PPC and DSP answer different questions, so compare them on different metrics. Judge Sponsored ads on ACoS, click-through rate, and incremental orders at the keyword level. Judge DSP on new-to-brand percentage, reach, and cost per detail page view. Comparing both on ROAS alone rewards whichever channel retargets your existing buyers hardest.
The short version
- PPC captures demand that already exists. Score it at the keyword level: ACoS, click-through rate, conversion rate.
- DSP creates and recovers demand. Score it on new-to-brand order share, reach, and cost per detail page view.
- Blended ROAS is the wrong scoreboard. Retargeting your own repeat buyers always looks efficient on paper.
- Sequence the channels. DSP amplifies a listing that already converts. It cannot rescue one that does not.
- This is also a staffing question. Two dashboards read properly every week takes real analyst hours.
Two channels doing different jobs
The reason PPC and DSP metrics cannot be compared head to head is that the channels intercept the shopper at different moments. Sponsored ads sit inside search results, in front of someone who typed a keyword with intent. DSP buys display and video placements on and off Amazon, in front of someone who has not searched yet, or who viewed your product and left without buying.
A metric only means something relative to the job. Punishing DSP for a higher cost per order is like punishing a billboard for not being a cash register. Praising DSP for a beautiful return built entirely on retargeting your own recent visitors is worse, because that budget mostly harvested orders you would have won anyway.
The metrics, side by side
| What you want to know | Ask PPC | Ask DSP |
|---|---|---|
| Is spend efficient? | ACoS by keyword, against a target you set | Cost per new-to-brand order |
| Is the creative working? | Click-through rate by placement | Detail page view rate on impressions |
| Is it growing the brand? | Organic rank movement on target terms | New-to-brand share of attributed orders |
| Is it incremental? | Branded vs non-branded keyword split | Frequency-capped or hold-out tests |
| Is the budget right? | Impression share on core keywords | Reach against the audience you defined |
Two rows matter more than the rest. The branded versus non-branded split on PPC tells you whether you are buying growth or buying your own name back. New-to-brand share on DSP tells you whether it is prospecting for strangers or re-billing you for customers you already earned.
The decision rule I use
- Prove listing conversion first. If the detail page does not convert search traffic, fix that before funding either channel. That work is listing optimization, not media buying.
- Fund PPC until the incremental keyword pool stops paying back. Search intent is the cheapest demand you will ever buy.
- Add DSP retargeting only once detail page traffic is large enough for the audience pool to matter.
- Add DSP prospecting last, and judge it on cost per new-to-brand order, nothing else.
- Re-check the reporting monthly. The moment DSP performance is quoted as one blended figure, ask for the split.
Whoever runs these channels needs the hours to read them. At Flapen we hold the load to about 1.4 brands per operator, and the honest reason is reporting like this: separating branded from non-branded and new-to-brand from retargeting takes analyst time every single week. An account manager carrying ten brands hands you one blended number, because that is all there is time to produce. Ask any agency you are evaluating how many accounts each manager carries before you ask about their DSP credentials.
What ad agencies will not tell you
DSP has a reporting surface that flatters it. Retargeting audiences are small and warm, so early campaigns post strong returns, and the deck you receive shows those returns folded into one figure. The question that cuts through: what percentage of DSP-attributed orders were new-to-brand, and what did each of those orders cost. If that answer does not arrive quickly, the campaign is harvesting, not hunting.
The second omission is sequencing. An agency earning fees on media management has little incentive to say your listing is not ready for more traffic. It is still the most common state I find accounts in when we take them over.
Related answers
- Amazon advertising KPIs that matter
- Amazon retail media attribution models explained
- Full-funnel Amazon audit vs PPC-only audit
- Best audit for Amazon PPC structure and neg match
- Amazon creative services: the complete guide
If you want PPC and DSP reported with the splits this page describes, that is standard practice at Flapen.
