Reducing Private Exchange Friction With a New Protocol Upgrade

casale-ddt“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 by Andrew Casale, VP Strategy, Casale Media.

Private exchanges were designed to bring publishers and buyers closer together, address concerns of channel conflict, and ultimately increase inventory value. We have yet to see this model create a unanimously positive feeling from both buyers and sellers. I contend this is largely because of friction that is inherent in the model of putting one’s most valuable opportunities behind a manual process.

The first wave of the private exchange was based on preferential treatment that gave top-tier buyers a first look at ad space in return for higher prices. However, this was met with lukewarm appetite by buyers who did not see enough value generated to justify higher pricing. Sellers of private exchanges adapted, and the latest iteration of the model now provides additional value in the form of more transparency, unique ad opportunities, unique inventory, and/or data in return for higher bid prices. You’d be hard pressed to find a buyer who wouldn’t see value in any of those categories. And while there are now happy buyers using and benefiting from this model, the process to buy needs improvement to reduce friction.

Last year many of the supply-side exchanges, myself included, proposed that bidders move toward supporting multiple bid responses in the real-time bidding protocol. After months of debate, multiple bidding is finally starting to become a reality with several key demand-side platforms now supporting it. The result? As expected, the bid stream becomes denser with this capability in place, which is leading to a higher volume of impressions cleared through the exchange. Now I propose that we continue on that path and look at the benefits multiple bid requests could create—another valuable protocol upgrade.

Today when a banner renders on a webpage that is activated for sale by an exchange, that impression is rapidly transmitted to every connected demand-side platform in the form of a bid request. The bid request is a static representation of the impression, and effectively grants DSPs an opportunity to bid. Part of the problem with the protocol is that it is based on one “static representation” of an impression. It’s all or nothing, and a supplier has to effectively pick how they want to represent that single opportunity and its reserve price to all buyers. When inventory is deemed premium or presents the possibility for channel conflict, more often than not that single opportunity is stripped of all value, made opaque, and all value is reserved only for transactions that are developed via a manual protocol (i.e. dealID).

There are only so many on the buy side that can effectively strike and execute private deals, and there are only so many publishers who justify the effort. For everyone else, we remain in a state where the best opportunities are rendered unavailable if the protocol does not change.

This is where multiple bid requests come in. One impression can spawn multiple opportunities, in parallel and in real time, each with their own reserve price. For example, a completely opaque impression can be offered up for bid at a low reserved price. Buyers won’t know what they’re getting, but the cost is low. That same impression can be made available as a semi-transparent request, where the buyer will know the domain, the reserve price is a little higher, and the buyer gains a measure of comfort about the environment of the impression. A final option can be full transparency at the URL level with the reserve price set at the highest amount. Different buys will start to fit into different buckets of desirability. And so while a DR campaign may be fine with the low-cost opaque option, a CPG buyer who would have passed on the option in fear of brand safety now has the opportunity to bid transparently. Today the technology has evolved enough to handle these various scenarios in real time. And while each bucket will not necessarily get a bid, the benefit here is we give all buyers a shot at all opportunities without manual work.

The argument against this would be that by adding multiple bid requests, the overall “perceived” supply in the market will expand. Today many remark that we already have practically unlimited scale at a pace of some 600,000 bid requests per second and growing, so why would we want to create more? The reason in my view is because the volume of premium, quality impressions is exponentially smaller than the entirety of display. To counter this reality, we want to give the most valuable publishers a louder voice in the sea of supply so that they can better stand out, as they rightfully should. By adding more bid requests we can expand the potential value that each independent impression can create for buyers. Transparency and price are really just one dimension where this can be applied. It can also be used for bigger formats – addressing the fact that supply is still dominated by only three static ad sizes, and moving us toward a model that would support something as innovative as the IAB’s Rising Star ad units. If every option has its price, why not make bigger formats available? If they are available to be bought without friction, at the publisher’s price, why not allow the same space—that would otherwise sell a less impactful unit at a lower price—to work harder?

If things stay as they are, we will continue to run into the friction created by sellers opting to create more value by tucking their best opportunities in behind a manual process and away from the open marketplace. Instead, if we start to sprinkle in new opportunities, where they are warranted and where the technology supports it, the programmatic buying channel will further grow its share of the media plan. I am not suggesting that multiple bid requests should be applied to every publisher, but the biggest and most well known could stand to benefit exponentially.

Follow Casale Media (@casalemedia) and (@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. Jeffrey Goldstein

    This sounds very familiar. I think I pitched this all to you guys a year ago?

    • Shael Fryer

      @Jeff – Either that or great minds just think alike!”

  2. Fabien Magalon

    I really like the idea.
    One concern only: how many bids requests per impression is acceptable ? I guess there needs to have a stop at some point.
    Benefit of using Deal ID today is to add unlimited additional dimensions and matching price-floors to every single impression. Not sure we could mirror this with multiple bid requests. I still like the idea a lot though.

    • Andrew Casale

      There would definitely need to be an upper limit or if left unchecked this could create more problems than it would solve.

      My thought isn’t for this to replace dealID. The later is incredibly valuable because of its simplistic design – that it can be applied to anything. Rather, the goal here would be to reduce the most common usages of dealID today which create the biggest burden so that we reserve dealID for only more complex executions.

  3. Interesting concept, Andrew. Do you not think the DSPs would quickly learn to group the three requests (transparent, semi transparent, opaque) as a single opportunity, using the transparent data to inform their decision but responding only to the opaque opportunity, as it could be had at the lowest price point?


    • Andrew Casale

      That’s definitely a valid concern. But if this is a direction we want to take to reduce friction which all parties would benefit from, it’s one I think all parties have a vested interest in resolving. The exchange can safe guard through obfuscation, or contractually the DSP can be obligated not to do this. Technically the same concern can be thought of with the usage of dealID, if a DSP knows the rules that govern a dealID they could make use of its designated purpose in unauthorized scenarios.

  4. I imagine this concept has two customers with two different requirements:
    On the buy-side, DSP’s will offer their clients the ability to purchase inventory at different price points across different targeting criteria.
    On the sell-side, SSP’s will need different pricing rules for their publishers inventory to match the bids and the levels of transparency.

  5. Jeremy Randol

    Yup, this was the tiered auction concept we’ve been thinking about for a while. As you know, we love the idea. Doug raises a point that some DSP and SSP partners have shared. Would love to better understand a path forward.

  6. Multiple bid requests has some distinct disadvantages. It triggers the danger of swelling supply in the various counting systems, as well as increased operational QPS loads. In my opinion it’s the wrong solution.

    I do agree with Andrew that we need to look at a protocol upgrade. Currently there is a draft within the OpenRTB dev group for a protocol extension for PMP. The draft describes a pmp child object that will contain an array of DealIDs, a type field, a pacing %, a price floor and possibly a list of data tokens for targeting. The proposal is still under consideration. This in combination with multi-bidding in responses will allow bidders to send back full bid responses for any of the DealIDs as well as open market bids.

    Rubicon has already extended our pre OpenRTB API with this object. We’re also going out to PMP capable DSPs with best practices on how this might be used. The most important point in this is an ideal bidding logic outline describing handling of DealID, SeatIDs and how the private flag should be interpreted in concert with multi-bidding.

    If we can get behind the outlines of this solution the RTB protocol will have evolved to support both second/reserve price auction as well as rich direct order automation between a buyer and a seller.