, bids, ACoS vs TACoS, dayparting, launch spend — and where AI tooling actually helps.
PPC as a system: campaign structure, bids, ACoS vs TACoS, dayparting, launch spend — and where AI tooling actually helps.
Amazon advertising is not a cost line; it buys the sales velocity Amazon's ranking runs on. Roughly 70 % of a search page is now paid [L97] [SS-03]. The practitioners agree on the shape of the curve: pay dearly at launch (assume ~50 % of revenue back into ads in month one [L19]), then let organic sales dilute the ratio — one launch fell from 21 % to 6.5 % total ad spend in 90 days while averaging 14 % ACoS [L30]. Structure, bids, placements and dayparting are all about spending that money so the ratio actually falls.
The two ratios and the one that pays the bills
- ACoS (advertising cost of sale) = ad spend ÷ ad-attributed revenue — a per-campaign efficiency metric [L19] [L40].
- TACoS (total ACoS) = ad spend ÷ total revenue, organic included — the north star, because it shows whether ads are building organic sales [L19] [L28] [L39].
- Break-even ACoS = margin after cost of goods, inbound shipping, returns, FBA and referral fees ([C6]); a keyword above it loses money on every ad sale [L44] [L47]. Break-even TACoS × trailing-30-day daily revenue is the most you can spend per day without losing money [L32].
- Lifetime ACoS (LACoS) re-runs ACoS over ~12 months of repeat purchases from Brand Analytics; consumables show 20–50 % repeat-purchaser sales, non-consumables ~8 %, and a consumable can justify a 90–100 % headline ACoS if LACoS is low [L41]. A shower head with replacement cartridges sits between the two.
- CAC = ad spend ÷ (total orders − repeat units); Brand Analytics counts "repeat" inside 90 days, the ad console counts "new-to-brand" as 365 days, so do not mix them [L88].
The most repeated warning: never judge PPC by ACoS alone — weigh it against the campaign's goal, listing quality and organic-rank movement [L51]. A keyword at 65 % ACoS was the top campaign of a $1M launch because its job was rank; the month came in at 16.6 % TACoS [L28].
The life of a product: expensive launch, then cash cow
PPC's first job is not profit; it is to prove click-through and conversion on specific search terms so the listing ranks organically for them [L56] [L93]. Hence the curve:
- $200+ home product: $0 → $100K in 90 days, 14 % average ACoS, TACoS 21 % → 6.5 % at ~$12K/week — "an expensive launch turning into a lean cash cow" [L30].
- Supplement: $1M in 12 months, ~$150K/month, its single-keyword ranking campaign at 65 % ACoS on purpose [L28].
- Flash-card brand: ad spend $5K → $8–9K/month while revenue quadrupled; a later "harvest" phase cut spend 10 % to test how much rank held [L39].
- Niche supplements: 60 → 600 units/day in a year by feeding ad data back into the thumbnail and listing, not only into bids [L35].
Underneath: a new listing has no sales history, so it sits past page one where 70 %+ of sales happen; the auction puts it on page one from day one, and a short "honeymoon" rank boost starts when inventory goes live — so listing content must be finished before launch [L44] [L45]. Rank is worth buying: organic #1 takes ~25 % of clicks and ~19 % of sales, #2 ~12 %/8 %, #3 ~8 %/5 % [L07]. Ads never go to zero afterwards, only lower [L51].
Where sources disagree: aggressive vs lean launch. Crescent Kao overbids Amazon's suggested range by ~$1 with odd-penny amounts, runs "dynamic up-and-down", and inflates the daily budget ($100 nominal to spend ~$20) to win top placement fast [L44] [L45] [L47]. Chris Rawlings derives bids from price × conversion × target ACoS, defaults to "down-only", treats budgets as a planner output, and says a lean launch shifts money toward product targeting and video, an aggressive one toward single-keyword ranking campaigns [L40] [L32]. My Amazon Guy adds the caveat both skip: with no reviews you often cannot convert on the "obvious" exact keyword at top of search, so aggressive spend there is a trap [L85]. For a small-budget single SKU, trust the formula path; overbidding is a data-collection tactic for sellers who can absorb ~50 % TACoS in month one [L19].
Campaign architecture: what each campaign is for
Only two placements exist — search results (top / rest of search) and product detail pages — and keyword vs product targeting merely signals intent; Amazon shows either in both [L40] [L26]. Every campaign carries one of four goals: performance (profit), ranking, brand defence, research [L29] [L19]. Hygiene the sources converge on: one product per campaign [L40] [L31]; one match type or targeting group per campaign [L24] [L45]; keywords split by search-volume tier so a head term cannot eat the budget [L51] [L66]; a naming template (product · ASIN · ad type · match) [L40] [L23]; and remove Amazon's hidden "keywords related to your product category" add-on from manual campaigns [L51].
| Campaign | Job | Setup the sources agree on | Notes |
|---|---|---|---|
| Auto — close match | discovery + rank at launch | "set bids by targeting group", one group per campaign [L24] [L56] | search-term report within 24 h [L56] |
| Auto — loose / substitutes / complements | discovery | separate campaigns; close + substitutes usually best [L24] [L30] | loose match won one $1M launch [L28] |
| Exact — single keyword | rank one alpha term | fixed or down-only, top-of-search modifier [L28] [L33] | judge by rank, not ACoS [L29] |
| Exact — multi-keyword by intent | profit | 4–10 intent clusters, down-only [L25] [L29] [L30] | per-cluster CTR diagnoses the listing [L25] |
| Broad (+ phrase negatives) | controlled discovery | same roots as exact; "+word" locks a term [L34] [L23] | best ACoS of a top-10 set [L28] |
| Category (refined) | wide, cheap reach | own subcategory; refine by price > yours OR rating < yours, one axis [L22] [L28] | lowest-ACoS campaign on one account; 12 % ACoS/$21.7K on a launch [L22] [L30] |
| Product targeting | cheap clicks, competitor pages | 6–12 ASINs: high-volume, weaker offer, missing feature, self [L26] [L28] [L23] | avoid as the ranking vehicle [L51] |
| Brand defence | block competitors on own pages | own ASINs, own brand terms [L26] [L30] | see disagreement below |
Phrase match is the least important type: Kao drops it because broad covers it [L44]; Rawlings keeps it only for semantic-family campaigns and says the "+word" modifier makes it "much less necessary" [L34] [L30]. Exact is no longer strictly exact — Amazon swaps words since 2024–25 [L34].
Harvesting makes the structure work: mine the search-term report, negate junk, move terms with 3–5 sales (1 for high-LTV brands) into exact campaigns where bids and placements can be controlled [L66] [L40]. Give a non-exact term ~20 clicks before negating unless obviously irrelevant [L51]; never negate on 50–100 % ACoS in the first days [L56].
Where sources disagree: auto campaigns. Helium 10's auditors cap auto at ~10 % of budget, harvesting only, because the head terms it surfaces are unwinnable [L69] [L72]. Rawlings says autos "once dismissed as money pits" now sit among the top-five campaigns of his launches when actively pruned, and runs a four-bid ladder (30/40/50/60¢) to find the sweet spot [L30] [L24]; Darren shows the same term can be cheaper via auto than exact because fewer sellers bother [L56]. Both fit different stages: at launch auto is discovery — small, close-match only [L56]; once negatives are mature, let its share float on performance [L19].
Where sources disagree: branded spend. Auditors call bidding on your own brand name "not incremental" [L69] [L72]. Two operators defend it: a greeting-card brand runs branded campaigns for "easy volume" and catalog-wide protection [L64]; a $10M/month tracker brand outbids cheaper rivals on its brand and inflates CPC on core terms to own share of voice, ignoring ACoS [L61]. Reconciliation: defence pays once competitors are actually buying your brand terms or your product page; a new single SKU nobody searches for yet has nothing to defend [L66].
Bids, budgets and placements
Bid formula. Starting bid = price × conversion rate × target ACoS; conversion rate is "unit session percentage" in Business Reports → ASINs [L40] [L19]. Use Amazon's suggested bid only as a CPC proxy when there is no history, never as the bid [L32]. Auto campaigns: start ~50p–£1, raise in 10–20p steps if the day's budget goes unspent; below £10/day you get no usable data [L56].
Budget formulas. New launch: daily budget = (CPC ÷ conversion rate) × target daily sales, CVR assumed 5–15 % or taken from category data in Search Query Performance, target daily sales near page-one competitors [L32]. Existing product: monthly budget = target TACoS × target revenue [L32], or daily = units/day × price × target ACoS [L40]. Then the "bombshell": with correct bids you rarely hit daily caps, so spend control lives in bids — some managers set $10,000/day budgets [L32] [L23]. The Budgets tab's "average time in budget" picks the lever: low ACoS at ~100 % time in budget → raise bids; low ACoS at 49 % → raise the cap [L32] [L66].
Bidding strategy. Down-only is the default for Rawlings, Darren and Kao-on-a-budget [L40] [L29] [L51] [L22]; up-and-down only to unstick impressions, then revert (a $6 down-only bid ≈ a $3 up-and-down bid, since Amazon may double it) [L29] [L24]; fixed bids for single-keyword ranking campaigns and to "force your way" onto competitor pages [L29] [L26] [L64].
Placement modifiers. Top of search has the highest CTR, conversion and ranking effect (9 % CTR vs 1.3 % rest-of-search vs 0.2 % product pages in one account) [L28] [L33]; product pages is the cheapest, lowest-volume placement [L33]. Leave all three at 0 % for ~2 weeks [L51] [L56], then start top-of-search at 20–40 %, not the 900 % that used to work [L33] [L44]. To cut a placement (modifiers cannot go negative), lower the base bid 10–20 % and raise the others to compensate [L33] [L44] [L56]. Audience adjusters ("high likelihood to purchase") at 20–30 % only on mature campaigns [L22] [L19]; the business-buyer modifier is a 2025 addition (US — verify for .ae) [L33].
Cadence. Run 7–10 days before touching anything [L44] [L45]; review last week / month / three months to separate noise from trend [L66]; "buy the data" — set a spend threshold per keyword, then keep or kill [L85]; a zero-sale keyword is dead once spend passes ~3× per-unit profit [L44] [L47]. Triage red (pause) / yellow (lower bid) / green (move to a dedicated, better-funded campaign) [L44] [L47], diagnosing in funnel order — impressions → CTR → conversion → ACoS — because ACoS is the outcome, not the cause [L44].
Dayparting: real, but last
Dayparting means rule-based bid changes by hour or weekday. The honest record: it tripled sales on one struggling product, roughly halved ACoS on another, and on a third brand sales fell after it started [L27]. Method: two 14-day hourly exports (Amazon's cap), pivot by hour, colour-code ACoS/CTR/CVR/CPC, apply moderated changes (+15 % to +60 %, deeper cuts on dead hours) only on campaigns older than a month, re-optimise every 3–4 weeks, never in the same week as bid changes [L27] [L19]. Amazon's native Budget Rules can only increase budget in a window; decreases need AdLabs, Adtomic or Perpetua [L27] [L19]. Do not toggle campaigns on/off by hour — it starves Amazon of data [L27]. And the widely quoted "49 % → 24 % ACoS in 36 hours" was not dayparting: it was a main-image change found by feeding Search Query Performance and search-term reports to Claude, lifting CTR 0.8 % → 1.8 % and conversion 19 % → 30 % — you cannot bid your way out of a margin problem [L17] (see [C2]).
Beyond Sponsored Products: Brands, Display, DSP, AMC
What a Sponsored Brands takeover buys is visible on SHIO's own head term: the leader's banner owns the screen before the first organic result appears [SS-03].
| Ad type | What it is | What the sources say works | Cost / access |
|---|---|---|---|
| Sponsored Brands (SB) | banner + video in search, bottom of product page | video should take 90 %+ of SB budget; single-product video straight to the product page; goal "drive page visits"; audience adjusters 15–30 % [L23] | Amazon's AI video generator "actually works", its AI images do not [L23]; layer in after ~2 weeks of static data [L51] |
| Sponsored Display (SD) | contextual, remarketing, audiences; no keywords | CPC not vCPM; start contextual + remarketing, test 30/60/90/180-day lookbacks side by side; default thumbnail; ACoS 10–36 % [L31] | views-remarketing for >$50 items, purchase-remarketing for consumables [L19] |
| Amazon Marketing Cloud (AMC) | 25 months of shopper behaviour; custom + lookalike audiences | feed cart-abandoner / high-converter audiences into Sponsored Products; a premium clothing brand cut ACoS 100 % → ~30 % [L97] | free, access via rep — US — verify for .ae [L97] [L19] |
| DSP | programmatic, on and off Amazon | judge by new-to-brand reach, not profit; retargeting-only DSP duplicates cheaper SD [L97] | $3–5K/month via agency seat (was $50K direct) — US — verify for .ae [L97] |
The logic for SB video and SD audiences: competitors crowd the keyword auction and leave audience and video placements cheaper [L68] [L72]. Sponsored Brands realistically wants 3+ SKUs [L45], but single-product video now goes to one product page and still shows at top of search [L23].
Where the money leaks: the traps list
- Over-committing to exact-match top-of-search on the "obvious" keyword with no reviews — it will not convert [L85].
- Too many keywords in one campaign; broad/auto left as the profit engine instead of a discovery engine; winners never harvested [L66] [L69].
- Judging by ACoS alone, or by a 30-day blended average that hides a recent turn [L51] [L93].
- Spending past a keyword's rank ceiling — if higher-ranked competitors convert better on that term, more spend cannot pass them [L93]; bid-automation software silently reverting manual changes [L93].
- Ads left on during tent-pole events without a deal, when CPCs spike and non-deal listings convert poorly [L85] (US — verify for .ae); and testing as if clicks still cost "two cents" — they do not, so targeting must be intentional [L85].
Where AI helps in PPC (the rest is [C7])
Two uses are PPC-specific and shown on real accounts. First, upload the 60-day Sponsored Products search-term report to Claude: it collates a keyword across campaigns (the old pivot-table job), ranks keywords by revenue with ACoS/ROAS/CVR, filters "efficient" ones by a minimum order count, flags high-spend/high-ACoS terms to negate, and groups terms by shopper intent — then save it as a reusable skill [L18]. Second, feed Search Query Performance plus search-term data to Claude to find the highest-leverage lever — CTR, not bids, in the 49 % → 24 % case [L17]. Rule-based automation exists too — Helium 10's bid rule raises a bid 10 %/day after 7 checks outside sponsored top-3 and lowers it 10 %/day once organic rank hits top-5 [L64], and Helium 10 Ads claims hundreds of targets in ~5 minutes a week [L69] — but the failure mode is automation overriding you [L93].
What this means for SHIO
Assume a single ASIN at retail price P (AED), no sales history, no Brand Analytics yet, small budget. Ratios come from the sources' formulas; replace them with real numbers after week one.
Starter structure (six campaigns, week 1).
| # | Campaign | Targets | Bid strategy | Starting bid | Share of daily budget | Source |
|---|---|---|---|---|---|---|
| 1 | Auto — close match | Amazon-chosen | down-only (switch to up-and-down only if no impressions) | ≈ 2.5–3.5 % of P | 20 % | [L56] [L24] |
| 2 | Auto — substitutes | Amazon-chosen | down-only | same | 10 % | [L24] [L30] |
| 3 | Exact — alpha keywords | 3–5 highest-intent terms (filtered shower head / shower filter / hard-water shower head / high-pressure shower head / vitamin C shower filter — validate volumes) | down-only; fixed later for the one ranking term | P × CVR × target ACoS ≈ P × 0.08 × 0.35 | 30 % | [L40] [L25] [L28] |
| 4 | Broad — same roots | same seeds, phrase negatives (e.g. "replacement filter", "handheld" if not offered) | down-only | 10–20 % under exact | 15 % | [L34] [L30] |
| 5 | Category — own subcategory, refined | rating < SHIO's OR price > P (one axis) | down-only, no modifiers | ≈ exact bid | 15 % | [L22] [L28] |
| 6 | Product targeting | 6–12 competitor ASINs: page-one for the alpha term, priced above P or rated lower | down-only, exact + one expanded twin | ≈ 70 % of exact bid | 10 % | [L26] [L23] |
Bid ratio: with an assumed 8 % conversion (inside the 5–15 % launch range [L32]) and a 35 % target ACoS, the formula gives 2.8 % of P [L40]; sanity-check against Amazon's suggested-bid range as a CPC proxy [L32]. Daily budget = (CPC ÷ CVR) × target daily sales [L32]: at CPC ≈ 3 % of P and CVR 8 % each ad sale costs ≈ 37 % of P, so three sales a day needs a total daily budget of roughly 1.1× P; a launch-month TACoS near 50 % is normal [L19]. Break-even ACoS comes from the fee stack in [C6]; only campaign 3's ranking term may run above it [L28] [L44].
Weeks 3–6, add: one Sponsored Brands single-product video to the product page on the 3–5 converting terms [L23] [L68]; one Sponsored Display views-remarketing campaign with 30/60/90-day lookbacks in one ad group [L31]; a self-targeting (own ASIN, expanded) campaign as cheap retargeting [L26] [L28]. Branded and defence campaigns only once branded searches appear in the search-term report [L66] [L69]. When replacement filters get their own ASIN, target each from the other (defence + repeat) [L26] [L41].
Measure weekly (one sheet). Spend, ad sales, total sales → ACoS and TACoS trend [L19]; time-in-budget per campaign [L32]; search-term report → negatives (after ~20 clicks or obvious mismatch), harvest at 3–5 sales [L51] [L66]; CTR and CVR per intent cluster vs account average — low CTR is a main-image problem, low CVR a listing/price problem [L19] [L25]; organic rank for the alpha terms [L51] [L64]; always in funnel order [L44].
Do not, in month one: touch placement modifiers [L51] [L56]; daypart [L27]; negate on ACoS under ~20 clicks [L51] [L56]; change anything inside 7–10-day windows [L44]; run branded or defence campaigns [L69]; buy bid automation [L93]; run DSP/AMC (US — verify for .ae; premature anyway) [L97]; pour budget into "filtered shower head" exact at top of search before reviews exist [L85] ([C4]); accept Amazon's suggested bid or budget as-is [L40] [L32]; try to fix low conversion with higher bids [L17].
The concept graph behind this chapter
While building the corpus, lore filed the ideas above as pages of a knowledge graph — each link opens that concept's own page, with its definition, the videos it came from, and how to apply it. The 36 most relevant of the subject's 385 published concept pages:
Day Parting (Amazon PPC) · Amazon Negative Keywords · Amazon PPC Auction Model · AWD (Amazon Warehousing and Distribution) · Amazon Seller 10-Step Launch Process · Amazon DSP (Demand-Side Platform) · "Amazon Escape" 23-Criteria Product Funnel · Amazon as the True Bottom of Funnel · Fixed Bids (Amazon PPC Bidding Strategy) · Advisor-Led vs. DIY Amazon Brand Sale (Fee Alignment & Risk) · TACOS vs. ACOS · Amazon Hourly Sponsored Ads Report Export · Branded Search-Lift & New-to-Brand Attribution for Off-Amazon Marketing · Amazon Brand Storefront as Ad-Free Conversion Destination · Dynamic Bids Up and Down (Amazon PPC Bidding Strategy) · Amazon Budget Rules (Native Day-Parting Tool) · Amazon Product Life Cycle (Launch, Expansion, Harvest) · Amazon Gated-Category Documentation Requirement · Five Reasons Amazon FBA Products Fail · Amazon Virtual Bundles (Buy Box Cross-Merchandising) · Six-Campaign PPC Launch Structure · Auto Campaign as ~10%-of-Budget Keyword-Harvesting Cap · Negative Product/Brand Targeting · Single Keyword Exact Match Campaign · New-Launch Ad Budget Formula (CPC ÷ CVR × Target Sales) · Amazon-Exclusive SKU Variants for Meta Ad-Account Protection · Daily Budget Formula (Target Units × Price × Target ACoS) · Amazon Creator Connections (Influencer Affiliate Program) · Campaign Planner (Pre-Launch Budget Worksheet) · One-Product-Per-Campaign Structuring Principle · Amazon-Only vs. Omnichannel Seller Skillset Framing ("Inside vs. Outside the Box") · Order-by-Spend Campaign Audit · Specialist-Partner Team Structure (Sourcing + Meta + Amazon) · PPC Starting Bid Formula (Target ACoS × Price × Conversion Rate) · Placement Bid Adjustments (Top of Search / Rest of Search / Product Pages) · Multi-Keyword Ranking Campaign (Broad Match)
Source legend
Not covered / open questions
- Every figure is US (UK for Darren): CPC levels, conversion rates, category targeting, SB video, SD audiences, AMC/DSP eligibility and the business-buyer modifier all need checking on amazon.ae; the corpus has no UAE PPC data ([C8]).
- No benchmark exists for shower-head or home/personal-care CPCs; the 2.5–3.5 %-of-price bid and 1.1× P daily budget are formula outputs, not observations.
- LACoS needs Brand Analytics, i.e. Brand Registry — status unknown for SHIO.
- Whether the "algorithm wants leeway" shift (broad, category, auto beating exact) holds on a smaller, less crowded marketplace is untested; and no source runs a controlled test of six campaigns against two at small budgets — the authors are agency operators with a stake in complexity.