“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 Mike Driscoll, CEO of Metamarkets.
Since late 2009 and the advent of the RTB era, online marketers have had access to a tremendous tool. RTB has allowed them to dynamically address audiences with vibrant, engaging advertising based on demographics, purchase intent and product searches. This real-time revolution has given marketers much greater targeting success than the days of relying on content.
Yet while RTB exchanges route billions of ads around the globe in milliseconds, sophisticated reporting on these systems takes days, consisting of cobbled together e-mail attachments shuffled among the stakeholders of publishers, platforms, DSPs, and agencies.
- Responding to algorithm crashes. On August 2, 2012, Knight Capital's high frequency trading algorithm broke catastrophically, causing the firm to garner $440 million in unwanted stock traded in just 45 minutes. While it's unlikely in the near term that a broken algorithm in a DSP is going to cause hundreds of millions in losses there have already been cases where an RTB buyer's system has gone on an unwanted buying spree that has remained undiscovered and caused net losses for that buyer's campaigns. With a system that shows these buys as they happen, these problems can be forestalled, and with learning systems, eventually eradicated.
- Detecting fraud. RTB marketers encounter two main sources of fraudulent behavior in the current exchange environment. First, the proliferation of third-party data sources has allowed sub-par data sellers to enter the market. These sellers are duping marketers faced with a bewildering variety of data choices into believing that “bad” data is actually of the type, and derived from the source, that the marketer is looking for. While I won't claim real-time analytics can change the nature of a data stream, it can allow marketers to much more quickly see performance degradation if bad data or inapplicable data is introduced into the optimization system, and change the campaign set-up. Since it's unlikely there will ever be a "data rating" service, Real-time access to traffic data is the only preventative that will work. The second type of fraud is better known: click fraud. Now, click fraud is not often discovered until reviews of monthly invoices are completed. Once again, if a media buyer can see, in real time, that a particular site or publisher is posting questionable click-through rates, she can immediately pause buys on that site and avoid any future disputes over an invoice, or avoid a claw back of previously paid money. Managing a real-time content distribution system is difficult enough without the added burden of also managing shifting cash flows to account for exogenous factors.
- Real-time optimization. I talked about "inapplicable data" in my last bullet. This problem is exacerbated for the sell-side, where SSPs are trying to leverage data sources that were designed for targeting, not pricing. It's a well-known secret that on the sell-side price floors are set through a little bit of historical analysis, a little bit of publisher intervention to preserve direct sales price levels, and more than a little guessing. With analytics in real-time, that guessing can become estimation done in iterations in the course of hours, rather than days or weeks. Until there are machine learning systems that can accurately predict bids and adjust floors responsively, analysis in real-time is the best way for publishers to ensure they are getting adequate prices without incurring too much dead weight loss. Loss caused by making poor price floor choices could cost sell-side RTB exchanges some of their premium publisher relationships.
- Guaranteed Buying & Futures. Sell-side exchanges are trying very hard to enable their publishers' direct sales forces to sell on a guaranteed basis on premium inventory through private marketplace or private exchange set-ups. As private marketplace gets wider acceptance in the publisher community, you can also expect these exchanges to start selling derivative products as well, just as in equity markets. The first will be options contracts for future placement of banners on a guaranteed basis. A media seller cannot track guaranteed sales (where the buyer will guarantee to buy a percentage of total inventory offered through a first look private marketplace) with reporting that takes several hours to collate. Without real-time analytics and forecasting, sell-side and buy-side RTB partners will risk failing to fulfill a guaranteed buy. Media buyers for premium campaigns won't put up with those companies for long. Moreover, for futures options, just ensuring that the contract starts properly will require a real-time view of exchange traffic. God forbid it takes several days to realize that a futures campaign hasn't gotten started at all. Restating revenue for already recognized options contracts is not something any company wants to go through.
There are numerous other reasons for RTB buyers and sellers to give their account managers access to a real-time analytics system, not least of which is the sophistication of the system that you can show off to potential advertiser and publisher client prospects. The vast majority of online and mobile marketers agree that real-time bidding is the future for banners, videos and even newer ad units. Only if we complement these real-time platforms with real-time analytics will we achieve the full promise of this advertising revolution.