You've been doing this long enough to know the trap. A campaign that looks fine. Revenue holds. ACoS is in range. Nothing on the dashboard screams. But something about the placement breakdown feels off and you can't quite say what.
You run the bulk sheet through BulkSheet Pro. The Campaigns page shows three placement recommendations on the same campaign — three different adjustments, all moving in different directions.
This is the story of what it saw.
First, the gate decides which lever fires
Before BulkSheet Pro recommends a single placement adjustment, it makes a decision at the campaign level: is this a bid problem or a placement problem?
The reasoning matters. A campaign that's underperforming has two possible causes. The first is that one or more keywords are bidding wrong — too high (overpaying) or too low (missing impressions). The second is that one or more placements are absorbing budget disproportionately to what they're returning — Top of Search eating spend while Product Pages converts better, that kind of thing. These two causes call for different fixes, and applying the wrong fix mostly just wastes a cycle.
For each campaign, BulkSheet Pro's diagnostic gate computes two numbers: the biggest gap between any keyword's ACoS and its target, and the biggest gap between any placement's ACoS and its target. Whichever gap is bigger by at least 10 percentage points wins. That lever fires. The other is suppressed for that campaign — only one lever moves per cycle.
When the two gaps are within 10 percentage points of each other, the gate considers them tied. The default tiebreaker: bid wins. Bid moves are the broader lever (they affect all surfaces simultaneously) and the more reversible one. There's one exception — if one placement in the campaign is consuming much more spend share than its sales share (what the gate calls a worst-imbalance signal), placement wins the tied race instead. This catches the case where the campaign's overall ACoS looks acceptable but one placement is silently bleeding while another is over-delivering.
This routing happens before you see anything on the Campaigns page. By the time placement recommendations appear at all, BulkSheet Pro has already determined that placement is the right lever for those specific campaigns.
One placement, one decision
Take the simplest version. A campaign called AirFryer-Manual-Exact has Top of Search set at +100%. Over the last 30 days that placement has 612 clicks, $148 in spend, and $190 in sales. ACoS at Top of Search: 78%. Your account target is 35%. The diagnostic gate determined that the placement layer is where this campaign's bigger problem lives, so we're in placement-rec territory.
Before any math happens on the 78% ACoS, BulkSheet Pro checks two reliability gates.
The first is whether the placement has enough activity to support a recommendation: at least 20 clicks AND at least 2 orders. Below that floor, the signal is too noisy to act on — BulkSheet Pro doesn't generate a rec, doesn't show a row, doesn't try to make a call. The data simply isn't there. Top of Search with 612 clicks and likely several orders passes this gate easily.
The second is whether to trust the raw ACoS as-is or to smooth it. BulkSheet Pro applies Bayesian smoothing to a placement's conversion rate — pulling the CVR of low-data placements toward a campaign or account prior. A placement with 22 clicks and 2 orders has a raw CVR of 9.1%, but that's based on thin data; if the campaign's overall CVR is 12%, the smoothed CVR lands somewhere in between, reflecting how much trust the engine should put in this specific placement's history. The ACoS that gets compared against your target is computed from the smoothed CVR. For Top of Search with 612 clicks, the smoothing pulls almost nothing — the raw 78% is robust enough to use directly.
Now the interesting math.
First, it has to decide which target ACoS applies. You set 35% as the account default, but BulkSheet Pro walks a four-level rule hierarchy before settling on that number: SKU rules first (the most specific), then keyword rules, then campaign-type overrides for Sponsored Brands or Sponsored Display, then the account default as the floor. You can set different targets on branded keywords versus generic ones, or on new SKUs versus mature ones, and the rec engine respects that hierarchy automatically. For this campaign no specific rule matches, so the account default rules: 35%.
Next, it calculates how far off target the placement is — but it does this relative to the placement's own ACoS, not as an absolute difference. The reasoning matters. A placement at 80% ACoS with a 35% target is 2.3 times over target. A placement at 800% ACoS with the same target is 23 times over. Both are "over" but the second needs to move dramatically more. Reading the gap as a ratio keeps the recommendation honest at any scale — it won't slam an aggressive cut onto a placement that's only slightly above target.
Then comes sensitivity. BulkSheet Pro's default sensitivity is 0.25 — what the settings call a "gentle nudge." This is intentional. A placement well over target gets a meaningful but not punishing reduction. If you want bolder moves you can tune it up. If you're running fragile new accounts you can dial it down. Either way BulkSheet Pro doesn't decide the aggressiveness for you — you do.
The result for Top of Search at +100% with that 78% ACoS: a recommended multiplier of +86%. About a 14-point reduction. Confidence: HIGH (612 clicks, well above the 20-click floor where BulkSheet Pro is willing to act on its own math).
This is the single-target story. The multi-target story is where things get interesting.
Same campaign, three different verdicts
Most tools would stop after the Top of Search recommendation. The campaign was flagged. A placement was adjusted. Ship it.
BulkSheet Pro doesn't treat the campaign as a unit. It treats each placement as its own conversation.
Looking at the same campaign, the other placements tell different stories:
Product Pages is set at +50%, has 891 clicks over the same window, and an ACoS of 22%. That's well below target — significantly. This placement is converting better than your goal asks of it. BulkSheet Pro doesn't only pull bad performers back. It also pushes good performers forward when the data supports it. Same formula, opposite direction. The recommendation: nudge Product Pages from +50% up to roughly +65%. A 15-point lift on a placement that's earning its keep.
Rest of Search is at +30%, has just 14 clicks, and an ACoS of 41%. It's slightly over target — but 14 clicks is below the 20-click reliability gate, and Rest of Search hasn't produced any orders yet. No rec generates at all. Not low-confidence, not visible on the page in the default view — just absent. BulkSheet Pro is transparent about data quality rather than pretending every signal is equal: below the floor, there's no honest recommendation to make, so there isn't one.
So in one campaign:
- Top of Search gets pulled back (overspending)
- Product Pages gets pushed forward (under-leveraged)
- Rest of Search isn't evaluated at all (clicks below the reliability floor)
Three different decisions on three different surfaces — independently calculated, independently presented. You see them as three separate items on the Campaigns page, each with its own confidence rating and its own direction. The campaign isn't one verdict; it's three.
Most tools can't do this because they treat the campaign as the atom. BulkSheet Pro treats the placement as the atom and the campaign as a container. The difference matters because the placement layer is genuinely heterogeneous — the same campaign can have a hot Top of Search and a cold Product Pages, and those placements are answering to different shoppers with different intent. There's no reason their multipliers should move in lockstep.
When both levers fire (the dual-fire case)
The diagnostic gate's default behavior is to route a single lever per campaign per cycle: either bid OR placement. There's one specific situation where it fires both.
For campaigns with multiple keywords where one of the placements is clearly bad, BulkSheet Pro generates both a bid recommendation and a placement recommendation — but it halves the nudge math on both. "Clearly bad" has a specific definition: the placement's ACoS has to exceed both an absolute threshold (1.5× the campaign's loosest target) AND a relative threshold (1.3× the best valid placement's ACoS in the same campaign). The absolute check catches outliers. The relative check ensures there's actually an allocation problem to solve, not just every placement being mildly bad together — if everything is uniformly underperforming, that's a campaign-level issue, not a placement-allocation issue.
When both criteria hit, the placement and bid recs share the work. The reasoning: bid and placement compound in the auction. A 20% bid cut combined with a 20% placement cut isn't a 40% combined effect on the auction price; the two cuts multiply, producing a 36% combined drop. The 0.5 scaling on each side keeps the combined movement close to what the math intended.
For single-keyword campaigns, this case doesn't apply — they always get one lever or the other. For multi-keyword campaigns without a clearly-bad placement, the gate routes one lever as usual.
Beyond the standard three
The example above used the three placements every Sponsored Products campaign exposes by default: Top of Search, Product Pages, and Rest of Search. But BulkSheet Pro recognizes two more placement modifiers, both also on Sponsored Products.
Amazon Business is a separate placement modifier — for campaigns enabled on the Amazon Business marketplace. If you sell B2B-relevant SKUs (office supplies, industrial inventory, larger pack sizes), this is its own surface with its own bidding dynamics. A B2B-relevant ASIN can convert well on Amazon Business while underperforming on the consumer surfaces, or vice versa. BulkSheet Pro evaluates it independently, with the same formula and the same confidence rules. If your campaigns don't have an Amazon Business multiplier set, this placement simply doesn't appear in your recommendations.
AMC Audiences are the audience-targeting side of the placement universe. Sponsored Products campaigns can layer audience-based bidding adjustments on top of the search placement multipliers — an AMC audience segment (retargeting, in-market audiences, lookalikes, loyalty cohorts, custom AMC audiences) gets its own multiplier that sits alongside Top of Search, Product Pages, and Rest of Search on the same SP campaign. A single SP campaign with three audience segments has three additional placement-style decisions on top of the standard three. Each audience segment's multiplier is evaluated by BulkSheet Pro the same way: target-ACoS-relative, sensitivity-tempered, confidence-aware. The campaign can have an aggressive Top of Search modifier AND a separate aggressive in-market audience modifier — and BulkSheet Pro will reason about each independently.
The principle generalizes: any surface where Amazon lets you set a separate multiplier is a surface where the performance can diverge from the rest of the campaign. BulkSheet Pro treats each surface as its own decision. The number of recommendations you see for a campaign is a function of how many surfaces you're running on — not how many recommendations the tool thinks it can squeeze out.
What BulkSheet Pro won't do
A few things you learn to trust:
It won't act on noise. Each placement needs at least 20 clicks AND at least 2 orders to participate in the rec engine at all. Below that floor, the recommendation isn't generated — not even as low-confidence. The data simply isn't there to support an honest call.
It won't trust raw ACoS on low-data placements. Bayesian smoothing pulls the CVR of placements with thin data toward the campaign or account prior, so the ACoS the engine evaluates reflects how much the data can actually be trusted. Robust-data placements like Top of Search with 600+ clicks are evaluated essentially as-is; thin-data placements get pulled toward the prior.
It won't break Amazon's rules. Placement multipliers are clamped between 0% and 900% — Amazon's hard limits. The rec engine respects them even when the math wants to push further.
It won't recommend what can't be applied. Legacy Sponsored Brands placements can't be edited through bulk uploads, so BulkSheet Pro doesn't bother recommending them. They still appear on the Campaigns page for visibility — they're cleanly excluded from rec actions.
It won't fire both levers at full strength on the same campaign. When the dual-fire case triggers (multi-keyword campaign with a clearly-bad placement), both bid and placement recommendations get scaled to half their normal nudge. This is the only situation where both levers move; in every other case, the gate picks one and suppresses the other.
It won't scale traffic that's failing across the board. A safety rule catches the case where every placement in a campaign is above target and the total ACoS exceeds your target by 10% or more. That's a campaign-level problem (probably bid-related), not an allocation problem, and BulkSheet Pro forces the gate to route to bid in that case. The mirror rule exists for the opposite — every placement under target plus total under target by 10% — where it again forces bid (upward this time) rather than trying to optimize placement multipliers that are already working.
It won't pretend a campaign is one decision. When placement wins the routing for a campaign, each valid placement gets its own evaluation, its own row, its own direction. You're the one looking at all of them and deciding what to ship — but BulkSheet Pro has already done the work of showing you they're separate stories.
And you can tell it which direction you care about. The mode filter — grow, reduce, balanced — lets you see only push-up recommendations (you're aggressively scaling), only pull-down recommendations (you're tightening up), or both (you're running the account holistically). The math doesn't change. What you see does.
The export
You hit Export, download the bulk sheet, and upload it to Amazon. Three placement changes on one campaign, plus 47 others across the account. The month that follows isn't a miracle — it's a measurable improvement. Account-level ACoS drops a couple of points. Top of Search spend comes back to earth. Product Pages picks up volume.
The thing about the placement layer isn't that it's hard to optimize. It's that it asks for three independent decisions per campaign instead of one, and most tools treat that as too granular to bother with. BulkSheet Pro made it the default.