Amazon DSP self-service has stopped being an enterprise-only tool. After managing DSP across roughly 90 brands in the $100K–$2M/month range over the last 18 months, we can tell you exactly when self-service DSP earns its slot in the mix — and when it is a tax on growth pretending to be a brand-building lever.
Most mid-size brands we audit are either over-leaning on DSP (running $25K/month and hoping the attribution lifts the ROAS story) or avoiding it entirely (when adding $8K–$12K of DSP would actually move the business). Here is the framework we use.
The honest spend threshold
Amazon does not enforce a hard self-service minimum on the public-facing tools, but the practical floor is $10,000–$15,000 per month through a managed self-service partner, sustained for at least 90 days before you read any signal.
Below that:
- Frequency caps on retargeting audiences kick in before the campaign reaches statistical confidence
- Audience reach is too thin to separate from Sponsored Ads pull-through
- Optimization passes can’t actually optimize — there is not enough event volume
Above $15K/month, DSP campaigns reach the impression density where the platform’s bidding model has room to work and you can read 90-day signal cleanly. Below $10K, you are paying retainer + media to learn what you would learn anyway from a Sponsored Display Audiences test.
If you are spending $50K/month total on Amazon advertising, $10K–$15K of that in self-service DSP is a defensible 10–15% slice. If you are spending $20K/month total, DSP is not your next dollar — Sponsored Display and Sponsored Brands are.
The 5 conditions that make DSP worth running
We pre-qualify DSP candidates against these. A brand needs at least 3 of 5 before we recommend pulling the trigger.
1. Branded search defense is healthy
Branded Sponsored Brands and Sponsored Products are already running at 65%+ impression share. If your branded search is leaking to competitors, fix that with Sponsored Brands defense first — adding DSP retargeting on top of a broken branded-search position is throwing dollars at a downstream problem.
2. Listing CVR is above category median
DSP drives traffic to a PDP. If the PDP is converting below the category median, DSP traffic converts even worse than your Sponsored Ads traffic, because DSP audiences are colder. Fix the listing first. DSP after.
3. You have at least 4–6 SKUs with $20K+/month revenue
Audience sizes inside DSP get useful when the brand has 4+ products that share a customer base. With 1–2 SKUs, retargeting audiences (in-market, lifestyle, lookalike) don’t have enough variety to populate without overlap, and frequency caps cannibalize each other.
4. You are willing to commit 90 days minimum
DSP needs a learning period. Brands that pause DSP at day 35 because “ROAS isn’t there yet” almost always pause it right before the audience seasoning kicks in. We have a hard rule with clients: 90 days minimum, no pauses, or do not start.
5. You have new-to-brand (NTB) as a real goal — not just a vanity line
DSP earns its slot when NTB matters. If your business is fundamentally a repeat-purchase business and 80% of revenue is returning customers, DSP is a slower lever than email retention. If you are trying to grow NTB share from 18% to 30%, DSP is one of the only tools that pulls it.
The 4 campaign structures we run
Self-service DSP for a $250K/month brand typically runs 4 campaign structures concurrently. Total spend: $12K–$25K/month split across them.
Retargeting (~35% of DSP budget)
Audiences: viewed product, viewed similar product, added to cart not purchased, viewed but didn’t add. Frequency-capped to 3–4 impressions per user per week. This is the cleanest ROAS campaign and the one we report on first.
Typical performance: 4–7x ROAS, attributing on a 14-day post-view + 14-day post-click window.
Competitor conquest (~25%)
Audiences built from people who viewed specific competitor ASINs. Highest CPM in the structure ($14–$22). Wins when your offering has a clear differentiator over the competitor — fails when you are commodity vs commodity.
Typical performance: 1.8–3.2x ROAS, but high NTB rate (often 45–60% of conversions).
In-market / lifestyle prospecting (~25%)
Amazon’s first-party audiences for category in-market shoppers. Best for newer launches and category expansion. Lower CPM ($6–$11) but lower CTR. The campaign you tune the longest.
Typical performance: 1.4–2.4x ROAS, but feeds the retargeting funnel.
Owned-audience (existing customer) re-engagement (~15%)
Audiences built from past purchasers — re-engagement for refill / next-product cross-sell. Lowest CPM, highest ROAS.
Typical performance: 8–14x ROAS, but capped on volume.
The 4 burn-cash patterns we audit out of accounts
These show up in roughly 60% of mid-size accounts we audit when DSP was set up by a generalist or run on auto-pilot for 6+ months.
Pattern 1: Display-only inventory, no Streaming TV or OLV
The brand runs DSP entirely on standard display banner inventory. Display is the cheapest, lowest-attention DSP inventory and the most heavily fraud-adjacent slot. A real DSP strategy has 15–25% in Online Video (OLV) or Streaming TV inventory for awareness, with the rest in display retargeting. Display-only DSP is glorified retargeting at premium CPM.
Pattern 2: 14-day attribution windows used to justify 28-day spend
The agency reports DSP ROAS on a 30-day post-view + 14-day post-click attribution window when the brand’s actual purchase cycle is 7 days. The “attributed” sales were going to happen anyway. We re-cut attribution to 7-day post-view for short-cycle consumables and the “8x ROAS” campaign drops to 2.1x.
Pattern 3: No frequency capping
Retargeting campaigns running at 12+ impressions per user per week. Spend bloats, customer experience suffers, and the unique reach number plateaus around day 8 of the flight. Frequency cap at 3–4 per user per week and reallocate the saved spend to prospecting.
Pattern 4: AMC queries never actually run
Amazon Marketing Cloud is bundled with the DSP buy but the brand has never run a path-to-purchase or audience-overlap query. The 4 queries we wrote about here are what makes DSP a real strategic tool vs an expensive ad-server. If your DSP agency cannot show you AMC pulls in the monthly report, ask why.
Self-service vs managed service — the real distinction
Amazon’s managed-service DSP requires a $50K/month minimum, which prices out most mid-size brands. Self-service through a partner agency starts at the $10K floor and runs all the way up to $200K+/month.
The functional difference is not the platform — both access the same inventory, the same audiences, the same bidder. The difference is whose team is pulling the levers:
- Managed service: Amazon’s team runs the campaigns. Less flexible. Slower to iterate. Fixed quarterly business reviews.
- Self-service through partner: Your agency (or in-house) team operates the platform. Daily optimization, custom audience overlap analysis, cross-channel measurement, and full AMC query access.
For brands in the $100K–$500K/month range, self-service through an experienced partner is almost always the right structure. The flexibility wins.
How DSP fits with Sponsored Ads — the 4-tier model
DSP does not replace Sponsored Ads. It sits on top of them. We run this 4-tier structure for most mid-size accounts:
- Tier 1 — Branded defense (Sponsored Brands + Sponsored Products on brand terms): 8–12% of total ad spend. Protects existing demand.
- Tier 2 — Non-branded acquisition (Sponsored Products + Sponsored Brands on category terms): 50–60%. The bulk of measurable acquisition.
- Tier 3 — Sponsored Display retargeting and audiences: 10–15%. Bridges Sponsored Ads to DSP — same audiences, lower CPM, less inventory access.
- Tier 4 — DSP (display + OLV + retargeting + conquest): 15–25% once the brand crosses the $100K/month ad spend threshold.
If you are running DSP without Tier 1 and Tier 2 buttoned down, you are skipping the cheaper-acquisition layers in favor of the more expensive one. Get Sponsored Ads right first. Add DSP when the math earns it.
The 90-day signal read
Brands that successfully add DSP in 2026 hit these milestones:
- Day 30: Frequency caps stabilized, audience seasoning visible, retargeting ROAS readable
- Day 60: First AMC query results — usually path-to-purchase showing DSP touch in 18–35% of converting paths
- Day 90: NTB% rises measurably (typical lift: 4–9 percentage points), branded search volume up 8–15%, retargeting ROAS at 4x+, prospecting at 2x+
Brands that fail to hit these milestones almost always show one of the 4 burn-cash patterns above. The fix is not “spend more on DSP” — it is “fix the structure.”
FAQ
Can I run DSP self-service without an agency?
Technically yes — Amazon has opened more direct self-service tooling in 2026. Practically, no, unless you have a dedicated in-house programmatic specialist. The platform rewards expertise. A solo seller running DSP part-time is going to underperform a $9K/month managed self-service partner running it full-time.
What’s the right starting budget?
$10K–$12K/month for 90 days. If the math at day 90 supports scaling, ramp to $15K–$25K. Do not start at $25K — you do not yet have audience data to deploy it.
Should I run Streaming TV inventory?
If your brand fits the category (consumer products with broad appeal, not B2B / niche industrial), allocate 10–15% of DSP spend to Streaming TV. The Sponsored TV self-serve tooling that opened in 2026 has lowered the floor for this.
How do I measure DSP success against Sponsored Ads?
Use AMC path-to-purchase queries — not the platform-reported ROAS in isolation. DSP’s value shows up as assists, NTB rate, and time-to-purchase, not just direct ROAS.
—
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.
