Amazon DSP & Sponsored Display — Interpreting ROAS and Setting Budgets

This guide is for sellers who are already running Amazon ads and are considering expanding into Sponsored Display or Amazon DSP. This isn't about new operational secrets; rather, it's an explanation of the rules behind the numbers you see in the reports for these two ad types. We'll walk through how ROAS is calculated, which purchases are attributed to ad performance, and what settings to check before you allocate your budget. Without understanding the calculation rules, the same numbers can easily lead to incorrect conclusions.

What's the Difference Between Sponsored Display and Amazon DSP?

Sponsored Display (now integrated into 'Display ads') and Amazon DSP (Demand-Side Platform) are both forms of display advertising, but they don't share the same billing model. Unlike Sponsored Products or Sponsored Brands, which are primarily CPC (Cost Per Click) based, DSP often bills on a CPM (Cost Per Mille) model, where you pay per 1,000 impressions.

When using a vCPM (Viewable CPM) bidding strategy, Sponsored Display only charges for ads that were 'actually viewed.' In contrast, DSP generally charges CPM for total impressions, regardless of viewability. Lumping these two into the 'same display ad' category can easily lead to misunderstandings about how your bill is generated.

Amazon Attribution Model: How ROAS Numbers Are Generated

ROAS (Return on Ad Spend) in your reports is a key metric for measuring ad performance, but attribution rules lie behind those numbers. Amazon primarily uses a Last-Touch Attribution model. This means sales credit is given to the last ad a customer clicked before making a purchase.

This is coupled with an Attribution Window. This is the period after an ad interaction during which a purchase is attributed to that ad. The click-through attribution window for Sponsored Display, Sponsored Brands, and Amazon DSP is generally 14 days. This means if a customer clicks an ad and makes a purchase within 14 days, that sale is attributed to the ad.

Another concept to be aware of is the Brand Halo Effect. If a customer clicks an ad and buys a different product from the same brand (not the advertised one), that sale can still be attributed to the ad. While this can boost your ROAS, it can be misleading if you're trying to assess the efficiency of a single product's ad in isolation.

Changes in View-Through Attribution (VTA) and Off-Amazon DSP

A more significant change has occurred in View-Through Attribution (VTA). Due to recently implemented policies, the VTA method for Sponsored Brands (vCPM billing), Sponsored Display (vCPM billing), and Amazon DSP campaigns displayed within the Amazon Store has undergone a major change.

Previously, it was a broad approach where attribution was granted if a customer simply viewed an ad (without clicking) and then purchased a brand product within 14 days. Now, a machine learning-based Shopping-Signal Enhanced Last-Touch Attribution model is applied, which analyzes whether an ad view actually influenced a purchase before granting attribution. As a result, for impressions within the Amazon Store, the effective VTA window may be shorter and more dynamic than before.

While this change aims to more accurately measure actual purchase influence, it also means that from a reporting perspective, VTA-driven ROAS might appear lower than in the past. Before concluding that performance has worsened just because numbers have dropped, you should first check if the measurement criteria have changed for that period.

However, Off-Amazon DSP ads are not affected by this change and continue to use the traditional 14-day click-through / 14-day view-through attribution windows. This means the traditional measurement approach is maintained due to the nature of DSP ads appearing on various websites and apps outside of Amazon. Therefore, there can be significant differences in ROAS interpretation depending on the ad type and placement.

New-to-Brand (NTB): A Metric for Identifying New Customers

It's hard to tell if a purchase comes from a repeat customer or a new customer acquired through advertising by looking solely at Total Sales or ROAS. The metric that helps distinguish this is New-to-Brand (NTB). NTB measures purchases made by customers who haven't bought products from your brand on Amazon within the last 12 months (365 days) and were acquired through advertising.

NTB metrics are available for Sponsored Brands, Sponsored Display, and Amazon DSP campaigns, but not for Sponsored Products. The items you can check in your reports include:

New-to-Brand Orders: The number of orders placed by new customers. New-to-Brand Sales: The total sales generated by new customers. % of Orders New-to-Brand: The percentage of total orders that are NTB orders. % of Sales New-to-Brand: The percentage of total sales that are NTB sales. New-to-Brand ROAS (DSP only): The efficiency of ad spend relative to sales generated from new customers.

NTB is an official Amazon metric that measures whether your customer acquisition strategy for long-term brand growth is working. If your reason for considering Sponsored Display and DSP is new customer acquisition, then relying solely on ROAS for judgment can be misleading.

How Operations Change When Targeting NTB

If your goal is to drive NTB sales, your campaign design and evaluation criteria should align with that.

Targeting New Prospects. With Sponsored Display's 'Views Audiences' feature, you can target prospects who have viewed competitor brands or related category products, focusing on discovering new potential customers who haven't previously viewed your product detail pages. Amazon DSP allows for more sophisticated Audience Targeting and Segmentation to reach audiences with low brand awareness both on and off Amazon.

Non-Brand Keyword Strategy. In Sponsored Display's Product Targeting, targeting generic keywords not directly related to your brand name can attract new traffic and aim for NTB conversions.

Setting KPIs for Upper-Funnel Campaigns. For campaigns focused on Brand Awareness and Consideration stages—such as Sponsored Brands video ads or specific DSP campaigns—it's more appropriate to set NTB sales and NTB percentage as key performance indicators (KPIs) and allocate budgets and optimize accordingly.

Balancing ROAS and NTB. When acquiring new customers, your initial ACoS might be somewhat higher. Instead of solely pursuing a low ACoS, it's crucial to evaluate efficiency from a long-term perspective, considering NTB metrics alongside it.

DSP Billing and Viewability Settings

Since DSP often bills on a CPM basis, whether an impression was actually seen directly impacts your costs. However, the default 'Viewability' setting in the 'Add Targeting' section of DSP campaign settings is 'Allow All.' This means you're agreeing to pay for all impressions, regardless of whether they were viewable. Leaving it this way can lead to ad spend being used on impressions that aren't seen by customers, such as those below the fold or quickly scrolled past.

Changing this setting to '70% or higher' ensures you only pay for impressions that have a high probability of actually being seen by customers. However, it's not a setting that must always be changed. If your primary DSP goal is brand awareness and broad reach is a priority, 'Allow All' might be a valid consideration. If ad efficiency and ROI are paramount, focusing your spend on viewable ads aligns better with your objectives.

The important thing is not to use the default settings without reviewing them. Platform defaults aren't always optimized for a seller's cost efficiency; 'Allow All' might be a setting for maximum reach. You can check this setting by navigating to Amazon Advertising Console → Campaign Manager → DSP Campaigns → select the relevant campaign → choose a Line Item → Add Targeting → Viewability.

If you're paying for impressions regardless of viewability, your budget can deplete faster than intended, and you might miss out on high-quality impression opportunities that could lead to actual conversions. During operation, regularly check the ratio of 'Viewable Impressions' to 'Total Impressions' in your reports.

Three Common Misconceptions When Reading Reports

First, the belief that a high ROAS automatically means a good ad. A high ROAS could be driven by repeat purchases from existing customers. Therefore, you need to check the NTB percentage alongside it to determine if the campaign is contributing to new customer acquisition.

Second, the idea that you should check reports daily and immediately adjust bids as soon as a campaign starts. Amazon data can take up to 12 hours to update, and order cancellations are reflected within 72 hours. Considering the 14-day attribution window, it's better to wait at least 15 days for Sponsored Brands and Sponsored Display, and at least 8 days for Sponsored Products, to accumulate sufficient data before performing weekly optimizations. Too frequent adjustments can lead to data confusion and inefficiency.

Third, the assumption that ad report sales represent your total sales. Ad reports only show sales directly attributed to that specific ad. Ads can also boost brand awareness and influence organic sales. Therefore, checking metrics that combine both ad-attributed and organic sales, such as Total ACoS (TACoS), will give you a more complete picture.

Practical Checklist

Differentiate billing models first. Sponsored Display's vCPM and DSP's standard CPM have different criteria. Don't compare these two channels using the same yardstick. Check DSP Viewability settings. For all active DSP campaign line items, verify if 'Viewability' is set to 'Allow All.' If efficiency is your goal, adjust it to '70% or higher.' Start new campaigns with the same criteria at creation. Always look at ROAS and NTB together. When analyzing Sponsored Display and DSP performance, always check NTB orders and NTB sales percentages together. (Advertising Console → Reports) Align optimization cycles with the attribution window. Adjustments should be made at least once a week; for a 14-day window, start with data from after 15 days. Off-Amazon DSP has different criteria. Understand that off-Amazon placements still use the traditional 14-day view-through attribution, not the changed VTA rules. Align campaign goals with key metrics. If your goal is awareness, focus on NTB; if it's conversion, prioritize ACoS and conversion rate. Before increasing budget on expensive channels, confirm conversions with smaller campaigns first.