Blog
/
Why Your Amazon DSP Isn’t Spending and Why Your DSP ROAS Looks Low
Why Your Amazon DSP Isn’t Spending and Why Your DSP ROAS Looks Low
Nivesh J.
KEY TAKEAWAYS
DSP underspend can be caused by a low maximum average CPM, restrictive frequency caps, strict viewability targets, small audiences, and missing creative sizes, which can limit campaign delivery.
Check delivery before increasing budgets: Use Amazon DSP's delivery rate, budget at risk, and pacing alerts to identify what is preventing campaigns from spending.
Understand the 2026 attribution changes: Amazon's January 1, 2026, view attribution update can lower reported ROAS without necessarily reducing actual sales.
Allow time for attribution data to mature: Evaluate DSP performance using data that is at least 14–21 days old to avoid misleading results.
Track ASINs and separate audiences: Untracked ASINs and blended prospecting and remarketing campaigns can distort reported ROAS.
Measure ROAS against break-even targets: Set ROAS benchmarks based on contribution margin rather than relying on industry averages.
Monitor marginal ROAS when scaling: Additional spend can generate diminishing returns, even when average ROAS looks healthy.
Fix delivery, measurement, and performance in order: Troubleshoot campaign settings first, validate attribution second, and optimize profitability last.
Quick answer: An Amazon DSP order usually underspends because a setting limits how many auctions it can enter or win, such as a low maximum average CPM, a tight frequency cap, strict viewability targeting, a small audience or missing creative sizes. DSP ROAS usually looks low because of how it is measured, through a shorter view window since January 1, 2026, immature recent data, untracked ASINs or mixed audiences, or because extra spend has hit diminishing returns. Check delivery first, then measurement, then performance, in that order.
Amazon changes DSP reporting often, so confirm labels in your own console.
Why Is My Amazon DSP Campaign Not Spending Its Budget?
A DSP order does not spend its budget just because you set it. It spends only when it finds eligible impressions that match your targeting and then wins the auction at your bid, so any setting that shrinks eligible supply or caps your bid can leave budget unspent.
Amazon’s own tools point to the same handful of causes. The table groups them by the symptom you will see.
Symptom | Likely cause | Where to look |
|---|---|---|
Order is on track to spend less than its budget | Maximum average CPM too low, frequency cap too tight, viewability target too strict | Pacing alert and line item settings |
One line item delivers, others do not | Missing creative sizes for the available placements | Creative recommendation icon on the line |
Delivery stalls after a few days | Audience too small for the frequency cap | Audience size against cap |
Deal or private marketplace line barely bids | Bid below deal floor, audience, location or viewability mismatch | Deal breakdown and missed bid reasons |
Spend is steady but below budget | Losing auctions at your current bid | Bid, supply and win rate |
Funnel showing how an illustrative 600,000 eligible daily impressions shrink to 72,000 won impressions and 432 dollars of spend against a 1,000 dollar daily budget
Figure 1: Illustrative delivery funnel. Each setting narrows the path from eligible supply to spend.
How Do You Check Whether an Amazon DSP Order Is Underdelivering?
Use Amazon’s delivery rate and budget at risk columns, then confirm with your own pace check. Amazon defines delivery rate as how fast an order is spending relative to its budget and the progress of the flight, capped at 100%, and budget at risk as the budget that may go unspent if the current pace continues.
Amazon says delivery rate is available from day two of the flight, is recalculated several times a day, and at launch was not available for impression based orders. It does not publish the formula or alert thresholds, so run a simple check of your own.
Pace check = Spend to date ÷ (Order budget × Share of flight elapsed)
Required daily spend = (Order budget minus Spend to date) ÷ Days left
Example (illustrative). A $30,000 order is on day 15 of 30 and has spent $9,000. Expected spend is $15,000, so the pace check is 9,000 ÷ 15,000 = 0.60. To finish, the order needs (30,000 minus 9,000) ÷ 15 = $1,400 a day, against an average of $600 so far. Spend would need to more than double, so raising the budget is not the fix. Find what is blocking delivery.
Amazon also runs pacing alerts for self service advertisers. The alert says an order is underpacing and on track to spend less than its budget, and ranks the line item fixes by impact.
Which Line Item Settings Stop Amazon DSP From Spending?
Three settings cause most underspend alerts: a maximum average CPM that is too low, a frequency cap that is too restrictive and a viewability target that is too strict. Amazon’s pacing alert recommends raising the CPM or the cap, or lowering the viewability target, to the values DSP suggests.
Setting | How it blocks spend | Fix Amazon suggests |
|---|---|---|
Maximum average CPM | Your bids sit below the clearing price for the inventory you target | Raise it to the suggested value |
Frequency cap | The same shoppers hit the cap, and no new impressions are eligible | Raise the cap to the suggested level |
Viewability target | Too few placements meet your threshold | Lower the target to the suggested setting |
Creative sizes | Placements need sizes you did not supply | Add the suggested approved creatives |
On creative sizes, PPC Land’s report on the recommendation explains that campaigns built without the right size variations can underpace, and that DSP can add creatives already approved on other lines in the same order with one click.
Amazon calls these recommendations guidance and leaves the final decision with you. Treat each as a trade, because a higher CPM or a looser viewability target also changes what you pay and what you reach.
How Small Can an Audience Be Before an Amazon DSP Order Cannot Spend?
A frequency cap puts a hard ceiling on spend for a defined audience, no matter how large the budget is. The ceiling is the audience size times the cap, spread over the cap period, times your CPM.
Frequency ceiling per day = Audience size × Cap per week ÷ 7 × CPM ÷ 1,000
Example (illustrative). A remarketing audience of 50,000 shoppers, capped at 3 impressions a week, with a CPM of $8, can absorb at most 50,000 × 3 ÷ 7 = about 21,400 impressions a day. At $8 CPM that is about $171 a day. A $500 daily budget on that line can never fully spend, however high you bid.
The same logic applies to supply. If you know roughly how many eligible impressions a day your targeting reaches, you can estimate the supply ceiling too.
Supply ceiling per day = Eligible daily impressions × Win rate × CPM ÷ 1,000
Example (illustrative). With 600,000 eligible impressions, a 12% win rate and a $6 CPM, the ceiling is 600,000 × 0.12 × 6 ÷ 1,000 = $432 a day. A $1,000 daily budget delivers 43%. Raising the bid lifts the win rate. Widening the audience lifts the eligible impressions. Raising the budget alone changes nothing. These inputs are assumptions you must estimate from your own reports, since DSP does not publish eligible supply per line.
Why Is My Amazon DSP Deal or Private Marketplace Line Not Delivering?
A deal line underdelivers when its bids are rejected before the auction or lose in it, and Amazon lists 15 reasons for missed bids on deal inventory. They fall into six groups, and the missed bid columns and Deals page show which ones affect your line.
The Amazon DSP delivery troubleshooting for deal inventory page names the reasons. The grouping below is this guide’s, not Amazon’s.
Group | Amazon missed bid reasons |
|---|---|
Traffic validity | Traffic quality filter, ads.txt filter, incorrect app ID or URL, no user ID |
Privacy and consent | Viewer privacy settings |
Targeting | Audience mismatch, viewability mismatch, location mismatch |
Frequency and creative | Order frequency limit, line item frequency limit, creative size limit |
Bidding and price | Bid below deal floor, max bid too low |
Auction outcome | Lost internal auction, lost external auction |
Amazon’s page gives the reasons and the metrics but no fixes. In practice, bid and targeting reasons are yours to change, traffic validity and privacy reasons are not, and a high share of those means the deal itself has limited usable supply.
In What Order Should You Troubleshoot Amazon DSP Underspend?
Work from the cheapest check to the most expensive one: status and budget, then bid, then targeting, then frequency, then creative, then supply. Changing several settings at once hides which one worked, so change one, wait for a few hundred auctions of data, and compare.
Step | Check | Change if |
|---|---|---|
1 | Order, line item and creative status | Anything is paused, rejected or out of flight |
2 | Budget and pacing | Delivery rate is low and budget at risk is flagged |
3 | Maximum average CPM | The pacing alert recommends a higher value |
4 | Frequency cap against audience size | The daily ceiling is below the daily budget |
5 | Viewability target | The alert recommends a lower target |
6 | Creative sizes | The size recommendation icon appears |
7 | Audience and location breadth | The supply ceiling is below the daily budget |
8 | Deal missed bid reasons | A deal line shows high bid or targeting reasons |
Ladder of eight DSP underspend checks from status and budget down to deal missed bid reasons
Figure 2: The troubleshooting order for DSP underspend.
Why Does My Amazon DSP ROAS Look Low?
DSP ROAS looks low for one of four reasons: the measurement rules changed, the data is still maturing, the setup misses sales it should count, or the spend has reached diminishing returns. Only the last one is a performance problem, so rule out the first three before you cut anything.
Cause | Type | Quick test |
|---|---|---|
View attribution change from January 1, 2026 | Measurement | Compare standard purchases with Purchases (all views) |
Recent days still filling in | Measurement | Look at data older than 14 to 21 days |
Untracked ASINs or mixed audiences | Setup | Check tracked ASINs and audience overlap |
Diminishing returns at higher spend | Performance | Compare marginal ROAS with break even ROAS |
Did the 2026 View Attribution Change Lower My Amazon DSP ROAS?
Very likely, if your ads serve on Amazon Store inventory and relied on views. From January 1, 2026, Amazon moved view conversions on that inventory to a shopping signal enhanced last touch model with a shorter window, so fewer views are credited and reported ROAS can fall without any change in real sales.
Amazon’s view attribution update says the change covers Amazon DSP campaigns serving in the Store and Sponsored Brands and Sponsored Display campaigns billed on vCPM, and leaves click attribution unchanged. It does not state the new window length. The legacy 14 day view numbers remain in metrics labelled Purchases (all views). PPC Land’s coverage adds that automated rules tied to ROAS targets may react to the shift, so review those thresholds.
View attribution gap = (Purchases (all views) minus Standard purchases) ÷ Purchases (all views)
Example (illustrative). If an ad group shows 1,000 Purchases (all views) and 780 standard purchases, the gap is 220 ÷ 1,000 = 22%. Compare 2026 standard results with 2025 results only after you adjust for that gap, or you will blame the campaign for a rule change. Our guide to Amazon DSP attribution walks through the rules in full.
Is My Amazon DSP ROAS Low Because the Data Is Still Maturing?
Often, yes, if you are reading the last two weeks. Amazon’s attribution documentation describes a 14 day lookback and reports conversions against the date of the ad interaction, so a sale made on day 12 appears under day 1, and recent days always look weaker than they will end up.
Use a settled window when you judge ROAS. A simple rule from this guide, which is judgment and not an Amazon rule, is to read a report date only after 21 days, and to compare the same age of data when you compare campaigns. A campaign that is 10 days old and a campaign that is 40 days old are not on equal terms, so never rank them on the same date range.
Are Untracked ASINs and Mixed Audiences Making DSP ROAS Look Low?
They can. A product left off the tracked list can only earn credit as brand halo, and only when it shares a subcategory with a tracked product, and one order that mixes prospecting and remarketing reports a blended number that hides both.
Blended ROAS = Σ (Spend share × ROAS of each audience)
Example (illustrative). An order spends 70% on remarketing at ROAS 8.0 and 30% on prospecting at ROAS 1.2. Blended ROAS is 0.7 × 8.0 + 0.3 × 1.2 = 5.96. Split the audiences into separate ad groups, and you can see that prospecting is carrying a ROAS of 1.2, which may be fine for awareness, while remarketing looks strong partly because it reaches people who were already likely to buy. Judge each against its own job, not one shared target.
What Is a Good ROAS for Amazon DSP?
A good DSP ROAS is one above your break even ROAS for the job the campaign does, and break even depends on your margin, not on an industry average. Amazon does not publish a benchmark that applies to every account, so work from your own numbers.
Break even ROAS = 1 ÷ Contribution margin before ad spend
Contribution margin | Break even ROAS | Read it as |
|---|---|---|
40% | 2.5 | Room to invest in prospecting |
30% | 3.3 | A common middle case |
20% | 5.0 | Tight, so favor remarketing and exclusions |
10% | 10.0 | Little room for any upper funnel spend |
Upper funnel campaigns can run below break even on last touch ROAS and still be worth it, if they raise new to brand customers or later sales. That is a decision to test, not assume. Our guide to Amazon DSP best practices covers how to set targets by objective.
Why Does DSP ROAS Fall When You Scale Spend?
Average ROAS falls when you scale because the extra budget reaches less valuable shoppers, and marginal ROAS, the return on the added dollars, falls faster than the average. Judge a scale up by marginal ROAS against break even, not by the headline number.
Marginal ROAS = Added sales ÷ Added spend
Example (illustrative). Spend rises from $10,000 to $20,000 and sales rise from $60,000 to $90,000. Average ROAS drops from 6.0 to 4.5, which still looks healthy. Marginal ROAS is 30,000 ÷ 10,000 = 3.0. At a 30% margin, break even is 3.3, so the added $10,000 returned $9,000 in contribution and lost $1,000.
Bar chart of illustrative average ROAS of 6.0 and 4.5 against marginal ROAS of 3.0 and a break even line of 3.3
Figure 3: Illustrative average and marginal ROAS when spend doubles.
What Happens to ROAS When You Loosen Settings to Force Spend?
ROAS usually falls when you loosen CPM, frequency or viewability to fix underspend, because you buy less selective impressions to get the delivery. This pattern is a judgment from how auctions work, not an Amazon statistic, so measure it on your own account.
Plan for the trade before you make the change. Set a break even ROAS for the line, loosen one setting, and read the result after the data has matured. If ROAS holds above break even, keep the change. If it does not, accept the underspend and move the budget to a line that can use it. Unspent budget is often cheaper than spend that earns below break even.
How Do You Fix Both Problems Without Wasting Budget?
Fix delivery first, measurement second and performance last, because each step changes what the next one shows. The table gives the first move for each symptom.
What you see | First move |
|---|---|
Pace check below 0.8 | Find the blocking setting from the pacing alert |
Frequency ceiling below daily budget | Raise the cap or widen the audience, or cut the budget |
ROAS fell in January 2026 | Compare standard purchases with Purchases (all views) |
ROAS weak on data under 14 days old | Wait until the window settles |
Blended ROAS hides weak ad groups | Split audiences into separate ad groups |
Average ROAS fine, profit flat | Check marginal ROAS against break even |
How Can Atom11 Help Diagnose DSP Underspend and Low ROAS?
Atom11 is an Amazon ads automation platform whose DSP module covers the operating jobs behind these checks: pacing across orders, frequency caps by audience, AMC audiences and DSP and Sponsored Ads reporting in one place. Confirm in a demo whether it surfaces Amazon’s own delivery rate, budget at risk and missed bid reasons for your accounts.
Diagnosis job | How the Atom11 Amazon DSP module helps |
|---|---|
Underspend across many orders | Cross order pacing toward one monthly budget, and bulk edits to orders and ad groups |
Frequency ceilings | Frequency caps set per audience |
Low ROAS reads | DSP and Sponsored Ads side by side, with promoted, halo and combined metrics |
Audience splits | AMC audience activation with drag and drop reports and no SQL |
Finding the cause of a change | Neo compares periods across more than 15 ad and retail signals |
Reporting | Exports to Power BI, Google Sheets and Looker |
Safe changes | Version control and a Performance Monitor |
Atom11 DSP pricing starts at $499 a month for accounts under $50K of ad spend and is 2% of ad spend above that, with Neo and the MCP included, and a managed service option is available. It is a weaker fit for teams that want a fully hands off autopilot. Check current terms on the pricing page, or book a demo and bring one underspending order to test.
Conclusion
A DSP that is not spending and a DSP with low ROAS are often the same account with two different problems. Underspend is a delivery problem, caused by a setting that cuts eligible supply or caps your bid, and the fix is to find the blocking setting and not to raise the budget. Low ROAS is more often a measurement problem, from the 2026 view attribution change, immature data, untracked ASINs or blended audiences, than a sign that the ads stopped working.
Work in order. Confirm delivery with Amazon’s delivery rate and your own pace check, settle the measurement by comparing standard and all views purchases on mature data, and only then judge performance with marginal ROAS against break even. Change one setting at a time, so you know which change paid off.
