I’ve been threatening on twitter (@MarkMcEachran) to write a book on the things publishers need to know about Real-Time Bidding (RTB). At the Rubicon Project, we like to talk about ‘RTB Done Right’ and how to help publishers understand what’s really going on. With a full-time day job I don’t have quite enough time to bang out chapter after chapter of really fascinating business and technical aspects of RTB, so I thought I’d do a Top 10 list. I started watching the Late Show with David Letterman to get the format down and then I popped over to a bunch of sites to make sure I wasn’t producing a fluff SEO piece. So based on what myself and our team has learned, here are the Top 10 Things Publishers Need To Know About Real-Time Bidding.
1. Transparency is a two-way street.
Some publishers are concerned about the data being passed to DSPs when a bid is solicited (URL, namely), but it should be noted that the DSPs deliver valuable information with each bid. When DSPs bid on an impression they send along the advertiser’s name and the maximum price they’re willing to pay for the impression. Standing alone that might not seem like much information, but if you keep track of all their bids you can build a picture of demand down to the advertiser level. Analytics packages on the SSP side are starting to take advantage of this data, exposing it to publishers in a variety of consumable formats. These reports are actionable in many ways; like increasing the floor for less desirable advertisers and, perhaps, lowering floors for quality advertisers or advertisers with a special relationship.
2. Floors are important.
Liquidity, liquidity, liquidity! I hear them screaming all the time; some DSP’s really, really hate the concept of floors. The fact is, though, you can capture more dollars in many cases with higher than expected floors when your floors act as a second bid in the absence of any other.
Consider the scenario of four impressions, and there is only one bid for each. This is a reasonable case in such an imperfect market as RTB. Two impressions each get a bid of $5 and the other two get bids at $20. What’s the best floor here? If your strategy is the capture as many bids as possible then you’d put your floor at $5, and you’d capture all those bids and earn yourself $20 (all four impressions fill at $5). But if you set your floor at $20 instead, you’d capture only two of the bids (at $20), and you’d have yourself $40!
How does this play out? In the short run we’ve seen the development of Rubicon’s dynamic price floor technology which can read the tea leaves of the market and maximize revenue in RTB. Simply put, dynamic price floor technology solves for the imperfections in the market by aligning floor prices with known demand throughout the user session. You can then pepper in expected yield from other demand source to balance out what you collect on the wasted impressions. “Waste” is an important concept in this market. The way I look at it is, in many ways, like DeBeers looks at the diamond market. You bring your best quality inventory to the premium market (Direct, DSPs) and let the rest go to industrial uses (ad networks) or to your secret vault (house ads, for us).
3. Data Leakage is a potential risk.
I’m going to go on record, it’s really out there. In RTB it’s really, really out there. I’ve seen it, I know how it works and I know where it doesn’t. First let me set something straight on one of the things that shouldn’t be scary but sometimes looks scary. Pixel drops by SSPs to synchronize user IDs with DSPs is not causing data leakage, at least when it’s done right it’s not. A given DSP only sees a user once every 3-30 days. They don’t get to follow that user to every page using this sync process so they can’t really build a profile on data gleaned from a single page hit every 3 days. Maybe they could figure out that the user has an affinity for the word “the,” but not much else.
Data leakage, unfortunately, happens out of sight for us. A nefarious DSP or other party plugged in to a DSP can read the stream of pages that come through in bid requests. These are tethered to a user ID or some unique combination of request parameters (IP + Browser + OS + Language, for example). So with a page URL, an identifiable user and tons of bid request volume a nefarious entity could reach out, scrape pages and build an audience profile for the users without any compensation to the publishers for the courtesy of the impression data. This type of activity is very difficult for SSPs to police from a technical perspective, but there are usually contractual obligations that DSPs are adhering to that, legally, should discourage it.
4. Not all DSPs are bidding down your inventory.
The truth is that most DSPs are bidding at pretty static prices for their campaigns. The prices might change based on which campaigns have filled and what campaigns are targeted to that type of user or inventory. A lot of the DSPs are using tried and true adserving technology logic as their bidding system. These things will evolve over time to bid in more intelligent patterns, but this is still a young market.
5. Most SSPs are using a Second Price Auction model.
If there’s anything that AdSense proved, it’s that the second price model drives up bid prices while delivering fair market value for the thing being purchased. The SSPs used this model as a jumping off point when they developed their systems. A few controls have been put in place to protect the publisher inventory from certain failings of the market. It’s imperfect, you see. An SSP can’t always expose every impression to every possible buyer. There are simply too many things that can go wrong with the Internet. As a result the current state of most of the RTB universe is that it’s a second price auction with floors.
6. Your late-in-the-session traffic is not magically worth more in RTB.
As with the traditional ad market, in RTB the high paying campaigns have frequency caps. Once those caps are exhausted the bid prices will go down. Enabling RTB on your inventory will raise the overall CPM, but it’s usually a result of parallel exposure of the impression to demand early in the session. Once you get into the later parts of the user session the premium campaigns are gone and you get the same mixed bag of stuff that you’re used to seeing.
7. Private Exchanges are in their infancy.
First off, what are private marketplaces? Simple, RTB-sold impressions with rules as to what a buyer (DSP, agency, trading desk, or advertiser) has access to what inventory (or audience), at what priority and at what price floor.
Today a private exchange is more about restrictive access than it is about doing deals with agency trading desks. There are two ways to go about this: the first is to just allow certain DSPs to touch the inventory and then ask them to restrict the buyers of that inventory on their side. This can come with or without a billing work-around, but that’s not really a publisher problem from an operations standpoint. Floors can be enforced or not, it depends on the deal. This is a phone-intensive operation generally handled by trade desks with four different companies in the mix.
The second way is to simply set up priority access around certain inventory, like your homepage. Some SSPs have a simple priority to buyer exposure. Imagine the three buyers get a first priority auction, if they don’t like the impression enough it goes to a second set of buyers and another auction is run, and so on. This is very general, but you get the idea. The technology and the mentality around the Private Exchange are evolving to increase efficiency so that impressions and money can flow a lot more smoothly through them. The next iterations should be interesting.
8. RTB alone should not be an island of demand.
Very few publishers can claim that RTB monetizes all of their inventory, and those that do might have suspiciously familiar CPMs. Today you can expect to monetize 10-50% of your inventory via RTB and still see the benefits of higher overall CPMs, or you can expect to monetize 50-75% of your inventory at less than optimal CPM’s. It’s a result of a storm of demand from many sources. RTB may provide the wind, but you still need the ad network’s rain and some hail, thunder and lightning wherever that may come from. Any worthy supply platform will be doing their best to provide publishers with the perfect storm.
9. RTB is still growing and growing and growing.
Ad networks appear to be entering the pool in earnest now, and more and more inventory is becoming available in the space. As we saw with new users getting online with a computer modem, this has a similar network effect of making the market, and all the stuff in it, more valuable. The efficiencies for buyers and sellers are talked about a lot, but what most folks don’t consider is that the users are probably seeing relevant advertising more frequently.
10. RTB is not the end
RTB is more like a tipping point, than the end of the road. Early advertising technology companies created ad servers, audience builders and campaign management systems. These technologies evolved slowly, being reused and slightly retooled to optimize online advertising for particular purposes. Serialized impression exposure with pass-back URLs seemed to be the epitome of monetization. But the tools took a long time to take the leap to parallelized impression exposure and bidding. What was required for that leap was an unprecedented amount of technical coordination amongst buying and selling platforms. Now that the engineers are talking to each other innovations beyond RTB are, likely, right around the corner.
So there you have it. I hope you’re not overwhelmed. Keep in mind that you’ve got partners working hard for you to understand all this and keep you in the black and safe.