After managing Amazon DSP budgets for dozens of brands over the past three years, we have watched the same pattern repeat. A brand crosses $100K/month in revenue. Their agency or rep proposes DSP. They allocate $5K, $10K, $15K to “test it.” Six months later, the campaigns are running, the dashboards look healthy, and nobody can answer whether it actually moved the business.
DSP is not a default-on lever for $100K+/month brands. It is a specific tool that pays off in specific conditions. When those conditions are met, DSP is the highest-ROI ad channel on Amazon. When they aren’t, you are funding programmatic display impressions that would have happened anyway through your Sponsored Products spend.
Here is the framework we use to decide, structure, and measure DSP for our clients.
When DSP Pays Off — The Three Conditions
We only recommend DSP when at least two of these three conditions are present:
1. You have a retargetable audience worth retargeting. This means meaningful detail page view (DPV) volume — typically 30,000+ DPVs/month across the brand. Below that threshold, your retargeting pools are too small for DSP minimums, and your CPMs are wasted reaching the same 4,000 people 18 times.
2. You have repeat-purchase or cross-sell motion. Subscribe & Save, replenishment categories, or multi-SKU catalogs where buying SKU A predicts buying SKU B. DSP shines at re-engaging existing customers and pushing cross-catalog discovery. If you sell a one-time-purchase $40 product, DSP is mostly a billboard.
3. You are saturated on Sponsored Products. TACoS is creeping up, ACoS on top campaigns is rising 15-25% YoY at the same impression volume, and you cannot find new keywords without bidding into negative ROI. DSP gives you a different supply pool. Without that saturation, you are paying display CPMs to reach people Sponsored Products would have reached cheaper.
If none of these are present, do not run DSP. Spend the budget on creative, listing optimization, or expanding Sponsored Products structure. That is the single most counter-intuitive piece of advice we give brands at this revenue tier.
The Spend Threshold That Makes DSP Work
DSP minimums in 2026 sit at $10K/month for self-service through a managed partner, or $35K/month direct with Amazon. Below $10K, your reach frequency is too thin and audience overlap is too high to drive measurable lift.
Our minimum recommendation: $15K/month for 90 days before reading any signal. At $5-10K/month, you are noise-trading against Sponsored Products spend. At $15K+ for at least 12 weeks, DSP campaigns reach enough impression volume to separate signal from variance.
Brands spending $5K/month on DSP and looking for ROAS reads after 6 weeks are the most common failure case we see. The math is not there. Either commit to the threshold or skip DSP entirely.
Campaign Structure That Actually Works in 2026
The default DSP structure most agencies set up — one retargeting line, one prospecting line, “lifestyle” audience overlay — is lazy. Here is the structure we run for $100K-$500K/month brands:
Tier 1 — Detail Page Retargeting (40% of spend). Audiences split by recency: 0-7 days, 8-30 days, 31-90 days. Different bid strategies and creative for each. The 0-7 day pool converts at 3-5x the 31-90 day pool. Treating them as one audience kills the campaign’s apparent ROI.
Tier 2 — Brand Halo / Cross-Sell (25% of spend). Audiences of past purchasers of SKU A targeted with creative for SKU B. Highest-ROAS line in most catalogs. We routinely see 8-12 ROAS on this line for brands with strong cross-sell logic.
Tier 3 — Lifestyle / In-Market Prospecting (25% of spend). Amazon-defined audiences (in-market for category, lifestyle segments). Needed for top-of-funnel but should never exceed 25% of DSP spend for performance brands. This is where most “DSP isn’t working” stories come from — it’s mostly Tier 3 spend with no Tier 1 or 2 underneath.
Tier 4 — Competitor Conquesting (10% of spend). Audiences of viewers of named competitor products. Aggressive bids, sharper creative, comparative messaging. Read separately from other lines because it has different attribution behavior.
The 40/25/25/10 split is a starting point. We move spend toward whichever tier shows the strongest 30-day ROAS, but we do not collapse the structure. Each tier needs to be measured independently.
Creative Requirements DSP Will Punish You For Ignoring
DSP creative is not Sponsored Products creative. We see brands recycle their hero image and call it a DSP ad. The performance gap between purpose-built DSP creative and recycled assets is consistently 2-3x in CTR and 1.5-2x in CVR for the same audience.
Required for DSP creative in 2026:
- Multiple aspect ratios. 300×250, 728×90, 160×600, 300×600, 970×250, plus mobile variants. Not optional — Amazon will under-deliver against the placements you don’t size for.
- Animated and static versions. Animated wins on prospecting, static wins on retargeting in our data. Build both.
- Audience-specific messaging. A retargeting ad for a 0-7 day pool says “complete your purchase” — a prospecting ad says “discover the category.” Same hero shot, different copy and CTA.
- Mobile-first design. Over 70% of DSP impressions in our accounts are mobile. Designs that lose impact at thumb-scroll size are 60%+ of the creative we audit.
Brands that allocate $50/month for DSP creative and $20K/month for DSP media spend are setting fire to the media spend. Plan 8-12% of DSP budget for creative production if you want the channel to perform.
Measurement: What “DSP ROAS” Actually Means
This is where most DSP reads fall apart. Amazon DSP attribution is a 14-day view-through, 14-day click-through model by default. That means an ad can serve an impression, the user converts 13 days later from a search ad, and DSP gets credit. Read straight DSP ROAS reports and you will overstate DSP’s contribution by 30-60%.
The right measurement framework:
- Incrementality, not attribution. Run holdout markets or holdout audience pockets when budget allows. Read brand-level revenue lift, not click-level ROAS.
- TACoS at brand level. If DSP is working, total ad cost as % of revenue should bend down or stay flat as you add DSP spend on top of existing channels. If TACoS rises and total revenue rises proportionally, you are buying revenue at the same efficiency you already had.
- New-to-brand metrics. Amazon’s new-to-brand reports separate first-time buyers from existing customers. DSP’s job in your stack is partly to drive new-to-brand. If new-to-brand % isn’t moving, DSP is mostly retargeting people who would have converted anyway.
- Repeat purchase rate. For replenishment brands, DSP retargeting should lift 90-day repeat rate. If it isn’t, the retargeting cadence is off.
Brands that read only the DSP dashboard and see “5 ROAS” stay in DSP for years buying their own existing customers. Read the holistic numbers or don’t run DSP.
When We Tell Clients to Pause DSP
We have pulled clients out of DSP three times in the past year. The common pattern:
- TACoS rising despite DSP spend
- New-to-brand rate flat or declining
- Repeat purchase rate flat
- DSP dashboard ROAS looks “fine” (3-5 range) but holdout testing shows minimal incremental lift
When all four are present, DSP is buying clicks that would have happened anyway. We pause DSP, redirect the budget to Sponsored Brands video and creative production, and watch incremental revenue rise even as reported ad ROAS drops on paper.
DSP is a tool, not a default. The brands that win with DSP have a clear job for it — retarget high-intent DPVs, cross-sell existing customers, conquest a specific competitor. The brands that lose with DSP run it because their agency suggested it.
FAQ
What’s the minimum revenue level to consider DSP?
We typically don’t recommend DSP below $100K/month, and even then only with the three conditions met. Below that, Sponsored Products and Sponsored Brands are still where the marginal dollar performs best.
Self-service DSP or managed?
Self-service through a competent partner gives you better economics if you have $20K+/month in DSP spend. Managed direct with Amazon makes sense at $50K+/month or for brands without an experienced DSP operator on staff.
How long until DSP ROAS reads?
Minimum 12 weeks at full spend. Reading at 4-6 weeks is the most common reason brands pull DSP before it works (or keep DSP running when it isn’t).
Does DSP help with off-Amazon traffic?
Yes, DSP can place ads on Amazon-owned properties (IMDb, Twitch, Fire TV) and on the open web through Amazon’s exchange. Off-Amazon DSP is a different beast with different math — that is a separate playbook we run only for brands above $500K/month.
Should we use DSP for new product launches?
Mostly no. New launches need review velocity, ranking momentum, and Sponsored Products presence first. DSP works better at month 4+ post-launch when you have DPV volume and review base to retarget against.
If you’re looking for a team that manages every lever — creative, advertising, and operations — Velocity Sellers works with brands doing $100K+/month on Amazon. Contact us for a free account audit.
