Amazon DSP for the brand that’s outgrown Sponsored Brands

Amazon DSP is the right next move for the brand that’s hit the ceiling on Sponsored Products and Sponsored Brands and is starting to feel the share-of-voice ceiling above them. It is the wrong move for almost everyone else. The math is unforgiving when DSP is treated as “more of the same” rather than as a different lever pulling on a different curve.

This is the readiness check, the campaign architecture, and the operational discipline we deploy when we add DSP to a mid-market account.

When DSP makes sense

Three conditions, and you need all three:

Condition 1: SP and SB are at saturation. If you can keep adding budget to Sponsored Products at the same incremental ACoS, you’re not at saturation, keep adding. If your SP cost-per-incremental-conversion has been climbing for two quarters and your share-of-voice on category-anchor terms has plateaued, that’s the saturation signal. DSP picks up the audience that’s no longer responding to the next dollar of SP/SB.

Condition 2: Annual Amazon revenue ≥ $5M, comfortably. Below $5M, the minimum viable DSP spend ($30-50K/quarter to actually buy meaningful audience and maintain optimization signal) is too large a percentage of total ad budget to absorb. The brands that try DSP at $2M revenue are usually pulling spend out of SP that was working harder. We’ve seen this play out a dozen times and the answer is the same: stay focused on SP/SB until you’ve genuinely saturated those, and then revisit DSP with the budget headroom to run it correctly.

Condition 3: Catalog ready for the traffic. DSP brings cold and warm audiences to your detail pages. If the page can’t convert a less-intent-rich audience than the bottom-of-funnel SP traffic, DSP ROAS will look terrible, and the right fix isn’t bid optimization, it’s catalog work. The brands that get DSP right typically run a Cosmo-aware A+ rebuild 60-90 days before launching DSP, so the page is ready when the colder traffic lands.

The architecture that works

We run DSP with three campaign tiers. Most brands try to skip the bottom tier because it looks like “the obvious one.” That’s the mistake.

Tier 1: Defensive bidding on owned audiences. Lookalike + retargeting on people who’ve viewed your detail pages, added to cart, or purchased in the last 30-90 days. This is the highest-ROAS DSP work, typically 4-6x return, and the most boring, which is why brands tend to under-budget it. We allocate 40-50% of DSP spend here in the first quarter and don’t move that allocation until the data tells us to.

The strategic role: DSP defensive bidding holds your customers against the competitor running their own DSP at you. Without it, every other ad dollar is working harder to acquire customers your DSP-running competitor is poaching from your owned audience.

Tier 2: Targeted prospecting against in-market audiences. Amazon’s audience taxonomy includes in-market segments tied to categories and lifestyle clusters. The targeting is reasonably good if you pick the segments that match your buyer profile. We typically allocate 30-40% of DSP spend to in-market prospecting in the first quarter and tune from there.

The metric that matters here is incremental new-to-brand (NTB) conversion rate. Not ROAS. Tier 2 is supposed to be expensive on a ROAS basis, it’s buying NTB customers who’ll have lifetime value, not direct-response economics.

Tier 3: Programmatic display against contextual signals. Display placements on Amazon-owned properties (Twitch, Fire TV, IMDb, the Amazon homepage and category pages). This is the “DSP looks like a DSP” tier and it’s the one to allocate to last. Typically 15-25% of spend in the first quarter, only after Tier 1 and Tier 2 are running cleanly.

The trap brands fall into: putting most of their DSP budget in Tier 3 because it feels like the most differentiated thing DSP can do. The economics rarely work as a lead tier. It’s a complement to Tier 1 and Tier 2, not a replacement for them.

The metrics that decide whether it’s working

ROAS as a single number is a bad way to evaluate DSP. We track four metrics in parallel:

• Tier-1 ROAS: floor 4x, target 5-6x, action threshold below 3x
• Tier-2 incremental NTB rate: floor 25%, target 35%+ of conversions are net-new
• Cost per incremental conversion (vs. SP/SB baseline): the lift attributable to DSP, not the gross attribution
• Share-of-voice trajectory on category-anchor terms: the structural goal of running DSP at all

The fourth metric is the one most agencies don’t track and most brands don’t ask for. Share-of-voice is what DSP is supposed to defend and grow, if it’s not moving, the campaign is failing the strategic test even if ROAS is acceptable.

What kills DSP campaigns

Three failure patterns, in order of frequency:

Failure 1: Catalog wasn’t ready. The detail page can’t convert the colder DSP traffic at acceptable economics, and the team interprets the resulting weak ROAS as a DSP problem rather than a catalog problem. They cut DSP spend, which doesn’t fix the catalog, and the cycle repeats six months later when someone proposes DSP again.

Failure 2: Underfunded. The brand allocates $15K/quarter to DSP, the optimization signal is too weak to learn from, the campaign never finds its footing, and the budget gets cut. Minimum viable DSP spend in 2026 is roughly $30K/quarter to maintain enough optimization data; below that, you’re paying to learn and not finishing the lesson.

Failure 3: Wrong tier mix. Brand allocates 60-70% of DSP spend to Tier 3 programmatic because that’s what feels like “real DSP,” ignores Tier 1 defensive bidding, and ends up with great-looking impression numbers and bad incremental ROAS. The fix is tier reallocation, not media spend changes.

DSP and the rest of the program

DSP isn’t an island. It works best when it’s wired into the rest of the operating system. The negative-keyword harvesting cadence that makes SP work also feeds DSP audience exclusions. The catalog-as-product ownership model ensures the detail pages are ready when DSP traffic lands. The Cosmo-aware A+ rebuilds are the highest-leverage prep work for a successful DSP launch.

Brands that run DSP well are running everything else well. Brands that try DSP as a fix for an underperforming program are buying themselves a more expensive version of the same problem.

What to do this quarter

If you’re considering DSP and you meet the three readiness conditions:

Run the catalog-readiness audit first. Identify the SKUs you’d send DSP traffic to. If their A+ hasn’t been touched in 6+ months, prioritize the rebuild before the launch.

Plan for $30-50K/quarter minimum on DSP itself, plus the agency or in-house cost to run it. If that’s coming out of SP/SB budget that’s still working, don’t launch DSP, keep adding to SP. The DSP launch should be incremental budget on top of a saturated SP/SB program.

Build the campaign architecture in the order described, Tier 1 first, Tier 2 second, Tier 3 last. Resist the urge to start with the programmatic-display work, however differentiated it feels.

Track all four metrics from week one. The brands that survive their first two quarters of DSP do it because they’ve been instrumented from the start, not because they got the bid math right.

Get an audit, we’ll tell you whether DSP is the right next lever for your account or whether the higher-leverage work is somewhere else first.