Amazon Sponsored Products Strategy: What to Actually Build

What a Sponsored Products strategy actually has to do

Most Sponsored Products accounts have one campaign per product, set to automatic targeting, running unchanged for a year. That's not a strategy, that's a default. A strategy assigns each campaign a specific job, because a single campaign can't discover new search terms, convert proven ones efficiently, and defend your product page from competitors all at the same time — the bids that work for one job are wrong for the other two.

There are three jobs. Discovery finds search terms and ASINs you haven't targeted yet. Harvest takes the terms that already convert and bids them tightly, cheaply, on their own. Defense targets your own product pages and close competitors' pages so a shopper who's already looking at your category doesn't drift to someone else. Build for all three and the account gets more efficient over time. Build for one and it plateaus.

The structure: five campaign types, not one

• Auto targeting (discovery): Amazon matches your ad to related search terms and product pages. Low bid, small budget, its only job is to surface search terms you didn't think of.
• Broad/phrase match (research): manually chosen keyword themes, run wide on purpose, to confirm which of the terms auto found actually convert at volume.
• Exact match (harvest): the proven terms from auto and broad, isolated into their own campaign with a bid you can control precisely. This is where most of the profitable spend should end up.
• Product targeting (defense): ASIN-level targeting on your own listings and on close competitors, run separately because the economics are different — you're buying placement, not a search query.
• Negative keywords/products: not a campaign, a discipline applied to every campaign above, so terms that waste spend in auto don't also waste it in broad.

A worked example: turning search term data into bids

Say you launch a new product with $30/day split across an auto campaign and a broad-match campaign built around ten keyword themes. After two weeks, pull the search term report. It isn't a summary of what happened — it's a list of what to build next.

Suppose three search terms have each cleared 20+ clicks and are converting under your target ACOS. Those three get moved into a new exact-match harvest campaign with a bid set to hold that ACOS, and they get added as negative exact in the broad campaign so you stop paying twice for the same click. Everything else that's cleared 20+ clicks with zero sales becomes a negative. Everything under 20 clicks stays — it hasn't told you anything yet, and pausing on ten clicks is a coin flip, not a decision.

Run that cycle every two weeks. The harvest campaign grows, the broad campaign gets leaner, and the auto campaign keeps finding new terms to feed the next cycle. That's the whole mechanism. Everything else is tuning.

How to know it's working, not just running

ACOS on its own is a trap. It tells you the cost of the sales the ad system is willing to credit itself with — it doesn't tell you whether that sale would have happened anyway from organic ranking, and it can look great on your best keyword while the account as a whole is flat or losing money. Look at TACOS (total spend over total sales) alongside ACOS, and judge the account by the whole campaign structure, not the top line item.

We see the same distortion at larger scale in DSP, where it matters even more because the media costs are higher and less visible. Across 30 of our managed advertisers in July 2026, the book delivered a 6.04x return on ad spend — measured across the whole portfolio, not the best-performing line. That same book ran at a blended $1.42 CPC; the $0.41 CPC figure that gets quoted in this category is online-video only, not the whole account. The principle carries over directly to Sponsored Products: a blended number across every campaign is honest, a number from your single best keyword is marketing.

The common mistakes, including ones we've made

• Pausing too early. Killing a keyword after three clicks and no sale isn't optimization, it's noise. Set a minimum-clicks threshold before you make any pause decision — we've paused campaigns ourselves before they'd cleared enough volume to mean anything, and the fix was writing the rule down before the next launch, not after.
• Running broad and exact on the same term forever. If a term is proven, it belongs in harvest and out of broad. Leaving it in both means you're bidding against yourself.
• Ignoring the placement report. Top-of-search and product-page placements convert differently and should be bid differently; one blended bid for both is leaving money on the table in one and overpaying in the other.
• Treating ACOS as the only number that matters. A campaign with a slightly worse ACOS but a much higher rate of new-to-brand customers is often the better campaign for the business, not the worse one for the account.

When it isn't working: what to check first

If ACOS is high, check the search term report before you touch the bid — you're probably paying for broad matches that drifted off-topic, and the fix is negatives, not a lower bid. If impressions are low, check your bid against the suggested range and check inventory; an out-of-stock or low-stock listing won't serve ads no matter how you bid. If conversion rate is low but clicks are healthy, the ad is doing its job and the problem is the listing — price, images, reviews — not the campaign.

Give any structural change at least two full cycles of data before deciding it didn't work. Restructuring every few days produces an account that never accumulates enough history on anything to actually optimize.

Where this fits with reMKTR

Sponsored Products is a keyword and bidding discipline you can run yourself with the structure above, and plenty of brands do it well without an agency. Where we come in is further up the funnel: we run Amazon DSP as a managed service, reconciled in Amazon Marketing Cloud so display and sponsored ads stop double-counting the same sale, because last-click attribution can't tell you if the display spend actually added anything — only a holdout can. If your Sponsored Products program is solid and you're trying to work out whether upper-funnel media is worth the money, that's the question we're built to answer.