Products
Resources
Other
How To Choose And Not Regret. 20 Questions To DSP Before Launch
September 17, 2026
There are enough DSP platforms in the advertising market today, each of which is able to purchase banner and video ads, work with audience segments, optimize campaigns and show statistics. If you open several presentations side by side, the differences begin to fade altogether. "Large reach", "high-quality inventory", "accurate targeting", "smart algorithms", and so on.
However, the advertiser is not buying the presentation. He buys technology, access to inventory and data, and with them a certain level of control over the advertising budget. Therefore, the choice of a DSP begins where the list of possibilities ends and uncomfortable questions begin.
We have collected 20 questions that you should ask a technology supplier before launching a campaign and will help you make the best choice.
However, the advertiser is not buying the presentation. He buys technology, access to inventory and data, and with them a certain level of control over the advertising budget. Therefore, the choice of a DSP begins where the list of possibilities ends and uncomfortable questions begin.
We have collected 20 questions that you should ask a technology supplier before launching a campaign and will help you make the best choice.
Under The Hood
DSP (Demand-Side Platform) is a system through which an advertiser automatically purchases advertising inventory. When a user opens a website or application, the platform informs them about the opportunity to show them ads. Buyers participate in the auction, and the winner receives a display.
Everything happens in a split second.
1. Do you have your own technology?
"The question"seems obvious, but the answer may be more complicated than a simple "yes" or "no."
The market has its own DSPs, licensed solutions, SaaS models, and White Label platforms. A company can sell its own service and interface to an advertiser, while a significant part of the technology stack belongs to another provider, like UMG. There is nothing criminal in this. When buying a car, we do not expect that the manufacturer independently melts steel for each bolt.
The question is who controls the critical components. Who develops the bidding engine? Who is responsible for the server infrastructure? Who is updating the platform? Who connects the new SSPs? Who can change the algorithm?
The more intermediaries there are already at the level of the technology itself, the more important it is to understand the boundaries of responsibility.
2. Where does the DSP live?
The DSP is constantly under high load: it receives requests for impressions, makes decisions about participating in the auction and calculates bids. Therefore, it is useful to find out where the infrastructure is physically located, how it is backed up, and what happens if a separate data center or system component fails.
" A separate advantage will be the presence of its own page with a fault detector or responsive support, which will promptly inform about the "fall" of the product.
3. What is your SLA?
"Reliable", "stable" and "fault-tolerant" are better words to forget here. The numbers are responsible for this. What uptime does the platform guarantee? How long will it take the technical team to respond to an emergency? Is there round-the-clock monitoring?
The auction does not close in the evening along with the office. The campaign continues to operate at night, on weekends and holidays. So a good additional question sounds like this: Who's going to answer the phone at three o'clock on a Saturday morning?
4. What if the platform spends too much?
An unpleasant but practical scenario. Let's say the budget limit has stopped working due to a technical error. The campaign should have stopped, but continued the purchase.
Who will detect the anomaly? After how long? Who can stop the campaign? At whose expense will the overspending be?
Mistakes are inevitable in technological systems. What is more important is not the promise of their complete absence, but the ability to quickly detect them and the agreed-upon responsibility of the parties.
Everything happens in a split second.
1. Do you have your own technology?
"The question"seems obvious, but the answer may be more complicated than a simple "yes" or "no."
The market has its own DSPs, licensed solutions, SaaS models, and White Label platforms. A company can sell its own service and interface to an advertiser, while a significant part of the technology stack belongs to another provider, like UMG. There is nothing criminal in this. When buying a car, we do not expect that the manufacturer independently melts steel for each bolt.
The question is who controls the critical components. Who develops the bidding engine? Who is responsible for the server infrastructure? Who is updating the platform? Who connects the new SSPs? Who can change the algorithm?
The more intermediaries there are already at the level of the technology itself, the more important it is to understand the boundaries of responsibility.
2. Where does the DSP live?
The DSP is constantly under high load: it receives requests for impressions, makes decisions about participating in the auction and calculates bids. Therefore, it is useful to find out where the infrastructure is physically located, how it is backed up, and what happens if a separate data center or system component fails.
" A separate advantage will be the presence of its own page with a fault detector or responsive support, which will promptly inform about the "fall" of the product.
3. What is your SLA?
"Reliable", "stable" and "fault-tolerant" are better words to forget here. The numbers are responsible for this. What uptime does the platform guarantee? How long will it take the technical team to respond to an emergency? Is there round-the-clock monitoring?
The auction does not close in the evening along with the office. The campaign continues to operate at night, on weekends and holidays. So a good additional question sounds like this: Who's going to answer the phone at three o'clock on a Saturday morning?
4. What if the platform spends too much?
An unpleasant but practical scenario. Let's say the budget limit has stopped working due to a technical error. The campaign should have stopped, but continued the purchase.
Who will detect the anomaly? After how long? Who can stop the campaign? At whose expense will the overspending be?
Mistakes are inevitable in technological systems. What is more important is not the promise of their complete absence, but the ability to quickly detect them and the agreed-upon responsibility of the parties.
"A Lot" Of Inventory
This is where presentations usually begin to compete with the number of sites, apps, and available screenings. But the absolute number of sources says much less than it seems.
5. Where do you get your inventory from?
It is better to ask the supplier to sort out the answer on the shelves:
Which SSPs are connected directly? Are there any direct integrations with major sites? How much do resellers make up? Is there a PMP (Private Marketplace) and exclusive sources?
Two DSPs may have the same number of available impressions and a completely different offer structure, where the ad will end up.
6. How many handshakes are there before the publisher?
In theory, the programming chain looks simple:
advertiser – DSP – SSP – publisher
But in practice, it may look like this:
advertiser – DSP – advertising exchange – reseller – SSP – publisher
And this is not the longest option yet.
Each additional participant performs a specific function and can be fully justified. The problem begins when the advertiser does not understand who is inside the chain at all.
First, each link has a potential impact on the economics of procurement.
Secondly, a long chain complicates the analysis of the impression source.
Thirdly, the more intermediaries there are, the more important transparency mechanisms are.
The industry has been solving this problem for years. If you want to understand in more detail how the relationship between DSP, SSP and other auction participants has changed, we talked about it in the article about the evolution of oRTB.
7. Can you show me the playgrounds? Preferably before launch.
Ask for an example of inventory or a forecast for a specific brief. Which sites are included in the purchase? What kind of applications? Which video platforms?
The "premium inventory" formulation is great because everyone considers premium in their own way. Sometimes ten really relevant sites are more useful to an advertiser than several hundred "premium" ones.
8. Will you show it after the launch?
The preliminary list is half the story. We also need to check where the ads were actually displayed. Is it possible to get statistics for a domain or an application? Is the purchase source visible? Is it possible to see the expense, CPM and performance of individual sites?
In this case, the automated reporting and forecasting tools of the platform are important. If an advertiser sees the sources, they can detect anomalies, exclude weak sites, create their own stop lists and understand whether the actual purchase corresponds to the original plan. You can find out more about the reports in uDSP in our earlier article.
5. Where do you get your inventory from?
It is better to ask the supplier to sort out the answer on the shelves:
- Web
- Mobile Web
- In-app advertising
- Online video
- CTV
Which SSPs are connected directly? Are there any direct integrations with major sites? How much do resellers make up? Is there a PMP (Private Marketplace) and exclusive sources?
Two DSPs may have the same number of available impressions and a completely different offer structure, where the ad will end up.
6. How many handshakes are there before the publisher?
In theory, the programming chain looks simple:
advertiser – DSP – SSP – publisher
But in practice, it may look like this:
advertiser – DSP – advertising exchange – reseller – SSP – publisher
And this is not the longest option yet.
Each additional participant performs a specific function and can be fully justified. The problem begins when the advertiser does not understand who is inside the chain at all.
First, each link has a potential impact on the economics of procurement.
Secondly, a long chain complicates the analysis of the impression source.
Thirdly, the more intermediaries there are, the more important transparency mechanisms are.
The industry has been solving this problem for years. If you want to understand in more detail how the relationship between DSP, SSP and other auction participants has changed, we talked about it in the article about the evolution of oRTB.
7. Can you show me the playgrounds? Preferably before launch.
Ask for an example of inventory or a forecast for a specific brief. Which sites are included in the purchase? What kind of applications? Which video platforms?
The "premium inventory" formulation is great because everyone considers premium in their own way. Sometimes ten really relevant sites are more useful to an advertiser than several hundred "premium" ones.
8. Will you show it after the launch?
The preliminary list is half the story. We also need to check where the ads were actually displayed. Is it possible to get statistics for a domain or an application? Is the purchase source visible? Is it possible to see the expense, CPM and performance of individual sites?
In this case, the automated reporting and forecasting tools of the platform are important. If an advertiser sees the sources, they can detect anomalies, exclude weak sites, create their own stop lists and understand whether the actual purchase corresponds to the original plan. You can find out more about the reports in uDSP in our earlier article.
Transparency
Programmatic is often compared in terms of price. Whoever has a thousand impressions cheaper seems to have won.
9. How much of a million rubles will be spent on the advertising itself?
"Try asking the question exactly like this:
"When transferring a million rubles, what part of this amount will be directly involved in the purchase of inventory?"
In addition to the inventory cost, there may be a DSP commission, data cost, verification, and other technological services inside.
This does not mean that the commission is something bad. A DSP, like any commercial structure, must earn money in order to maintain servers, keep algorithms running, and feed the families of the people who take on this job. But it is useful for the advertiser to understand exactly what he is buying.
10. Is the commission the same everywhere?
Not obligatory.
The economics of banner advertising, video, CTV, private transactions, and purchasing using paid data may vary.
Therefore, when comparing two media plans, one final CPM is not enough. One supplier may be more expensive to buy high-quality inventory and charge a lower technological commission. The other is the opposite.
On paper, the figure will be the same, but there will be two different purchases behind it.
11. What will we pay extra for?
Audience segments can be paid for. Verification, non-standard analytics, research, or individual integrations are similar. It is better to get a list of additional expenses before launching. Especially if the DSPs are compared as part of a tender. A media plan without some of the technological costs may look cheaper before receiving the first invoice.
9. How much of a million rubles will be spent on the advertising itself?
"Try asking the question exactly like this:
"When transferring a million rubles, what part of this amount will be directly involved in the purchase of inventory?"
In addition to the inventory cost, there may be a DSP commission, data cost, verification, and other technological services inside.
This does not mean that the commission is something bad. A DSP, like any commercial structure, must earn money in order to maintain servers, keep algorithms running, and feed the families of the people who take on this job. But it is useful for the advertiser to understand exactly what he is buying.
10. Is the commission the same everywhere?
Not obligatory.
The economics of banner advertising, video, CTV, private transactions, and purchasing using paid data may vary.
Therefore, when comparing two media plans, one final CPM is not enough. One supplier may be more expensive to buy high-quality inventory and charge a lower technological commission. The other is the opposite.
On paper, the figure will be the same, but there will be two different purchases behind it.
11. What will we pay extra for?
Audience segments can be paid for. Verification, non-standard analytics, research, or individual integrations are similar. It is better to get a list of additional expenses before launching. Especially if the DSPs are compared as part of a tender. A media plan without some of the technological costs may look cheaper before receiving the first invoice.
Who Is An Audience?
Targeting is one of the main reasons to use DSP at all.
Modern platforms allow you to select an audience based on geography, interests, behavior, devices, and many other attributes. But there is a small problem here, because a segment with a beautiful name does not mean high-quality data.
12. How do you know anything about the audience?
DSP can work with its own DMP and external data sources: applications, publishers, and other vendors. UMG, for example, has its own DMP with data from several sources.
So "how many segments do you have?" is the question.
"Where did these segments come from?" should be replaced with another one.
The data source, the audience formation methodology, and the ability to combine different types of signals say much more about targeting than the number of lines in the segment catalog.
13. How is a particular interest assigned to the user?
Let's say the platform has a segment called "interested in buying a car." But one vendor included users of car websites there. The second is for owners of certain apps, and the third is for people with appropriate purchasing behavior. The name is the same, but the audience is different.
A good DSP will explain the methodology of segment formation, which signals are used and why they allow us to conclude about the user's interest.
14. And when was he interested?
Yesterday and nine months ago – different intentions. For some characteristics, the duration of the signal does not play a big role. For others, recency is critical. A person may be interested in traveling for a long time, but searching for a specific flight ticket has a much shorter window of relevance.
Therefore, it is worth finding out the lifetime of the segments and the refresh rate. Especially if the campaign is not focused on interest, but on expressed purchasing intent.
15. Can I bring my details?
Many advertisers already have their most valuable data inside the company: CRM, website visitors, app users, customers, and loyalty program participants. Therefore, it's worth finding out if DSP can work with first-party data.
And there's another important question: what happens to this data after the campaign ends? It is better to discuss the rules for storing, using, and deleting advertiser data before transferring it to the platform, and some jurisdictions generally require a more rigorous approach to data processing.
Modern platforms allow you to select an audience based on geography, interests, behavior, devices, and many other attributes. But there is a small problem here, because a segment with a beautiful name does not mean high-quality data.
12. How do you know anything about the audience?
DSP can work with its own DMP and external data sources: applications, publishers, and other vendors. UMG, for example, has its own DMP with data from several sources.
So "how many segments do you have?" is the question.
"Where did these segments come from?" should be replaced with another one.
The data source, the audience formation methodology, and the ability to combine different types of signals say much more about targeting than the number of lines in the segment catalog.
13. How is a particular interest assigned to the user?
Let's say the platform has a segment called "interested in buying a car." But one vendor included users of car websites there. The second is for owners of certain apps, and the third is for people with appropriate purchasing behavior. The name is the same, but the audience is different.
A good DSP will explain the methodology of segment formation, which signals are used and why they allow us to conclude about the user's interest.
14. And when was he interested?
Yesterday and nine months ago – different intentions. For some characteristics, the duration of the signal does not play a big role. For others, recency is critical. A person may be interested in traveling for a long time, but searching for a specific flight ticket has a much shorter window of relevance.
Therefore, it is worth finding out the lifetime of the segments and the refresh rate. Especially if the campaign is not focused on interest, but on expressed purchasing intent.
15. Can I bring my details?
Many advertisers already have their most valuable data inside the company: CRM, website visitors, app users, customers, and loyalty program participants. Therefore, it's worth finding out if DSP can work with first-party data.
And there's another important question: what happens to this data after the campaign ends? It is better to discuss the rules for storing, using, and deleting advertiser data before transferring it to the platform, and some jurisdictions generally require a more rigorous approach to data processing.
Traffic Verification
You can choose the ideal audience and a high-quality platform, and then discover that some of the advertising contacts occurred only in the report.
16. How does your anti-fraud work?
"We have our own anti-fraud system" is the beginning of the answer, but not the answer itself.
What does it define? What types of invalid traffic are being filtered? Does the verification take place before the purchase of the display or after? Are external systems used? What does DSP do if a specific source starts generating suspicious activity? Automatically reduces the purchase? Excludes him? Is he sending it for additional verification?
Anti-fraud is useful not by the very fact of its existence, but by the decisions that the platform makes based on its data.
The availability of additional traffic filtering systems will, of course, be a plus. For example, UMG, in addition to internal analytical tools, uses the internationally recognized Protected Media anti-fraud system from Mediaocean to guarantee customers the quality of their placements.
17. Where will the Internet be next to my advertisement?
Brand Safety (the safety of the Internet environment for a brand) is usually remembered after some particularly unfortunate screenshot from an advertiser. It's better to remember this earlier. Which categories of content can be excluded? Is it possible to block individual domains and applications?
At the same time, there is no absolute universal "safe environment". For a manufacturer of household appliances, a news website can be a completely normal placement. For another brand, the proximity of certain types of news will be undesirable.
Ideally, you should have not only Brand Safety, but also compliance of a specific environment with the rules of a specific advertiser.
18. And who will check the inspector?
DSP informs you that the show has taken place. DSP reports that the display was of high quality. It is logical to ask if this can be independently verified. Which verification systems does the platform support? Is it possible to measure IVT, viewability, and other quality parameters separately?
This issue becomes especially important in video ads and on Smart TV, where the usual browser-based measurement methods do not always work. There, the industry uses separate technological standards like VAST and Open Measurement, as we wrote earlier.
By themselves, additional meters do not guarantee a perfect result. But the possibility of independent verification is much more useful than the principle of "trust our internal statistics."
16. How does your anti-fraud work?
"We have our own anti-fraud system" is the beginning of the answer, but not the answer itself.
What does it define? What types of invalid traffic are being filtered? Does the verification take place before the purchase of the display or after? Are external systems used? What does DSP do if a specific source starts generating suspicious activity? Automatically reduces the purchase? Excludes him? Is he sending it for additional verification?
Anti-fraud is useful not by the very fact of its existence, but by the decisions that the platform makes based on its data.
The availability of additional traffic filtering systems will, of course, be a plus. For example, UMG, in addition to internal analytical tools, uses the internationally recognized Protected Media anti-fraud system from Mediaocean to guarantee customers the quality of their placements.
17. Where will the Internet be next to my advertisement?
Brand Safety (the safety of the Internet environment for a brand) is usually remembered after some particularly unfortunate screenshot from an advertiser. It's better to remember this earlier. Which categories of content can be excluded? Is it possible to block individual domains and applications?
At the same time, there is no absolute universal "safe environment". For a manufacturer of household appliances, a news website can be a completely normal placement. For another brand, the proximity of certain types of news will be undesirable.
Ideally, you should have not only Brand Safety, but also compliance of a specific environment with the rules of a specific advertiser.
18. And who will check the inspector?
DSP informs you that the show has taken place. DSP reports that the display was of high quality. It is logical to ask if this can be independently verified. Which verification systems does the platform support? Is it possible to measure IVT, viewability, and other quality parameters separately?
This issue becomes especially important in video ads and on Smart TV, where the usual browser-based measurement methods do not always work. There, the industry uses separate technological standards like VAST and Open Measurement, as we wrote earlier.
By themselves, additional meters do not guarantee a perfect result. But the possibility of independent verification is much more useful than the principle of "trust our internal statistics."
Five Screens, One Viewer
The user stopped living inside the same browser a long time ago.
He reads the news from his smartphone in the morning, works from his laptop in the afternoon, and turns on the TV in the evening. For an advertiser, it can be one person with multiple IDs.
19. What do you consider the frequency?
Setting the parameter "no more than three impressions" is not difficult. But who gets three impressions?
One cookie? The device? The user? The household?
If the campaign is running on the web, apps, and CTV at the same time, the task becomes much more difficult. You can accidentally show an advertisement to one person three times on a laptop, three times on a smartphone, and several more times on a TV, and each individual system will honestly report that the frequency requirements have been met.
Therefore, when purchasing omnichannel, it is worth asking not only about the frequency limit, but also about which identifiers it is based on and how well the DSP is able to distinguish the audience between the media.
He reads the news from his smartphone in the morning, works from his laptop in the afternoon, and turns on the TV in the evening. For an advertiser, it can be one person with multiple IDs.
19. What do you consider the frequency?
Setting the parameter "no more than three impressions" is not difficult. But who gets three impressions?
One cookie? The device? The user? The household?
If the campaign is running on the web, apps, and CTV at the same time, the task becomes much more difficult. You can accidentally show an advertisement to one person three times on a laptop, three times on a smartphone, and several more times on a TV, and each individual system will honestly report that the frequency requirements have been met.
Therefore, when purchasing omnichannel, it is worth asking not only about the frequency limit, but also about which identifiers it is based on and how well the DSP is able to distinguish the audience between the media.
The Key Question
20. Which of the above can I check for myself?
At this point, you can ask to open the interface.
Show yesterday's campaign. Where was the advertisement placed? How much money was spent on each site? How has CPM changed? Which segment of the audience worked better? How many times have users seen the ad? What campaign settings have changed?
The more responses a client can receive directly from the system, the less they depend on tables that someone has manually compiled after the campaign.
At this point, you can ask to open the interface.
Show yesterday's campaign. Where was the advertisement placed? How much money was spent on each site? How has CPM changed? Which segment of the audience worked better? How many times have users seen the ad? What campaign settings have changed?
The more responses a client can receive directly from the system, the less they depend on tables that someone has manually compiled after the campaign.
Trust And Verify
Ultimately, a DSP is not a cabinet with a "Start" button. Good technology doesn't have to be simple.
It has a long chain of technologies, data, and market participants behind it. It is impossible to check absolutely everything before the first campaign.
In this sense, good questions sometimes work more effectively than any presentation.
It has a long chain of technologies, data, and market participants behind it. It is impossible to check absolutely everything before the first campaign.
In this sense, good questions sometimes work more effectively than any presentation.