Why Is Top of Search the Cheapest Placement on Your Own Brand Terms?

01The answer: same campaign, same day, half the price

The cleanest test does not compare averages. It takes one branded campaign on one day and divides its Top of Search CPC by its product-page CPC on that same day — same campaign, same bids, same modifiers, same competitor set. Each bar below counts campaign-days by that ratio. On brand terms the whole distribution sits left of 1.0. On the generic control it sits right of it.

So the ranking on a brand's own exact terms is Top of Search < rest of search < product pages — the exact reverse of the generic control, where Top of Search costs 1.5× product pages. The question is not whether this happens. It is why, and what it means for how you set bids.

02It holds across brands, markets and currencies — with one exception

Click-weighted CPC by placement for each brand-market with at least 300 branded clicks. Local currency, so compare within a row, not across rows. Top of Search is the cheapest placement in all eleven rows.

Paired per brand (Top of Search ÷ product pages, same campaign-day, ≥3 clicks each): Brand B 0.49× on 87% of 2260 days; Brand A 0.55× on 95% of 495 days; Brand D 0.46× on 100% of 212 days; Brand C 0.71× on 88% of 117 days; Brand E 0.63× on 89% of 104 days; Brand F 0.24× on 100% of 84 days. No brand in the table comes close to parity.

Two brand-markets sat outside this table, and we went back and tested them, because excluding them was our choice rather than the data's. One reproduces the result and should have been in the study: a hair-care brand in the US with 12,566 branded clicks — more than four of the rows above — pays $2.23 at Top of Search against $3.37 on product pages, a median 0.67× on 111 paired days, cheaper on 89% of them. Our notes recorded it as having too few branded clicks, which was simply an error.
The other does not reproduce. An intimates brand in India, on 1,227 branded clicks, pays ₹26.01 at Top of Search against ₹33.22 on product pages — still cheaper, but on only 58% of 45 paired days — and its cheapest placement is rest of search at ₹10.88, which breaks the Top of Search < rest of search < product pages ordering this study is about. It is the one brand-market we have found where the pattern does not hold.
So the claim this section can carry is narrower than "every brand, every market, every currency": the ordering holds in every brand-market with real branded volume that we have looked at, with one low-volume exception, and only inside accounts we manage.

03What you pay versus what you bid, per placement

To see why, you need the bid that was live on the day, not just the CPC. We snapshot every keyword bid and every campaign's placement modifiers daily, so for 530 campaign-days (81 branded keywords, single-keyword campaigns, dynamic bids down-only) we know the exact effective bid per placement: base bid × (1 + placement modifier). Dividing the realized CPC by that number gives the share of your bid Amazon actually charged.

This is the mechanism in one number. On a brand's own exact term at Top of Search, Amazon charges about a tenth of the effective bid — 10.7% click-weighted, 11% on the median day, and under a third of the bid on nine days out of ten. On product pages the same campaigns pay 88% of the bid, and on the median day exactly 100%. Generic keywords pay 83–95% of the bid everywhere.

Read 10.7% as one account's number, not a constant. The strict cohort needs a campaign where exactly one keyword took clicks, and only one advertiser in this sample runs branded campaigns that way: roughly nine in ten of these campaign-days are the same skincare brand in Germany. Take that brand out and the branded Top of Search paid share lands near 40% on the days that remain. The figure also depends on a number the advertiser chose: paid share divides by the effective bid, so a 200% Top of Search modifier cuts the ratio threefold by construction. On the modifier-free days — where the effective bid is simply the base bid — branded Top of Search pays 39% (see section 07), against 83–95% for generic keywords. The gap between brand terms and generic terms is real in every cut we have; its size is not 8× but closer to 2–3× once the modifier is matched, and it varies from about 12% to about 60% across accounts.

Amazon's Sponsored Products auction is a second-price auction on ad rank: you pay just enough to beat the runner-up, adjusted for relative click-through rate. On your own brand term at the top of the page, the runner-up is a competitor conquesting your name with an ad shoppers rarely click — your click-through rate there is 17.0% against their generic-level CTR — so the clearing price collapses to a fraction of your bid. Your bid is not setting the price; the runner-up is. On product detail pages the same keyword drops your ad into a category-wide detail-page auction (CTR 0.07%), where dozens of product-targeting and auto campaigns bid at your level and you pay what you bid.

04Raising the bid does not raise the Top of Search price

If the runner-up sets the price, your bid should have almost no effect on what you pay at Top of Search — until you drop below the runner-up and lose the slot. Each dot is one campaign-day (≥2 clicks): the effective Top of Search bid on the x-axis, the realized Top of Search CPC on the y-axis. Generic keywords hug the diagonal. Brand terms lie flat along the floor.

Across the branded bid quintiles the effective bid rises almost fourfold (1.21 → 4.54) and the Top of Search CPC stays between 0.28 and 0.35. On generic terms the CPC follows the bid quintile by quintile. Within a single keyword over time the same picture: where a branded keyword's effective Top of Search bid changed during the window, a 10% higher bid meant about 3% higher CPC (median elasticity 0.33, 8 keywords); on generic keywords it meant 8% higher (0.84, 38 keywords). On product pages the branded elasticity is 1.16 — there, your bid is the price.

05The Top of Search modifier buys share, not price

Most branded campaigns in this sample run a Top of Search modifier between 50% and 200%. If your bid does not set the Top of Search price, the modifier should not either. It doesn't — but it does change where your clicks come from.

Read the left column top to bottom: on brand terms a 0% modifier pays 37% of the base bid at Top of Search and a 101–200% modifier pays 30% — the price is flat while the effective bid doubled or tripled. On generic terms the same modifier buckets push the price to 196% and 5.1× the base bid. What the branded modifier does buy is share: Top of Search goes from 51% of the campaign's clicks at 0% to 90% at 101–200%, because a higher effective bid wins more of the auctions where a conquesting competitor was briefly ahead.

The within-campaign test points the same way, on a much thinner base than the bucket chart above. Hold the keyword bid genuinely fixed from one day to the next, require at least 3 Top of Search clicks on both days, and this sample yields only nine clean modifier changes — every one of them a large move. When the effective bid was raised by a median 63%, the next day's Top of Search CPC was 0.71× (n = 2); when it was cut by a median 35%, the CPC was 0.85× (n = 7). Both directions sit inside ordinary day-to-day noise, and the regression slope across the nine is slightly negative.

An earlier version of this section reported 128 events and an elasticity of 0.09; that figure was not reproducible and this replaces it. Dropping the requirement that the bid stayed put multiplies the event count roughly eightfold, but it then admits days where the bid moved as well as the modifier, and it becomes dominated by a single advertiser's automation nudging the modifier one to five points at a time — moves far too small to reveal a price response. An elasticity computed that way is measuring noise. On the nine events where the arithmetic is clean, the direction is consistent and the magnitude is not identified. Read "the modifier does not move the Top of Search price" as a claim about direction — supported here and by the flat bid quintiles in section 04 — not as a measured elasticity.

06Product pages: the expensive placement hiding inside your brand campaign

A branded exact keyword does not only trigger on the search results page. Amazon also serves it on product detail pages it considers relevant, and that inventory behaves like a generic auction: you pay 88% of your bid, the click-through rate is a rounding error, and the conversion rate is less than half of Top of Search.

Product pages take 26% of branded spend and return 9% of branded sales, at an ACOS of 28% against 7% at Top of Search. That is not because product-page shoppers are bad shoppers — the 15% conversion rate is respectable. It is because the click costs 1.4× more, and it costs more because on that placement you are paying your bid instead of the runner-up's. Note the impression column: 95% of a branded campaign's impressions are product-page impressions nobody clicks, which is why campaign-level CTR on brand campaigns looks so much worse than the search-page reality.

07Where the effective-bid math breaks: audience, B2B and intraday noise

All of the above relies on knowing the effective bid, and on a down-only campaign a click should never cost more than it. So any campaign-day where CPC exceeds the computed effective bid is a day where something else was adjusting the bid. We classified every campaign-day in both cohorts by what was live on it.

With a stable bid and no audience or B2B modifier, CPC exceeds the effective bid on 0.5% of days — the model is exact. Every other row is a source of noise you need to account for before reading a CPC-to-bid ratio:

• Audience (shopper cohort) modifiers — boosts of up to 900% for shoppers in a chosen segment. They are invisible in the placement report, so their clicks land in the same Top of Search / product-page rows at a higher price. CPC exceeded the bid on 26% of those days, mostly by a little (more than 10% over on 4.3%).
• Amazon Business (B2B) modifier — the same problem, worse: Amazon does not report Business as a placement, it folds B2B clicks into the standard three. With the modifier active, one day in five pays more than the computed bid and the 90th-percentile day pays 2.5× it. With both modifiers active, 75% of days exceed the bid.
• Intraday bid changes — if a bid tool moves the keyword bid at 09:00, the day's CPC blends two bids. A quarter of such days exceed the morning bid. Any daily analysis has to freeze on days where the bid did not change.
• Dynamic bids up-and-down — Amazon may lift the bid by up to 100% at Top of Search and 50% elsewhere. In the generic control it did: Top of Search paid 1.63× the computed bid and rest of search 1.49×. Too few branded campaigns ran up-and-down to measure it there.

None of this changes the headline: the 11% Top of Search paid share is measured on down-only campaigns, and the small modifier-free subset (13 campaign-days) comes out at 39% — still a fraction of the 88% on product pages, and of the 83% generic keywords pay at Top of Search.

08Spotlight: one brand keyword, 50 days, three prices

One single-keyword campaign on the brand's hero product term (skincare, Germany), every day it collected clicks from July to September. The base bid was €1.13 with a 200% Top of Search modifier, so the effective Top of Search bid was €3.39. Watch the three lines.

Over the 50 days the keyword bought 1,400 Top of Search clicks at an average €0.27 — about 8% of a €3.39 effective bid — and 39 product-page clicks at €0.96, which is 85% of the €1.13 bid that applies there. The product-page line rides the base bid almost every day; the Top of Search line never gets near its own bid. The base bid was cut to €0.81 on 21 July and raised to €1.05 on 26 August; the Top of Search CPC noticed neither. On the last two days the modifier was set to 0% and the bid to €0.75 — an effective Top of Search bid a fifth of what it had been — and the clicks still cost €0.24 and €0.26, with the campaign still winning Top of Search. That gap between the two purple lines and the blue one is headroom you are paying nothing for, and that a bidding tool reading "CPC well below bid" would happily spend.

09Stable week after week

Weekly click-weighted paid share (CPC ÷ effective bid) on the strict branded cohort, June to September. The Top of Search line sits between 8% and 23% every single week; product pages between 82% and 113% (above 100% only in the week where audience-modifier days dominate). Rest of search is thin and noisy.

10What we're doing with this at SYNQ

11Method & caveats

• Data. Sponsored Products daily reports from the Amazon Ads API for accounts SYNQ manages: campaign placement report (Top of Search on-Amazon, Detail Page on-Amazon, Other on-Amazon; Off Amazon excluded), targeting report (per keyword per day), plus a daily 07:30 UTC snapshot of every keyword bid, campaign bidding strategy and placement / audience / B2B modifiers taken by our platform since 9 June 2026.
• Branded cohort (broad, sections 01, 02, 06). 8,322 campaign-days from 156 branded Sponsored Products campaigns across 7 brands and 11 brand-markets (US, UK, DE, FR, ES, IT), 1 January – 10 September 2026, 324,203 clicks and 368,392 in local-currency spend. A campaign-day qualifies when every keyword that received a click that day contains the brand name, and the campaign carries at least one branded exact-match keyword. Brand-markets with fewer than 300 branded clicks were dropped. Match type is mixed inside this cohort — 43% of its clicks come from exact keywords, 46% from phrase and 13% from broad — so it is a branded cohort, not an exact-match one. We checked every customer search term that bought a click in it: 1.7% came from a query that does not contain the brand, and most of that residue is apostrophe and spacing variants of the brand itself. The traffic is branded; only the match type is mixed.
• Strict cohort (sections 03, 04, 05, 08, 09). 530 campaign-days, 81 keywords, 56 campaigns, 9 June – 10 September 2026: single-keyword exact-match branded campaigns on dynamic bids down-only, where exactly one keyword received clicks, its bid was captured that morning, and the campaign's modifiers were captured the same day. Effective bid per placement = bid × (1 + that placement's modifier). Days where the bid differed from the next morning's snapshot are excluded from the paid-share statistics. The strict cohort is dominated by one skincare brand in Germany; the broad cohort and the per-brand table show the CPC ordering is not brand-specific.
• Generic control. 7,166 campaign-days from 371 single-keyword exact-match campaigns on non-brand terms in the same accounts and period, built identically.
• Paired ratios use campaign-days with at least 3 clicks on each of the two placements being compared, and report the median of the log ratio. Paid share is Σ cost ÷ Σ (effective bid × clicks), i.e. click-weighted; medians and percentiles are over campaign-days. Elasticities are the slope of log CPC on log effective bid within one keyword across days.
• What we cannot see. Amazon does not report the runner-up bid, hourly prices, or which product pages served the ad. "Second price set by the runner-up" is the standard description of the Sponsored Products auction and the only mechanism consistent with a flat price across a fourfold bid range; we have not observed the auction directly. Currency is mixed in pooled CPC figures (USD, GBP, EUR); every ratio is computed within a campaign-day and is currency-free.
• Selection. These are brands with established branded search demand and a Top of Search position they usually win. A young brand whose term is dominated by a larger conquesting competitor will see a higher paid share. The one brand-market we have found where the placement ordering does not hold is a low-volume account in India where rest of search is the cheapest placement (section 02). The generic control is the same accounts' non-brand exact campaigns, not a random sample of Amazon advertisers.
• Campaign-level modifiers are one number per campaign and placement; where a multi-keyword branded campaign appears in the broad cohort, its keywords share the same modifiers but may carry different bids, which is why bid-based statistics use the single-keyword strict cohort only.