Unlearning Amazon PPC — Target-Level, Hour-by-Hour Bidding with Patrick Butscher

Independent PPC consultant (and active seller) Patrick Butscher joins AMALYZE founder Christian Kelm to talk about unlearning Amazon Ads habits. They dive straight into target-level, hour-by-hour bidding on Amazon Marketing Stream (AMS) data, why placement modifiers mislead, how attribution beyond 7 days changes the math, and the mindset required to let software do what humans can’t.

Start here: ads won’t fix a broken offer

Christian opens with a reality check: if your product doesn’t retail — low CTR, weak reviews, uncompetitive price — ads can’t save it. A CVR of 0.5% won’t jump to 50% because you turned on PPC. Patrick agrees: PPC only scales what already works.

“If your product is bad on Amazon, advertising won’t change that in any meaningful way.” — Christian Kelm

They revisit this again when variants or repricers distort your average cart: if you bid on a €90 variant but most buyers switch to the €60 sibling, your ACoS expectations implode. AMALYZE’s cart view makes this obvious; the standard console doesn’t.

Why Patrick chose AMALYZE: flat files, facts, and AMS

Patrick describes himself as a “numbers-first” operator. He loves flat files, hates guesswork, and hadn’t seen another tool that mirrored raw Amazon data the way AMALYZE does — then the Advertising Add-on layered AMS control on top. Crucially:

• The system only uses Amazon’s own data streams (Marketing Stream hourly signals, search term reports, etc.).
• It optimizes at the last mile — target level — with full respect for lookback and attribution windows.
• It’s structure-agnostic at import, then helps you segment what actually deserves distinct goals.

Time leverage: from impossible to automatic

The headline benefit for Patrick isn’t a cliché “save time.” It’s making the impossible routine. Christian shares one account’s math: in 30 days the system performed ~6.7 million target checks — roughly 223,000/day or 9,300/hour — before even deciding to adjust bids. No human or team can replicate hourly decisioning across thousands of targets, 24/7.

Practical effects:

• Teams finally move beyond Auto-only because they have the bandwidth to launch and manage Sponsored Brands/Display.
• Long-tail coverage expands safely: one client had run “10 keywords per ad group” forever; with hourly management, they could add dozens/hundreds of long-tails without drowning.
• Reactivity replaces weekly bulk edits. If you optimize only on Fridays, you leave six days of suboptimal bids on the table.

Target-level ACoS means recalibrating your brain

Patrick’s biggest shift was psychological. Instead of “20% ACoS at campaign level,” each target gets a goal and the system moves it hourly toward that goal. That forces adult math:

• Permission to play: if you were comfy at 30–35% ACoS but need 10% to be profitable, expect to lose auctions and volume. There’s no “more sales, lower ACoS, lower spend” trifecta.
• Market isn’t obliged to your budget: some brands demanded “1000% ACoS” during a blitz; even after bids went from €8→€15, CPC sat around ~€1.20 because no one else bid that high — supply of impressions capped.
• Per-target plateaus are real: you can’t drag an inherently 150% ACoS keyword to 7% overnight. Some targets simply don’t have that floor or ceiling.

Christian’s favorite framing: if the machine can return €1,000,001 for every €1,000,000, how much would you feed it? The answer is limited by stock, not fear — several clients had to throttle spend because operations couldn’t keep up with demand.

Attribution beyond 7 days changes the bids you can afford

Console reports typically stop at 7-day attribution. AMS carries 14- and 30-day attributions. In one onboarding, Christian showed that including 14/30-day clicks-to-sales added ~8% sales that weren’t visible in the console. Practical implication: you can rationally bid 30–40% higher CPC at the same true ACoS, outbidding competitors who price off 7-day only.

Patrick’s takeaway: none of this is magic. The data already exists; most teams just don’t use it in their bidding math.

Modifiers: why Patrick runs fixed bids (and Christian agrees)

They devote a big chunk to the danger of dynamic bids and placement modifiers:

• Dynamic bidding (up/down) can double your effective bids invisibly; placement modifiers go up to +900% for Top of Search, Rest of Search, and PDP — at campaign level.
• Because modifiers sit above target-level bids, they spray the uplift across everything and Amazon provides no clean per-target feedback on the actual uplift used.
• The math breaks: if your target ACoS is 25% but you run +60% placement, your effective ACoS target becomes 25 / 1.60 ≈ 15.6% just to compensate — and that’s before the bid/CPC “scissor” where you bid €1 but actually pay €0.60 (or vice versa once modifiers kick in).
• Bid management becomes contradictory: lower the base bid to hit ACoS and the modifier might still crank the paid CPC higher; the system can end up “reducing” base bids while effective CPC rises.

Patrick’s practice: fixed bids only. If you truly want placement-specific control, you would need tri-campaign setups per target (Top/Rest/PDP), which is operationally insane at scale and still lacks clean feedback. Category realities also matter: in jewelry, Top-of-Search share is inherently small because shoppers browse for style; modifiers won’t change that.

Structure for control, not comfort

AMALYZE can ingest any mess, but Christian and Patrick both stress segmentation for the few queries that actually deserve it:

• Pull out mission-critical themes (e.g., Herren/Männer vs. Damen, or occasion-driven terms like Vater/Papa or Männertag) into dedicated campaigns with their own ACoS.
• Don’t co-mingle those with your low-ACoS “workhorses” in one ad group — you lose the ability to set different goals.
• Expect odd long-tail wins. AMALYZE’s recommendations pull from any source (including “exact” that still throws off misspellings/stopwords/pluralization), check for duplicates account-wide, and propose sensible, pre-calculated starting bids tied to your goal and historic performance.

Patrick admits many recommendations “felt wrong” at first — queries like “männliches Armband” or unexpected furniture cross-terms — but the data wins. He processed ~500 recommendations in about 15 minutes; depending on results, that can create thousands of new targets and negatives across Auto/Phrase/Exact without chaos.

Auto → Exact migration without spreadsheets

Their preferred end state is simple: every spend-driving query becomes an exact target you control 24/7. AMALYZE’s workflow continually mines search term reports over your lookback window and proposes:

• Adding positives into the right campaigns.
• Seeding fresh bids calculated for your ACoS and average cart, not guessed.
• Setting negatives precisely where needed (Auto/Phrase/Broad) to dry up duplicative spend and move traffic into Exact.

It’s non-linear: Exact can seed fresh Phrase/Broad tests and vice versa, always deduped across the account and respecting SKU relationships.

Repricers, prices, and variant CPO traps

Two hard-won lessons:

• Repricers can torch your PPC math. If you swing price ±20–40% over a few days, every historical ACoS/CPO assumption is invalid. Hourly bidding can react to targets and lookbacks — it can’t rebuild your economics mid-flight every hour without you resetting goals, which then creates attribution chaos.
• Variant cart mismatch is real. If you model bids on a €90 SKU but shoppers buy the €60 sibling, your “true” average cart is €60 — your bid ceiling must be lower. Patrick has seen clients only realize this once AMALYZE surfaced ad cart values; the standard console hides that context.

Christian and Patrick repeatedly come back to CPO math. Example: to keep single-digit ACoS on a €7.50 item, your CPO must be ≤€0.75. At 20% CVR, your max CPC is €0.15. Bump CPC to €0.17 and you’re already at ~11.3% ACoS — tiny CPC moves can have huge ACoS swings.

Transparent decisions: Protocol and Simulation

Two workflow pieces Patrick calls “game-changers” because they build trust and speed learning:

• Protocol: a complete log that explains each hourly decision in plain language — e.g., win rate falls from 88%→58%, impressions drop X%, CPC drops €0.15, ACoS moves toward 15% goal. You see why the system acted.
• Simulation: before activating settings, you’re prompted to preview the first optimization hour and the expected trajectory over your lookback (7/14/30 days). It forecasts ACoS/CPO/impressions/clicks/sales/cost and highlights which targets drive the change. If you’ll crater win rate on a key term, you can adjust goals or segment first.

Low-data doesn’t stall the engine. Decisions don’t wait for 20 clicks/sale thresholds; the system incorporates time and attribution logic to move bids responsibly even with sparse signals.

Onboarding cadence, support, and who shouldn’t use it

Patrick’s advice to freelancers and in-house teams: yes, you can run this yourself — but do the onboarding and “work with the tool,” especially at the start. Christian’s team forces tight feedback loops (early, frequent check-ins, and they peek into accounts at least every ~14 days). They’ll ping you if you leave lookbacks stale, forget to activate segments, or wander far from your goal.

Who shouldn’t use it? If you spend only €600–€700/month and your time cost is negligible, the math won’t pencil — the tool runs ~€300/month and shines from ~€10k/month in ad spend unless time saved is itself worth the fee.

Budget fear is overblown. Patrick tried to “force” spend; most campaigns hit a market cap long before budgets do. Christian’s reminder: Amazon’s search supply is finite; you can’t multiply demand with budget alone.

Practical diagnostics you can do today

• Pull your Search Term Report and actually chart the hockeystick: sort by spend/sales and see how much budget sits right of zero sales. On Windows, “F11 on column 11” is Christian’s quick chart gag.
• Compute your true CPO ceiling: Avg Cart × Target ACoS = Max CPO; Max CPC = Max CPO × CVR. Now compare with your real CPCs.
• Spot variant/cart mismatches: are you advertising high-ticket SKUs while the cart skews to cheaper siblings?
• Switch off placement modifiers in a test cohort and run fixed bids with target-level goals; compare stability and speed-to-goal against your modifier-heavy setup.

Key takeaways

• Fix the store first: ads scale working offers; they don’t repair low CVR or bad retail.
• Let data set CPC ceilings: use 14/30-day attribution and per-target CPO math to decide what you can truly bid.
• Ditch placement/dynamic modifiers if you want control. Fixed bids + hourly target-level optimization beat campaign-level multipliers for clarity and predictability.
• Segment by business reality, not comfort. Pull mission-critical queries into their own campaigns with distinct goals; let recommendations grow the long tail safely.
• Price and variants can wreck ACoS. Repricers and variant switching alter your average cart; re-check your math when economics shift.
• Trust, but verify. Use Protocol and Simulation to understand and anticipate what the engine will do — then get out of its way.

Bottom line

Patrick Butscher’s core message is deceptively simple: unlearn the habits that made manual PPC barely manageable. Hourly, target-level bidding on AMS data makes “impossible” optimization practical — but only if you stop sabotaging yourself with placement modifiers, weekly bulk edits, and fuzzy ACoS thinking.

If you can articulate per-target goals, accept that the market (and your stock) cap your spend more than budgets do, and keep your price/variant hygiene in check, you can both grow and lower ACoS at the same time — not by magic, but by math and workflow. AMALYZE’s role in his stack is to turn that math into 24/7 execution with transparent logs, forecasts, and a ruthless migration of queries into exact control. For serious sellers and agencies, that’s the difference between chasing PPC and finally owning it.