Amazon DSP common mistakes
Amazon DSP · Understand the traps and how to avoid them.
Treating Amazon DSP as a retail-only ad console and assuming campaigns can only target Amazon shoppers or only run on Amazon-owned sites.
Candidates come from sponsored ads experience where inventory is Amazon's own marketplace; they carry that mental model into DSP and miss that DSP buys programmatic inventory across the open internet plus Amazon-owned properties.
What to do instead
Memorize the two-sided structure: supply (Amazon-owned properties, Amazon Publisher Services/exchanges, third-party exchanges and SSPs) and demand controls (audiences, deals, bid strategies). Be able to state which inventory sources are Amazon-exclusive versus open-web before answering any supply or targeting question.
Confusing Amazon-built audiences (in-market, lifestyle, remarketing) with advertiser-uploaded audiences and third-party segments, and picking the wrong one for a stated objective.
All audience types appear in the same targeting UI, so candidates skim labels and default to whichever segment name sounds closest to the prompt instead of mapping audience type to data source and match method.
What to do instead
Build a comparison table from memory: Amazon first-party (shopper signals), advertiser first-party (pixel/remarketing, hashed CRM), and third-party segments. For each, note data origin, typical use case, and whether identity resolution is needed. Practice classifying five scenario prompts per session.
Selecting a deal type by name rather than by the access and pricing mechanics it implies, e.g., treating a private marketplace deal as guaranteed inventory.
Deal terminology overlaps with general programmatic vocabulary, and candidates answer from surface familiarity instead of recalling which deal types guarantee volume, which set fixed or floor prices, and which are open auction.
What to do instead
Learn each deal type by three attributes: is inventory guaranteed, is price fixed or a floor, and is the deal invitation-only or open. Then answer deal questions by matching the scenario's requirement (guaranteed delivery vs. price control vs. reach) to those attributes, not to the label.
Setting line item bid strategy and budget without checking that they align with the campaign goal and flight, then reporting the mismatch as a platform limitation.
Candidates memorize bid strategy names in isolation and never practice the decision sequence (goal, then bid strategy, then budget/bid, then pacing), so they pick a strategy that optimizes for the wrong outcome.
What to do instead
Rehearse the setup order explicitly: objective, inventory/deal, audience, bid strategy matched to objective, budget and flight, creative. For each bid strategy, write one sentence on what it optimizes and when it is inappropriate.
Assuming a creative format can run anywhere and ignoring size, placement, and device constraints when recommending a format for a scenario.
Candidates learn formats as static images of ad units rather than as specifications with inventory and device requirements, so they recommend an interactive or video format for a placement that cannot serve it.
What to do instead
For each major format, record supported devices/environments, typical placements, and required creative assets. When a scenario names a device or placement, filter formats by those constraints first, then choose by objective.
Reading attribution and reporting metrics as if all conversions are incremental, and comparing metrics across different attribution windows or report types as if equivalent.
Dashboards present conversions, view-through, and click-through side by side; candidates report the largest number without checking window length, attribution model, or deduplication rules.
What to do instead
Before interpreting any metric, state its definition, attribution window, and whether it is modeled or observed. Never compare figures built on different windows. Practice explaining why a view-through conversion differs from a click conversion and how that affects reported ROAS.
Diagnosing underperformance by changing many levers at once (audience, bid, creative, budget) in a single optimization step.
Candidates treat optimization as a checklist of fixes rather than an experiment, so they cannot attribute the result and may mask the real cause of delivery or performance issues.
What to do instead
Use a structured diagnostic: check delivery/pacing first, then inventory and deal health, then audience size and frequency, then creative, then bid and budget. Change one variable per cycle and define the success metric and evaluation window before making the change.
Answering policy questions by intuition about what seems acceptable advertising rather than by Amazon's specific creative and content restrictions.
General advertising common sense differs from platform rules on prohibited content, creative claims, and restricted categories, and candidates have not studied the actual policy categories.
What to do instead
Study policy by category (prohibited products/content, restricted categories, creative and landing page requirements, data use). For any policy scenario, identify the category first, then the specific rule, then the compliant alternative.
Treating brand safety and suitability as a single toggle and ignoring how third-party verification and exclusion controls interact with supply source selection.
Candidates conflate brand safety (avoiding harmful content) with suitability (contextual fit) and assume enabling one setting covers all inventory, including deals and Amazon-owned properties.
What to do instead
Separate the concepts: safety controls (blocklists, verification vendors, inventory exclusions) versus suitability preferences (contextual targeting, curated supply). Map which controls apply to open exchange versus deal-based inventory, and check them before finalizing supply strategy.
Confusing campaign-level and line-item-level settings, and placing targeting, bid, or budget at the wrong hierarchy level.
The DSP structure (campaign, line item, creative) mirrors other ad platforms but with different inheritance rules, and candidates transfer assumptions from sponsored ads or social platforms.
What to do instead
Draw the hierarchy from memory with the settings owned at each level. For any setup scenario, state the level at which each setting lives before answering, and note which settings are inherited versus overridden at the line item.