Fragrance and scenting brand: a placement audit on a seven-figure month

A ranked list of where spend worked, and a warning against the obvious fix.

The situation

This brand sells scenting systems and diffuser oils, much of it to hospitality buyers. It runs a very large Amazon US program, over 200 Sponsored Products campaigns and a 30-day spend in the high six figures, at roughly 28% ACoS. This is a healthy account. The question was not what was broken. It was where moving money would pay.

What the placement report showed

Amazon sells ads in three places: top of search, rest of search and product pages. We analyzed each across every campaign.

• Top of search was efficient and tiny. It ran around 23% ACoS and 26% conversion, took about a quarter of spend, and received only 3.7% of impressions. Product pages ran near 30% ACoS and 15% conversion, took 42% of spend and 61% of impressions.
• A weak slice was easy to find. 61 campaign-and-placement combinations had spent at least $300 at an ACoS above 45%. Together they were about 12.6% of spend, returning sales at a 56.7% blended ACoS. Four of every five of those dollars were product-page placements.
• Some campaigns were lopsided. 13 campaigns with more than $2,000 in spend sent under 15% of their budget to top of search. In the widest gaps, top of search ran at 15 to 17% ACoS while product pages ran at 36 to 53%.

The fix we advised against

The obvious conclusion is to shift budget from product pages to top of search. We modeled it and recommended against a blanket shift. Top-of-search return on ad spend was only 1.31 times the product-page return. A shift only pays while the marginal return on added top-of-search spend stays above about three-quarters of its average, and extra dollars there buy progressively worse clicks. At a realistic 60% haircut on the marginal return, a 20% shift loses sales rather than gaining them.

The better plan was narrower: act campaign by campaign, cap each change, and leave alone anything performing, such as the brand's largest manual campaign, which ran near 19% ACoS.

Checking that changes land

We built the changes as an upload-ready file with multipliers capped between 0% and 50%, so they would not fight the bid-management platform's own algorithm. A later check made the point that you should verify, not assume: by correlating cost-per-click ratios against the intended multipliers, we could see the first file had apparently not taken effect. We now treat "did the change land" as a step in every placement project.

The second period

Across the next 30 days spend and ACoS were flat, but conversion fell from 19.5% to 17.9%. The placement mix had actually improved, which means within-placement conversion fell by more than the headline number. That is not a bidding problem. The next place to look is price, stock, reviews and competitors on the two hero listings.

The takeaway

On a healthy account the work is mostly deciding what not to touch. The value of the audit was a ranked list of cells and a reason to resist a tempting blanket change.