The biggest top-of-search premiums were on the worst words

A common piece of Amazon PPC advice goes like this: top of search is the expensive placement, so shift budget toward rest of search and product pages, where clicks are cheaper. In the account behind our shower curtain liner audit, following that advice would have made things worse. Top of search was where the account made its money. The waste was in which words carried the biggest top-of-search premiums — and in four campaigns that were quietly spending on product pages.

Where the money went, by placement

Thirty days of the account’s placement report (Jul 21 – Aug 19, 2026, $410,274 of ad spend):

The cheapest clicks were the least profitable ones. A product-page click cost 38% less than a top-of-search click, but converted at a third of the rate.

The audit put a number on it: the most a click can cost in each placement while holding the account’s 36.6% ACOS. That ceiling is average price × placement conversion × ACOS — with a $10.49 average price:

Top of search was the only placement where the account paid less than it could afford. On product pages it paid almost twice the ceiling.

Why not just move budget down the shelf?

Because the conversion gap isn’t only about the slot. Shoppers who click a result at the top of a search for shower curtain liner are, mostly, people about to buy a shower curtain liner. Shoppers who click a sponsored product on someone else’s detail page are browsing. Shifting budget to cheaper placements buys more of the second kind of shopper — the clicks get cheaper and the orders get rarer.

So the audit used one lever only: each campaign’s top-of-search premium and base bid. No placement-level budget shifts.

Finding 1: the premiums were upside down

Start with how each family of search terms actually performed:

The account’s best word — the product’s own name, converting at 54% — ran at a 100% top-of-search premium, and the audit’s verdict was to hold it there. The curtain words convert at roughly two-thirds of that rate and cost 14 points more ACOS. Yet the campaigns the audit flagged for repricing — mostly curtain words — carried premiums of up to 200%:

At the far end were four “scenario” campaigns built around broad shopping phrases:

The account’s highest premium, 330%, sat on two of its worst-performing campaigns. These weren’t small samples either: each verdict rested on 313 to 1,309 clicks.

The campaign names add a clue. All four are named … – TOS 200 – Down – Relaunch, yet two now run at +330% and the others at +225% and +250%. The premiums had been changed since the campaigns were named — just not in the direction the results pointed. Whatever the history, the premiums no longer tracked what the words were earning.

What the audit recommended

• Pause the four scenario campaigns. Estimated saving ≈$4,623 a month against ≈$3,143 of sales given up. Spanish-language demand stays covered by a Sponsored Brands campaign already running at 32.9% ACOS on cortina para baño.
• Walk the curtain-family premiums down a ladder, one step a week — for example 200% → 160% → 120% on the main Shower Curtains campaign. Budgets and base bids stay untouched: these words still drive more than 13% of ad units.
• Gate every step. Move to the next step only if, after a week, the placement’s ACOS is at or under 42% and weekly units have fallen by no more than 15%. If either breaks, go back one step.

The audit estimated ≈$12–17K a month and 1.5–2 points of ACOS from the ladder. That’s a projection, labeled as one in the report, and the gates exist precisely because it might be wrong.

Finding 2: four campaigns leaking into product pages

The second problem was smaller and stranger. Four campaigns had spent $11,331 on placements they weren’t built for, at 63–83% ACOS:

The account happened to contain a perfect control. Test 2 ran the same exact keyword as Test 1 — and leaked only $332 into product pages. Same keyword, same product, a tenth of the leak. So the leak wasn’t a property of the keyword. It was a property of the settings.

How a campaign leaks

A top-of-search premium raises your bid only for top-of-search. Every other placement gets the base bid (unless you’ve set a separate adjustment for it). A campaign with a high base bid and a low premium is therefore competitive everywhere — including product pages — while a campaign with a low base bid and a high premium concentrates its spend at the top.

Test 1 had the first shape: a $0.99 base bid with a 100% premium, for an effective top-of-search bid of about $1.98.

The fix keeps the top-of-search bid where it is and takes away the rest:

The same pattern applied to the other two keyword campaigns ($1.17 → $0.90 with 50% → 120%; $1.93 → $1.40 with 50% → 130%), and the product-targeting campaign got a plain bid cut. The gate: product-page and rest-of-search spend should fall below 5% of each campaign within two weeks; if top-of-search impressions drop by 20%, raise the base bid back 10%. Estimated saving ≈$6–8K a month.

Check your own account

1. Pull the placement report for the last 30 days and compute the affordable CPC per placement: average price × placement conversion × your target ACOS. Wherever you pay more than that, the placement is losing money at your target.
2. List your top-of-search premiums next to each campaign’s ACOS and conversion. Sort by premium. If the highest premiums aren’t on your best-converting words, they’re upside down.
3. Look for twins. Find keywords that run in more than one campaign and compare how much each spends outside top of search. A large gap between two campaigns on the same keyword is a settings problem, not a keyword problem.
4. Change premiums in steps, with a gate. One step a week, a stated ACOS ceiling and a stated unit floor, and a rule for stepping back.

Placements are one of nine dimensions a Sapientia audit scores. In this account it scored 64 — not the weakest dimension (that was campaign structure), but the one where most of the audit’s estimated savings were.