How is traffic allocated by the exchanges to different DSPs? (8)
How is traffic allocated by the exchanges to different DSPs?
As I understand it, it is a common aspect of media buying that once you have a profitable campaign, you should scale to other traffic sources, including other DSPs. However, most of these DSPs connect to the same ad exchanges.
Do all the DSPs have access to all of the inventory of each exchange, and that traffic is split in real time by bids to the different DSPs or are placements or certain amounts of traffic allocated to specific DSPs such that you can scale the same campaign to another DSP and select the same exchange but not compete against yourself?
For example, the ad exchange Smaato connects to over 340 DSPs. Is it necessary to sign up for all of them in order to have access to all of Smaato’s inventory?
If not then is scaling to other traffic sources only meaningful if the new DSP connects to exchanges different than the ones you had access to before?
Yea I'd like to know some of this information too.
My feeling is that each DSP and exchange probably has individual arrangements eg % and quality of traffic.
I could be wrong though.
As I understand it (could be wrong), an auction takes place across all the DSPs that have access to a specific inventory, with the winning bidder getting the impression. Impressions aren't pre allocated - that's where the real-time bidding comes in.
Different DSPs have access to different exchanges though (and maybe different parts of the exchange's inventory?).
So placements are already separated among the DSPs? No one DSP has access to all placements in an exchange's network?
From my tests on 3 different DSPs all connected to smaato, I believe that smaato can allocate different traffic to each DSP, as a percentage.
On DSP 1 Im receiving quite a lot of traffic from smaato at a relatively low bid. On DSP 2 I was receiving 5 times the traffic for half the price of DSP 1, but after about a week, either smaato didnt like that or DSP 2 is messing with me and I have to bid almost 7 times that to receive any traffic at all. On DSP 3 im getting traffic at around double the bid of DSP 1. Also, the traffic quality is best for me on DSP 1 with the cheapest current bid.
Originally Posted by lanza123
Yeah, so it seems that traffic is distributed amoung DSP's and not given to the highest bidder that is connected to the exchange. This is opposite of what you should expect on a marketplace...Nevertheless, usually, this is not that bad, because you don't want to bombard all traffic available with your ads. However, some apps performing so amazingly well, that you wish you could get more, and in that case it is bad. In addition, if the prices differ amount the DSP's, it means you should find the cheapest DSP. So according to your test results, you should start working with several DSP's.
Question: I think it is more important to focus on getting something to work and then move away from DSP/RTB, to go or example to inmobi, leadbolt or admob. Why would you stay working on DSP/RTB?
Cheers
Normally an exchange just sends out bid requests across all dsps but there are filters/blacklists/floor prices for advertiser id, dsp etc that the ssp can enforce. You also have ad classification (initially done by the dsp). Quality probably mostly differs by the filter tech used by the dsps. Also its possible that the dsp only actively responds to a certain % of bid requests. Pricing differences can occur due to the fee DSP's charge (which is included in your bid for most dsps)
All in all, many variables which results in different performance across DSP's. The DSP should be able to tell you exactly what variables are in effect for your campaign (like: ssp not accepting bids from dsp, ssp not accepting bids from advertiser, ssp not accepting classification)
In the end, the property owner decides what ads run on their property
This is a very interesting subject, and a great question!
We're now building a DSP into
There are a lot of factors which affect the bidding, but in theory it can be said that it's beneficial to run the same campaign via several DSPs who are integrated with say a single exchange like Smaato (obviously DSPs are connected with different exchanges, but to address OPs question, we'll assume there's one exchange in this scenario).
-DSPs have different margins, so the actual bid sent may be significantly lower on DSP 1 vs. DSP 2 if your bid is say $10 CPM on both and one has a margin of 30% and the other 10%.
-DSPs often throttle the requests received from exchanges, the volumes that are dealt with are so extreme that not all DSPs that can handle 100% in real-time.
-DSPs categorize your campaigns, each does so manually, differences here would affect the traffic you receive.
-A DSP may be slow and not fit in the required time to respond with a bid, missing out on the auction.
-Exchanges may not always take the highest bid as the winning bid on a particular placement in a single request - for example they may want to give a 'taste' of the traffic to lower bids to encourage them to bid higher.
-DSPs support different ad sizes and ad types, so only a part of an exchanges inventory may be available on one DSP.
-Publisher (owner of app/site on exchange) may block a campaign, category, advertiser, or DSP.
Reality:
-DSPs are integrated with different exchanges, there are more exchanges out there than you may think.
-The contracts DSPs sign with exchanges are usually straight forward but they may vary from DSP to DSP.
-DSPs offer different levels of optimization
-Exchanges may do proxy bidding, meaning the highest bid from a DSP may not be the price they charge the DSP. In such a case you could be competing against yourself.
There are probably a slew of other factors, but as we've not done too many live tests I don't want to speculate too much.
To conclude, I would recommend scaling across different DSPs keeping in mind the following:
-Find out what exchanges they are connected too, some also have access to direct publisher
-Keep in mind the time you will need to dedicate to optimizing more campaigns on additional DSPs
-Check how your campaigns perform on a granular level and see if a given DSP offers the same level of optimization
I'd personally test as many of them as possible and select a few to stick with based on your results, the ease of campaign creation and optimization, campaign/creative approval, inventory research capabilities/tools, general reliability, and overall UI/UX, some are surprisingly terrible..
Best,
Robert
Home > Paid Traffic Sources > Mobile