Why One DSP Just Isn’t Enough

veerle-exchange-lab1“Data Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Veerle De Lombaerde, global product director at The Exchange Lab.

Consider the following scenario: You spent months educating your client about the value of programmatic buying. Based on your client’s internal CRM data, you defined the right audience — and you also came up with a test-and-learn strategy that allows you to expand beyond the known best-performing audience. When you launch the programmatic-buying campaign, it’s a big success, according to every key performance indicator. The only problem now? You can’t deliver scale.

Most likely, you’ve chosen to use a single Demand-Side Platform to reach that audience. It’s a confusing market for those of us who work in the trenches of programmatic buying, let alone for our clients, but — even though they may not know it — the decision about whether to go with a single- or multi-platform approach actually does matter to clients. Sticking with a single platform can significantly limit a programmatic-buying campaign’s ultimate impact and success.

Here are four big reasons why you should work across multiple platforms, instead of just one:

1) Technology:

Every DSP submits its bids differently, which means that all bids are not created equal. Because of this, a $2 bid in Appnexus will not necessarily beat a $1.50 bid from The Trade Desk for the same inventory. A bid must win internal auctions within its DSP before it even makes it to the SSP auction, when the SSP decides which of those bids, from all of the DSPs, will win the impression. DSPs consider budget, performance against target, time of day, advertiser category, publisher block lists and minimum bid, to name just a few of many, many factors — beyond price — that can alter the likelihood of an advertiser winning a bid.

2) Scale:

As I mentioned above, choosing only one platform can limit how big a campaign can grow. This is because good optimization strategies often decrease the size of the audience available on a platform, for obvious reasons — a more specific audience is, necessarily, a smaller one. Running a campaign across multiple platforms increases the potential audience, bringing scale back into the mix. This is even more important when you’re targeting a niche audience, for which delivering scale is a bigger challenge. Building that audience in more platforms can only bring you more opportunities to reach it.

3) Risk Management:

As the publishers work to optimize yield by altering their back-end inventory allocations to SSPs and exchanges, it’s not uncommon for an excellent, optimized strategy to fail when the inventory mix on which you’re bidding suddenly changes — or even disappears all together. When this happens, the ability to shift to alternative platforms can provide options and redundancy, saving a campaign from getting sent back to the drawing board. In addition, running a campaign across multiple platforms simultaneously can maximize your chances for a win, similar to spread betting.

4) Insight:

In a constantly changing ecosystem, one of the most important aspects of developing your programmatic strategy is insight development.  To do this, you first need to be able to convert your detail-level data so that you can make apples-to-apples comparisons. You also need to link your objectives to your executional tactics and determine the best KPIs across platforms, strategies and tactics. If you accomplish this, you’ll continuously build and refine your strategy, creating a constant innovation cycle in optimization.

Multiple platforms can help. Running your activity across multiple platforms means you can test, learn and apply the learning on a continuous basis, without losing scale or being overexposed to your platform’s SSP mix. You may find, for example, that one platform is better at delivering against an audience within a certain price range, while another is great for back-end conversions on your site. Providing this type of cross-platform insight allows the development of a proper test-and-learn strategy, a cornerstone you can use to build long-term marketing plans across programmatic media.

Tapping Into Innovation And Evolution

All the changes happening in programmatic buying come with new challenges and opportunities. As new technologies and platforms evolve, for example, both demand- and supply-side technologies will have to improve and differentiate themselves. Some recent examples of new mediums include radio and digital billboards. Each of these new additions will result in new platforms specialized for that medium, so the ability to leverage multiple platforms — and tap into these new platforms as they arrive — will increasingly become a competitive advantage.

The supply side of the ecosystem already is pushing forward with a multiple-platform approach: Some publishers, including premium ones, are signing up with multiple SSPs to manage their inventory to optimize internal yield management and enable advertisers to find the inventory in more than one place. I’d argue that, now, the demand side also should be making the switch. A multiplatform approach would stimulate innovation on both sides and enable the delivery of better performance for clients.

Follow The Exchange Lab (@exchangelab) and AdExchanger (@adexchanger) on Twitter.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Good article. Thanks. I agree with a lot of the thoughts shared here, interested to know what your opinion is of running the same exchanges on multiple DSPs. Clients always ask whether this will result in double-bidding and wastage, but in my opinion it enhances scale and, due to the differing bid win rates on DSP platforms, it most likely allows you to hit different inventory as a result.

    Have tried blacklisting certain exchanges on one DSP and whitelisting the same on another to avoid this, but I’m not sure it’s as good a strategy as running them all on multiple exchanges together. Interested to know what others think about this.

  2. Although trading desks should absolutely be using more than one DSP, I am finding it difficult to agree with testing a campaign on multiple DSPs at the same time. I think this can be a dangerous practice that can result artificial cpm inflation – as the buyer is running the risk of bidding against themself.

    Most (major) dsps, can hit the same inventory – with the main exception being the inability to reach FBX via Google’s Doubleclick Bid Manager. In order for the test to work successfully (to avoid the risk of bidding against oneself) exclusions on certain ssps and inventory would need to be setup in the the orders to make sure none of the same inventory is being bid on in the different dsps. This kind of setup would be quite tedious and doesnt
    seem very efficient.

    Testing on multiple dps on the same campaign also can be done during different time periods. Though it is is not exactly apples to apples(depending on the cpm averages and trends in the market during the different time periods), it can give you an indication of which dsps works best for which type of campaign. Perhaps thru testing you may find strict
    performance based campaigns (in general or perhaps for a specific segment such as auto) works better in one dsp, whereas more branding type campaigns with less strict goal works better in another.

    The above mentioned reasoning of a dsp only submitting one bid on behalf of all buyers of the dsp is not the case if the buyer has their own buyer ids linked to the ssps they are bidding on. A buyer having their own buyer id allows (in addition to that buyer
    id submitting their own bid next to to dsp seats bid rather than competing against it) the publishers transparency into who is buying their inventory and at what price. This can help strenghten the relationships between buyers and publishers and if a buyer is buying a substantial amount of the inventory of a particular publisher/network, could open the door for opportunities to get inventory with a sense of exclusivity (be it on a floor prive level or specific inventory).

    Again, I completely agree with a trading desk’s need to use more than one DSP – but would advise not to test one campaign across multiple dsps at the same time. Not unless proper settings are put in place to avoid cpm inflation.

    • I totally disagree, with the billions of ad requests happening during every bid, its almost impossible to artificially inflate the CPM.

      I agree with the author on this one, with Publishers maximizing yield with multiple SSPS it makes sense to do the same on the demand side.

  3. Testing multiple DSP’s simultaneously will hurt your campaign, it’s call self-competition. Running multiple bids in different DSP’s will increase the likeliness that your bidding against yourself, driving up CPM’s that ultimately can affect your performance. If you test a Retargeting campaign with multiple DSP’s that has access to the same inventory like FBX, you’re most likely bidding against yourself and you would’ve won that bid at a much lower CPM if you would’ve consolidated. If you do decide to run with multiple DSP’s, make sure efficiently plan it out to give DSP exclusivity when trading, such as day-parting, geo-targeting, etc. so your DSP’s are not overlapping each other.

  4. I think it’s funny how every comment is, “ooo ooo your going to compete against yourself”. This may be the case to an extent and more so in certain situations but if you bid .51 in three systems, the most you’re going to pay is .51, once you find out which ones works the best for you (on price and KPIs), you consolidate to that one system and see how much you can get out of it. Then see if it makes sense or not to turn some of the others back on (scale, reporting preference, etc.). You need to see who has the better set up for your pool of inventory and I think the commenters about competition missed this part:

    “Every DSP submits its bids differently, which means that all bids are not created equal. Because of this, a $2 bid in Appnexus will not necessarily beat a $1.50 bid from The Trade Desk for the same inventory. A bid must win internal auctions within its DSP before it even makes it to the SSP auction, when the SSP decides which of those bids, from all of the DSPs, will win the impression. DSPs consider budget, performance against target, time of day, advertiser category, publisher block lists and minimum bid, to name just a few of many, many factors — beyond price — that can alter the likelihood of an advertiser winning a bid.”

    Without testing it you don’t know jack about who is going to be the cheapest or best performing and I think the above factors are going to trump any competition you create for the first few days of the campaign. Unless you are very familiar with the pool of inventory, the budget is small or the targeting is too niche; it just doesn’t make sense to put all your eggs in one basket, every time. This is however, a form of investing.

  5. @ Jamie – I see your point however if you set a flat bid of .51 cents in multiple dsps for the same campaign during the same time period you still can cause cpm inflation. As setting a bid of .51 doesnt mean you would necesarily pay the .51 (as the winning bid is usually one or two cents above the second highest bid. In this case your bid would be the second highest bid so chances are you would always pay the .51 or so (even though you may have been able to get it for lest). Also, setting flat bids in dsps(or at least some to my knowledge) as opposed to dynamic cpms will prevent you from taking advantage of the dsp’s algorythm and its ability to set a bid based on the different variables its seeing on the available impression available at the locatioin is bidding at that particular moment (other types of data the algorythm uses in its decision making process). Again I agree on testing by all means – but I just think it needs to be done responsibly and not at the expense at the advertiser (having higher cpms than necessary).

  6. Veerle

    Thank you for your comments, it’s always good to get some interaction and discussion flowing around these articles. I do think it’s important to evaluate and have a considered approach/ awareness around the potential of bidding against yourself, however I agree with James on this point, the marginal chance of this event occurring is unlikely due to the vast number of permutations in play from both the demand and the supply side. Coming back to part of the main focus of the article; addressing the scale problem, it’s not about using limited test or small budgets across multiple DSPs; it’s about delivering scale beyond a single one and taking a proven strategy into a larger ecosystem. Analysing your overlapping publisher mix will enable you to make smart decisions on where to take the potential risk of bidding against yourself versus the potential increase in scale and cost effective difference in the buying points. A multiple DSP strategy does require consideration of more factors, but in our experience the benefit outweighs the extra work you need to put in during your optimisation.

  7. James–I appreciate your response to my comment, like I said testing multiple DSP’s simultaneously should be done with thoughtful consideration so you can decrease your chances of overlapping your efforts.

    I don’t ridicule self-competition when it comes to my advertiser’s money, I make sure that if their running on multiple DSP’s that they do it w/ careful consideration.

  8. Agree with most of the commenters. Using more than one DSP for a given campaign and tactic is not a good decision. Forget bidding against yourself, but unless you have a very strong attribution algorithm running downstream from both of them, you’re pitting two DSPs against each other in a last-ad-seen game. Anytime a DSP knows it’s in a last-ad-seen shootout, the thoughtful marketing goes out the window and the games begin.

    I completely agree that every advertiser should have multiple DSPs in place and at their disposal because they each have strengths and weaknesses. Within a given campaign and tactic, though, use the benefit of single-source frequency management and thoughtful, easier attribution to produce better RTB results.

  9. seb robin

    hi mate
    interesting thoughts & comments here..

    now based on my 6 years experience i can just say the following :
    1/even if you “fear” to bid against yourself..there is always another buyer (TD, advertiser, adnetwork) bidding ALSO against you.. so there is a little chance that you bid against yourself..you rather bid against others more than against you..

    2/ if you’d like to compare 2 DSP..there is 2 ways : either ISO- campaign..(same budget, campaign, creatives, goal, max bids, strategies and so..) and you see the results of who’s getting the most out of your constraints of campaigns..or (if you fear self competition..) again, same campaign constraints, at different period of time..now the minus in this second option is that since time period is different, so if the competition you face..then you do not have the correct “testing parameters”..
    there is no ideal situation in fact..but year after year and after thousands of campaigns (or sooner), you just realize that somehow..each DSP’s different from the other in the way it manages your spend and in the way it chooses “audience and publishers/exchange/ssp” to deliver against..
    that’s kinda logical if you go back to what makes the difference of each solution..they were not built the same and their algos are just different from each other….it’s like a car… it may drive you to your destination but with different +/- depending on the way you drive..so in the end it’s a personnal choice too.. but definitely a multiple DSP is the norm..

    now my question would rather be..if you’re an advertiser or a TD..how do you choose which DSP will run your campaign if you have 2 or 3 solutions available ? what kpi’s do you use to choose ?

  10. CPM inflation aside, how does one control frequency caps with multiple DSPs? While I can agree that multiple DSPs can offer scale and potentially better algorithms, I doubt the incremental gains are enough to negate the downside (CPM inflation, frequency caps, conversion deduplication etc.)