Fictional brands, illustrative numbers — each one worked end to end: background, strategy, weekly execution, the dashboard, the results, and what I'd carry into the next launch.
Aurelia Home is a hypothetical premium kitchen appliance brand. The brief: launch a new stand mixer and kettle range on Amazon DE, FR, IT, and ES simultaneously, build Sponsored Ads from zero, own the Brand Store, and report weekly on efficiency.
Aurelia Home had strong retail distribution in Germany but no Amazon presence. Category research showed the top 3 competitors were winning on Sponsored Brand video placements and had significantly stronger Brand Store content than the product photography Aurelia had on hand at launch.
Zero historical data meant every campaign started on estimated bids. Four marketplaces meant four sets of translated copy, four Brand Stores, and four budgets to balance — with a single combined monthly cap, not four independent ones.
Launch broad/auto campaigns for two weeks purely to harvest converting search terms, then migrate winners to exact match with individual bids.
Commission a lifestyle main image before scaling spend — treat creative as a conversion lever, not an afterthought to bidding.
Compare ACoS across Sponsored Products/Brands/Display every week and move budget toward whichever is most efficient, not a fixed split.
Impression Share started at 14%; negative-keyword list grew fast as broad match surfaced irrelevant traffic.
Lifestyle shot vs. plain white background — lifestyle won, lifting unit session % from 9.1% to 12.4%.
Top converting terms moved to dedicated exact-match campaigns with individual bids; broad match scaled back.
Added Sponsored Brands to build Store traffic; Impression Share reached 24%.
Cut Sponsored Display 40%, redirected into Sponsored Products; Impression Share closed at 31%.
| Week | Spend | Ad Sales | ACoS |
|---|---|---|---|
| 1 | €5,800 | €18,500 | 31.4% |
| 2 | €6,100 | €20,800 | 29.3% |
| 3 | €6,300 | €23,200 | 27.2% |
| 4 | €6,000 | €24,100 | 24.9% |
| 5 | €6,400 | €27,300 | 23.4% |
| 6 | €6,200 | €28,900 | 21.5% |
| 7 | €6,600 | €31,500 | 21.0% |
| 8 | €6,800 | €34,000 | 20.0% |
| Total | €50,200 | €208,300 | 24.1% blended |
TACoS held at 11% across the period — advertising drove a healthy but not dominant share of revenue, evidence the organic flywheel was doing real work. NTB orders sat at 42%, confirming the campaigns reached genuinely new customers.
The single highest-leverage change was free — the main image test — not a bid increase. Sponsored Display looked fine in isolation but was the clearest reallocation target once compared against Sponsored Products on ACoS. Negative-keyword harvesting works best as a weekly habit from week one, not a monthly cleanup.
Lumeo Audio's flagship headphones were already ranking well, but ACoS had crept up to 42% as new competitors entered the category with aggressive bids. The brief: protect sales volume while rebuilding efficiency — without a full relaunch.
Lumeo had two years of stable Amazon history and decent organic rank, but had never run Sponsored Brand video — the format competitors were now using to win the headline placement Lumeo used to hold by default.
Volume couldn't drop during the fix — Lumeo's Amazon revenue fed a quarterly forecast the wider business depended on. Every change had to be tested against a control group before being rolled out fully.
Move top 20 converting keywords to exact match with individual bids before touching creative — cheapest lever, fastest to test.
Test video against static Sponsored Brands on a 50/50 budget split to isolate the format's actual lift.
Analyze hour-by-hour conversion data before assuming budget should be flat across the day.
Found broad-match keywords carrying 60% of spend but only 30% of orders — the core of the ACoS problem.
Top 20 keywords moved to individual exact-match bids; broad-match budget capped.
50/50 split against static; video ran roughly 2x the CTR within the first week.
Found a clear evening conversion peak (7–10pm local); shifted bid modifiers accordingly.
Shifted remaining static Sponsored Brand budget fully into video; held Sponsored Display at a capped 15% of spend.
| Week | Spend | Ad Sales | ACoS |
|---|---|---|---|
| 1 | €5,040 | €12,000 | 42.0% |
| 2 | €5,580 | €15,500 | 36.0% |
| 3 | €6,016 | €18,800 | 32.0% |
| 4 | €6,148 | €21,200 | 29.0% |
| 5 | €6,396 | €24,600 | 26.0% |
| 6 | €6,600 | €27,500 | 24.0% |
| Total | €35,780 | €119,600 | 29.9% blended |
TACoS held around 9% throughout, confirming the ACoS spike was a bid-discipline problem, not weakening organic demand. Sponsored Brand video became the highest-CTR format in the account within three weeks of launch.
An ACoS spike doesn't always mean weakening demand — check TACoS before assuming the worst. Broad-match keywords are worth auditing individually the moment ACoS trends upward; they're usually where inefficiency hides. Video format tests are worth running earlier than most accounts do.
Verde & Co had healthy ad efficiency but weak Store performance — high bounce rate and multiple third-party sellers competing for the Buy Box on core SKUs. The brief: fix content and Buy Box health before touching bids at all.
Verde & Co sold through both 1P and unauthorized 3P resellers, which meant multiple offers competing for the same Buy Box — and a Brand Store still using launch-era stock photography two years on.
The ad account looked healthy by every standard PPC metric. The real leak was downstream — Store Insights showed most paid traffic bouncing before it ever reached a purchase decision, which no amount of bid optimization would fix.
Rebuild A+ Content and Store modules first; hold ad spend flat to isolate the real driver of any improvement.
Enroll in Brand Registry protections and tighten MAP pricing enforcement against unauthorized resellers.
Add size/usage comparison modules — the most-requested pre-purchase question in customer reviews.
Bounce rate confirmed at 68%; heaviest drop-off traced to the homepage hero module.
New comparison and usage-scenario modules replaced generic stock photography.
Filed against three unauthorized resellers undercutting price and winning the Buy Box.
Ad spend held flat throughout to confirm the lift came from content and Buy Box, not advertising.
| Week | Store Bounce Rate | Buy Box % | Milestone |
|---|---|---|---|
| 1 | 68% | 61% | Store Insights audit |
| 2 | 58% | 61% | A+ Content rebuild live |
| 3 | 49% | 82% | Brand Registry enforcement filed |
| 4 | 41% | 97% | Ad spend held flat — measurement week |
Sales-per-visitor more than doubled without any change to ad spend or bids — proof the leak was downstream of the click, not in the campaigns. Buy Box recovery alone likely accounts for a meaningful share of the lift, since lost Buy Box sessions convert at close to zero.
Healthy ACoS can hide a broken destination — always check Store Insights before assuming ads need work. Buy Box loss to unauthorized resellers is an easy blind spot on established listings; it's worth a quarterly check even when nothing else looks wrong. Content fixes are cheap relative to their leverage and deserve to be tried before bids.
Bellamare Beauty is a hypothetical premium skincare D2C brand. The brief: audit and rebuild the owned website's structured content and review infrastructure so the brand shows up accurately inside AI-generated answers — coordinating an external CMS/dev agency to implement the technical changes.
Bellamare's owned site ranked well on traditional search but almost never appeared when the same questions were asked of AI assistants — the content was written for keyword match, not for a generative engine to extract and cite confidently.
GEO measurement has no industry-standard tool yet — there's no "GEO Analytics" dashboard equivalent to Google Search Console. The team had to build its own tracking proxy, and every technical change (schema, page structure) required a CMS agency to implement, not an in-house dev team.
Implement product and FAQ schema across the top 40 revenue-driving pages before any content rewrite — the extraction layer has to exist before content quality matters.
Treat the review platform as a GEO asset, not just a CX widget — push structured review coverage on core SKUs as deliberately as any content deliverable.
Brief, steer, and QA the external CMS/dev agency against a clear schema and page-speed spec — the brand team owns the "what" and "why," the agency owns implementation.
Ran a fixed panel of 50 realistic consumer prompts across major AI assistants; Bellamare was mentioned accurately in only 8.
Briefed the CMS agency on product/FAQ schema spec for the top 40 pages; QA'd markup against Google's structured data testing tools.
Coordinated with CRM/CX team to prioritize review requests on core SKUs lacking structured coverage.
Rewrote top 20 pages around specific, verifiable claims (ingredient concentration, clinical test summaries) instead of generic marketing copy.
Mentions rose to 31 of 50; iterated on the weakest-performing product category.
| Week | AI/LLM Referral Sessions | Milestone |
|---|---|---|
| 1–2 | 120 → 140 /wk | Baseline audit, 50-prompt panel run (8/50 mentions) |
| 3–4 | 165 → 190 /wk | Schema implementation with CMS agency |
| 5–6 | 220 → 255 /wk | Review platform coverage push |
| 7–8 | 290 → 320 /wk | Top 20 pages rewritten for content depth |
| 9–10 | 350 → 380 /wk | Prompt panel re-run (31/50 mentions) |
Structured review coverage more than doubled on core SKUs, and the prompt-panel mention rate nearly quadrupled — directional evidence the site became meaningfully more extractable, not proof of a fixed ranking the way a keyword position would be.
GEO measurement is genuinely immature — a self-built prompt panel is an honest, testable proxy, not an industry-standard metric, and that distinction should stay explicit in any report. Schema and reviews compound each other; neither alone moved the needle as much as both together. Agency management is a spec-writing discipline as much as a relationship one — vague briefs produced vague implementations on the first pass.
The launch funnel, the metrics cheat sheet, and the budget rhythm behind every campaign on this site — one free PDF, no signup.
Knowledge base, not live AI — instant answers