Amazon DSP gets sold as an awareness machine. The standard play runs top down, with video at the top introducing the brand to strangers and remarketing at the bottom mopping up whatever trickles through. We run it the other way, and when we open a new DSP flight we start with zero upper funnel spend at all.
Display remarketing carries the volume because the audience already raised a hand. Shoppers who viewed the product page or abandoned a cart have declared intent with their own behavior, so the job left is to finish a sale already in motion. Display is the hammer for that job. Impressions are cheap, frequency is controllable, and the return is measurable inside a window short enough to act on. Prospecting to cold audiences asks display to persuade, which it does poorly, while remarketing asks it to remind, which it does well.
The budget follows that rule. We run flights on a 60/30/10 split, and the largest share belongs to remarketing display because the money should sit where the intent already exists. The smaller shares expand outward only from audiences that the remarketing core has already proven can convert.
Video is where discipline matters most, because video is where DSP budgets go to die. Our rule is that video has to earn its DSP budget in Sponsored Brands first. The Sponsored Brands video auction gives fast, cheap, keyword level feedback on whether a creative actually stops a shopper and converts one, and a video that cannot win there will not be rescued by DSP inventory at DSP prices. Running unproven video on DSP is just an expensive way to discover that the creative is weak. When a video wins in Sponsored Brands, it has demonstrated demand capture with real purchase data behind it, and only then does it deserve programmatic scale.
Products match audiences, not funnel stages. The funnel stage tells you how warm a shopper is, but it says nothing about which product belongs in front of them, and that is where most flight plans go wrong. On one portfolio we ran, we built an audience of medical professionals, and the product that belonged in front of them was the footwear they actually wear on shift rather than the fragrance line from the same catalog, because the audience connection was real for one product and purely aspirational for the other. A nurse standing on a hospital floor for twelve hours has a concrete problem that one of those products solves, and no amount of funnel logic makes the other one relevant.
Every dollar you push upstream comes out of shoppers already reaching for the product, and We have never seen an awareness metric that justifies that trade. The upper funnel is a privilege you earn, not a starting point. It is not worthless, but it only turns on after remarketing audiences are saturated, video has proven itself in the sponsored auctions, and the account is converting everything the lower funnel produces. At that point, moving toward colder audiences becomes a rational bet instead of a hopeful one.
← Back to all posts