Amazon DSP cheat sheet

Amazon DSP is a demand-side platform for programmatic buying of display, video, audio, and connected-TV inventory, both on Amazon-owned properties and across the open web via third-party exchanges.

Core hierarchy: advertiser account → campaign → line item (formerly order) → creative; line items hold budget, flight, targeting, and bidding settings.

Supply comes from Amazon-owned sites and apps, Amazon publisher direct deals, and third-party exchanges/SSPs accessed through the DSP.

Amazon DSP buys primarily on a fixed-price or auction basis; auctions are typically second-price or first-price depending on the exchange, with Amazon's own auctions historically second-price.

Amazon identities (Amazon shopper IDs, hashed emails) enable first-party audience targeting that is largely unavailable on non-Amazon DSPs.

Reporting entities include campaign, line item, creative, and audience; metrics can be broken down by supply source, device, and geo.

DSP access is generally through managed service or self-service; self-service requires platform certification and minimum spend thresholds set by Amazon.

Amazon Marketing Cloud (AMC) is a separate clean-room product for SQL-based analysis of DSP and sponsored ads signals; it is not the DSP UI itself.

Audience types in DSP include in-market, lifestyle, interest, remarketing (advertiser hashed audiences), lookalike, Amazon shopper (contextual/behavioral), and third-party segments.

In-market audiences target shoppers actively researching or purchasing in a category; lifestyle segments reflect ongoing purchase behavior over a longer window.

Remarketing audiences are built from advertiser-supplied hashed data (email, phone, address) or from pixel/Amazon Ads tag events; match rates depend on data quality.

Lookalike audiences expand a seed audience to similar shoppers; seed size and recency affect reach and performance.

Contextual targeting matches page content or product category rather than user identity and does not require an identity match.

Audience overlap and deduplication matter: stacking multiple segments can inflate reach estimates; use audience overlap reports before launch.

Frequency and recency windows (e.g., 30/60/90-day) change segment size materially; shorter windows are smaller but more intent-rich.

Exclusions (e.g., recent purchasers) reduce wasted spend; exclusions apply at the line-item level and can be combined with inclusion segments.

Supply sources include Amazon-owned and operated inventory (e.g., Amazon.com, IMDb, Twitch, Fire TV), Amazon publisher direct, and third-party exchanges/SSPs.

Deal types: open auction (RTB), private marketplace (PMP) deals, preferred deals (fixed price, non-guaranteed), and programmatic guaranteed (fixed price, guaranteed volume).

Amazon Publisher Direct (APD) provides access to Amazon-owned and select third-party publisher inventory through deals negotiated with Amazon.

Deal IDs are attached to line items to restrict bidding to a specific deal; without a deal ID the line item bids in the open auction.

Supply Source targeting lets buyers include or exclude specific exchanges, domains, apps, or Amazon-owned properties.

Viewability, brand-safety, and fraud-prevention filters (e.g., IAS/DoubleVerify integrations) can be applied as supply filters.

Inventory types map to creative formats: display, online video, streaming TV/OTT, audio, and out-of-home (where available).

Deals may carry floor prices; bids below the floor do not win, and deal pacing differs from open-auction pacing.

Campaign level holds advertiser, budget cap (optional), flight dates, and frequency caps; line items inherit and can override some settings.

Line item settings include budget, flight, goal (e.g., reach, conversions, CPC), bid strategy, targeting, supply, creative rotation, and pacing.

Bid strategies include fixed bid, dynamic bids (up or down), and rule-based or goal-based bidding where available; the strategy determines how the base bid is adjusted at auction time.

Base bid is the maximum CPM the line item will pay before strategy adjustments; effective bid is what is actually submitted.

Pacing options (even vs. ASAP) control how budget is spread across the flight; ASAP front-loads spend and can exhaust budget early.

Frequency caps limit impressions per user per time window and can be set at campaign or line-item level.

Budget recommendations and forecasting tools estimate reach and impressions based on targeting, bid, and flight; estimates are directional.

Line items can be duplicated to test bid, audience, or creative variations; naming conventions should encode test variables for reporting.

Common formats: standard display (static, responsive), rich media, online video (in-stream/outstream), streaming TV (OTT), audio, and sponsored display-style native placements.

Responsive e-commerce ads and other Amazon-native creatives pull product imagery, price, and ratings from the product detail page; the ASIN must be in the advertiser's catalog.

Video creative specs include resolution, aspect ratio, file size, duration, and bitrate; streaming TV typically requires HD video and may disallow interactive elements.

Creative approval is required before delivery; disapprovals cite policy reasons and must be fixed and resubmitted.

Creative rotation settings control whether creatives serve evenly or optimize toward the best performer.

Click-through URLs must be secure (HTTPS) and land on a functional page consistent with the ad's claims.

Third-party ad tags (e.g., VAST for video, HTML5 for display) are supported; tag macros pass targeting and measurement parameters.

Interactive and shoppable formats (e.g., add-to-cart, product carousel) can shorten the path to purchase but require catalog and policy compliance.

Key metrics: impressions, clicks, CTR (clicks ÷ impressions), CPM (cost ÷ impressions × 1000), CPC (cost ÷ clicks), CPA (cost ÷ conversions), ROAS (sales ÷ spend).

Attribution windows (e.g., 1-day, 7-day, 14-day, 30-day) determine which conversions are credited to an impression or click; longer windows capture more but dilute causality.

View-through conversions (VTC) credit conversions to impressions without a click; click-through conversions credit clicks.

Amazon Attribution measures off-Amazon performance for advertisers using Amazon ads to drive traffic to their own sites.

Amazon Marketing Cloud enables custom SQL queries across DSP and sponsored ads signals for path-to-conversion and overlap analysis.

Brand Lift and Brand Search Lift studies measure upper-funnel impact (awareness, consideration, search behavior) via survey or search-lift methodology.

Reports can be scheduled and broken down by line item, creative, audience, supply source, and geo; exported data should be reconciled against billing.

Third-party measurement vendors (e.g., viewability, verification, MRC-accredited) can be integrated via tags for independent validation.

Diagnose underperformance by segment: low impressions → bid/targeting/supply constraints; low CTR → creative/audience mismatch; low conversion → landing page or audience quality.

Bid optimization: raise bids on high-converting segments, lower or exclude low-performing ones; adjust based on marginal CPA vs. target.

Budget reallocation: shift spend toward line items or audiences with lower CPA or higher ROAS, subject to statistical significance.

Frequency management: excessive frequency raises CPM waste and can drive negative sentiment; monitor frequency reports and cap accordingly.

Creative testing: run controlled A/B tests with even rotation, sufficient sample size, and consistent targeting to isolate creative effects.

Pacing checks: verify line items are on pace to spend full budget by flight end; adjust bids or budgets early rather than at flight end.

Supply optimization: exclude low-viewability or low-converting domains/apps and lean into deals or supply sources that perform.

Statistical caution: avoid reacting to small-sample fluctuations; require adequate impressions/conversions before making bid or budget changes.

Prohibited content includes (among others) illegal products, tobacco and vaping, unapproved pharmaceuticals, weapons, and adult content; restricted categories require pre-approval.

Creative must be accurate and not misleading; claims require substantiation, and pricing/availability shown must match the landing page.

Advertisers must have rights to use all creative assets (images, logos, music, trademarks) and must not infringe third-party IP.

Data use is governed by the Amazon Ads agreement: advertiser-supplied audiences must be lawfully collected with proper consent, and PII must be hashed before upload.

Brand-safety controls include supply-source exclusions, third-party verification integrations, and content-category filters.

Amazon prohibits ads that mimic Amazon's own UI or misrepresent an Amazon endorsement; use of Amazon trademarks is restricted.

Policy violations can result in creative disapproval, line-item pause, account suspension, or removal from the DSP.

Compliance with regional regulations (e.g., GDPR/CCPA consent, ad-format rules) is the advertiser's responsibility and is enforced through Amazon's data policies.