Most DSP reports lead with ROAS, which is the one number that makes DSP look worst and tells you least. New to brand share is what says whether the channel is doing its job. Here is what to ask for instead, with two real accounts read properly.
Amazon DSP
Ask most agencies how your DSP is performing and you will get a ROAS figure. It is the wrong lead, and it makes a well run DSP look like a poorly run search campaign. New to brand share is the number that says whether the channel is doing its actual job, which is reaching people who do not know you yet.
DSP’s job is to find buyers who have never heard of you. Those people take longer to buy, buy less on the first visit, and a good share of them convert later through search, where the credit lands somewhere else entirely. Compress all of that into one return figure and you have thrown away the thing you were paying for.
Judging DSP on ROAS is like judging a shop’s front window on how many people bought something while standing in it.
What is on this page
What new to brand actually measures
A new to brand purchase is one made by a shopper who had not bought anything from your brand on Amazon in the previous twelve months. Amazon works this out from purchase history you cannot see, which is precisely why it is valuable: nobody else can calculate it for you.
It is reported for Sponsored Brands, Sponsored TV and DSP rather than Sponsored Products. That distinction matters more than it looks, because it tells you what Amazon thinks each channel is for. Sponsored Products captures demand that already exists. The others are supposed to create it, so they get the metric that measures creation.
So the question new to brand answers is simple and important. Of the sales this channel produced, how many were customers you did not have?
Why ROAS fails here
Three reasons, and they compound.
Cold audiences buy slower. Somebody meeting your brand for the first time on Fire TV does not buy that evening. They buy in a fortnight, after seeing you again and reading some reviews. Any return figure measured over a short window misses most of them.
The credit lands elsewhere. A shopper introduced by DSP frequently completes the purchase through search, because that is how people navigate Amazon. Your Sponsored Products report gets a cheap sale and your DSP report gets spend with nothing next to it. Judge each channel on its own report and you will conclude that search is brilliant and DSP is broken, when what actually happened is that one handed to the other.
First orders are smaller. New customers try the small size, buy one rather than the bundle, and come back for more later. The first transaction understates what you bought.
None of that means return does not matter. It means a single return figure, read early, systematically undervalues the one channel you are running specifically to reach strangers.
What to ask for instead
Four things belong at the top of any DSP report. If your agency leads with ROAS and buries these, ask why.
New to brand sales, as a share of total. Buyers who had not purchased from you in the previous twelve months. This is the number that says whether the channel is doing its actual job.
New to brand cost. What it cost to acquire one. Compare it against your customer lifetime value, not against your ACOS target. Those are different questions and only one of them is relevant here.
Detail page views and branded search volume. Both should rise while DSP runs. If neither moves, the creative or the audience is wrong, and you will know that long before the sales data is conclusive.
Search performance on the same terms. A working DSP makes Sponsored Products convert better. Some of the return you paid for shows up in that line, and if nobody is looking at it, nobody is counting it.
Agree this before the campaign starts. Every argument about DSP performance is really an argument about which numbers count. Settling that in week zero takes ten minutes. Settling it in week nine, with money spent and opinions formed, takes considerably longer and usually ends with the campaign switched off.
Getting a DSP report you cannot really interrogate?
Send it over. We will tell you what it is actually saying, free, whether or not you work with us.
Two accounts, read properly
Numbers from our own book, not illustrations.
The first is seventeen months of continuous running: $101,970 of media, $320,437 of combined product sales, a 3.14 combined ROAS. On its own that return looks unremarkable, and plenty of brands would have cancelled at month four.
Then look at the fourth figure. $233,105 of those sales, 72.7% of everything the channel produced, came from people who had never bought the brand. That is not a modest return on advertising. That is a customer acquisition programme that happened to pay for itself three times over while it ran, and the customers it acquired are still buying.
The second account is a single month on a tuned setup, and it shows what happens once the audiences have learned: $4,856 of media returning $35,310 at a combined ROAS of 7.27, with 81.5% new to brand. Same channel, same principles, further along the learning curve. Nobody achieves the second number in month one, which is exactly why the first account matters more as an example.
What a low share tells you
If new to brand is under about 40% of DSP sales, the campaign is mostly retargeting people who were going to buy anyway. That is not useless, it is just not what DSP is for, and it is a far more expensive way to buy those sales than search would have been.
The usual causes are predictable. Budget has drifted towards retargeting audiences because they perform best on the short term report, which is a self reinforcing loop. Prospecting audiences were set too narrow, so the campaign keeps finding the same familiar people. Or frequency caps are loose, and the same small pool is being hit repeatedly rather than the pool being widened.
The fix is a deliberate split between prospecting and retargeting, decided in advance and protected from the temptation to move money towards whichever line looks best this fortnight. Retargeting will always look better in a short window. That is not evidence that it deserves more budget.
The number that decides whether it was worth it
New to brand cost only means something next to customer lifetime value, and this is where most sellers get stuck, because they have never worked theirs out.
A rough version is enough to make decisions. Take your repeat purchase rate over twelve months, multiply by average order value, multiply by contribution margin. If a customer buys 2.4 times a year at £42 with 30% contribution, they are worth about £30 to you annually. Paying £18 to acquire one through DSP is a good trade. Paying £45 is not, however healthy the new to brand share looks.
For consumables and subscription friendly products this arithmetic usually justifies far more aggressive acquisition spending than an ACOS target ever would. For a one time purchase with no repeat, it justifies much less. Same channel, same metric, opposite conclusions, and only your own numbers can tell you which one you are.
ROAS asks what this month’s spend returned. New to brand asks how many customers you now have that you did not have before. Only one of those is a business.
Want a DSP report that leads with the right number?
This is how we report every DSP account we run, and you can have the template either way.
A monthly report worth reading
Keep it to one page. New to brand sales and share. New to brand cost against your lifetime value figure. Detail page views and branded search over the period. Sponsored Products conversion rate on the terms DSP is supporting. Then, last rather than first, the return figures.
Read it as a trend rather than a snapshot. Share climbing month on month means the prospecting is working. Cost per new customer falling means the audiences are learning. Branded search rising means the creative is landing. Any of those moving in the wrong direction is worth investigating well before the return figure confirms it, which is the whole advantage of reporting this way.
If you are still deciding whether the channel belongs in your account at all, the wider case is in Amazon DSP explained, including how to tell whether you have genuinely run out of room in search first.
How to push the share higher
Protect the prospecting budget. Decide the split in advance and hold it for a full quarter, because retargeting will always win the weekly comparison.
Widen the audiences. In market and lifestyle audiences reach further than product viewers. They convert worse per impression and they are the only place genuinely new customers come from.
Cap frequency. The twentieth impression to the same person does not buy anything the fourth one did not.
Refresh creative. Fatigue shows up as falling click through and rising cost per acquisition, usually around week six to eight.
Make sure the landing page can close a stranger. Cold traffic is unforgiving, and a listing that just about works for warm search traffic often falls apart with it. The checks in your listing is a mobile listing now are the right starting point.
Want DSP reported in a way you can actually act on?
Twenty minutes on a call and we will show you what your numbers are really saying.
Frequently asked questions
A new to brand purchase is one made by a customer who had not bought anything from your brand on Amazon in the previous twelve months. Amazon reports it across Sponsored Brands, Sponsored TV and DSP, and it is the closest thing you have to a measure of genuine customer acquisition rather than repeat business.
For DSP, above roughly 60% is healthy and 70% to 80% is what we see on well run accounts. Under about 40% usually means the campaign has drifted into retargeting people who were going to buy anyway, which is not useless but is an expensive way to buy customers you already had.
Because DSP reaches people who do not know you. They take longer to buy, buy less on the first visit, and many convert later through search where the credit lands elsewhere. Compressing all of that into one return figure throws away the thing you were paying for.
Divide total media spend by the number of new to brand orders. Then compare that figure to what a customer is worth to you over a year rather than to your ACOS target, because you are buying customers here, not individual transactions.
A first Subscribe and Save order from someone new to the brand counts as new to brand. Their subsequent repeat deliveries do not, which is exactly right: the acquisition happened once, and everything after that is the value you acquired.
Not in the same way. New to brand reporting is available for Sponsored Brands, Sponsored TV and DSP rather than Sponsored Products. That is one reason judging those channels against Sponsored Products metrics goes wrong so often: they are reporting different things on purpose.
Yes, and if it does not, something is wrong with the creative or the audience. Rising branded search is one of the clearest signs that awareness advertising is landing, and it usually appears before the sales do.
Eight to twelve weeks at minimum. The audiences need volume to learn, and the second and third purchases from customers you acquired in month one only start showing up in month three, which is precisely where the economics turn.
Send the right reports. Get a straight answer back.
We read the account properly and write back with where the money is going, what it is buying, and the three things we would change first. You keep it either way.
No obligation · no card · a person replies, usually the same working day