Real time bidding (RTB), also known as programmatic (ad) trading, is for many intents and purposes the nom du jour of the digital advertising industry. The spotlight has continued to grow with successful recent IPOs of Rocket Fuel (FUEL) and Criteo (CRTO), and upcoming ones like Rubicon (RUBI). A common analogy for such businesses is that of the high frequency trading that is routinely observed today on most major securities exchanges, e.g. the NYSE, and which has even made its way into academics' simulations of agent-based market models. Providing a little historical context will illustrate how we've gotten to this point in advertising technology and why market bellwethers like Kellogg are experiencing multi-bagger ROI gains from the utilization of RTB service providers in their ad budgets.
In ad exchanges, one may think of the aforementioned "securities" as quantifiable increments of advertising inventory (spots, if you will) amid various evolving channels: TV, newspaper, radio (traditional); banner display (early/static Web); and mobile, video (Web 2.0, targeted). The evolution toward the latter is a function of technological (Moore's Law, 4G/LTE) and social (chat, text, tweet) factors. These elements describe a trend toward accelerated content delivery and consumption with far greater categorical specificity against a rapidly shrinking time scale. Short answer: more content, faster and on-the-go, to a very specific and transient (if you're not quick) set of eyeballs.
Buyers generally represent advertising networks (or ad agencies in the old, traditional sense) who bid against content publishers' (seller) allocations of ad space. Traditional media inventory transactions were negotiated between agencies and publishers separately in turn (think Mad Men), but the acceleration described above has necessitated a shift toward effectuating greater volume today via a growing use of ad exchanges.
Ad exchanges are synonymous and largely synchronous with online, and as such, have evolved concurrently with the prevailing software. One of Rocket Fuel's salient themes in pitching their investment thesis during the 2013 IPO roadshow was its optimal positioning on the leading end of ad software development that evolved from an age of databases & apps, through Big Data to autonomous AI. The first two ages underscore the inherent value add of scale and volume within ad exchanges. The third reinforces the former two and adds a component of efficiency and transparency heretofore unseen in these markets. The RTB infrastructure forms the basis for the latter age.
We can now refine the generic buyer/seller terms to the topical nomenclature of today's ad exchanges: DSPs and SSPs (demand and supply-side platforms), respectively. A DSP or SSP generally employs RTB and essentially functions as a 'trading desk' in that it can uniquely bid or offer more precise lots of unsold advertising inventory with high frequency and precision. Low frequency platforms that don't use RTB will be more inclined to buy or sell in bulk at pre-negotiated wholesale rates along longer time scales. Unlike the world of securities exchanges, where sharp traders can still develop an edge building concentrated books with low frequency fundamental or chart-based trading, there appears to be very little such alpha in trading against such asymmetries. The time factor is too critical.
The first order pricing mechanic between DSPs and SSPs here is generally set on a cost-per-mille or thousand impressions (CPM) basis in which inventory is readily quantifiable and derived from a content network, as opposed to a banner (e.g. search engine result) with a less defensible yield structure. Supply is scalable because direct ad performance (clicks) does not usually constrain SSP/publisher capacity. RTB initially emerged as an answer to demand-side needs among advertisers, and generally remains more diffuse and transformative on the buy side, for now. Hence we see the emphasis on 'bidding' in the titular term.
Whether DSP or SSP, the best ad tech service providers will minimize transaction costs, optimize inventory, and maximize advertiser/publisher ROI (conversion, click-through, recall), thereby commanding the strongest valuations. The former is a function of lower bid-ask spreads, greater price discovery, and precision-based adherence to stricter orders of market efficiency. The latter two are emergent properties of the positive feedback loop established via machine-learning. In other words, the DSP or SSP AI utilizes prior and ongoing ad campaigns to continuously improve its predictive modeling, combing through petabytes of data for quality inventory or bidders (i.e. no bots) through which to target or deliver user-specific demographic/location/weather data in real-time, learning from each exchange. As always, transparency is the name of the game.
ROI gains to advertisers and publishers from the use of RTB campaigns have continued to be cited at outsized levels of return. Back in 2009, Turn reported advertisers seeing up to 135% improvement on click-through rates and 150% improvement on conversion rates. PubMatic, an SSP, was reporting 2009 RTB gains of 64% among publishers in the earliest stages of its pilot. In late 2012, Kellogg's associate director of global strategy reported in a Forbes interview that digital media ROIs have increased as much as six times across their brands. Rocket Fuel's management commissioned a three year study from Forrester consulting that determined FUEL's RTB-specific value add to be an incremental 144% over an unnamed group of ad networks (presumably non-RTB for the most part). Lastly, in Criteo's December '13 investor presentation, they cite a specific case of 28x ROI in a single quarter for a major multi-channel US retailer.
The autonomous AI feature gives incumbency a sustainable comparative advantage at virtually any scale or sequence of the ad exchange. All proprietary algorithms are not created equal, but if talent levels and inherent management of the software engineers are not wildly divergent, a more seasoned RTB infrastructure should generally remain 'smarter' over time than a newer variant, barring some game-changing innovation or coding master class.
Consider the following: FUEL, founded in 2008, presently trades at 3.7x forward revenue. CRTO, established in 2005 but focused on CPC (click) based ad inventory and less premium display in the EMEA region, still holds a 3.1x forward multiple. FUEL actually attributes less than 20% of current revenue to social, video or mobile channels. The under-penetrated channel mix is likely part of the low-hanging fruit which the market is pricing into these strong multiples.
Indeed, by 2017, eMarketer forecasts that video advertising markets ($20BN today) are expected to more than triple in size as video becomes the preeminent delivery channel (40%) for digital advertising, reaching two billion people worldwide. IDC expects RTB ad spend to be $20.8 billion in 2017, reflecting a 51% CAGR from 2012 due to further adoption in transacting premium inventory.
The punch line is that video market CPM in Q3'13 was about $11 vs. $1-$1.50 for mobile and other display. We would expect that CPM gap to obviously narrow somewhat as mix shifts to video, but the delta is so large that further adoption should invariably be accretive to well-positioned DSPs and SSPs with a viable RTB product. Let us assume that FUEL's channel mix follows the market and video becomes 40% of revenue in the next 3-4 years. Pro forma for the CPM mix shift, that top line could be double and the sales multiple closer to 5x. This pro forma exercise all assumes zero growth in ad impressions, so you can imagine the actual 3-year forward multiples would likely be another few turns lower. FUEL was trading at 5.7x forward 2014 revenue around the time of CRTO's IPO.
That's all great for the ad tech darlings like FUEL, but where does this leave everyone else? Are FUEL and its similarly sized peers too far ahead of the game that they will decimate any smaller incumbents or potential new entrants? Not necessarily.
The Big Data opportunity set is measured in exabytes and the best RTB operators (proxy or in-house) appear to just be cracking single digit or teen petabyte volumes for running their DSPs or SSPs. CRTO's IPO prospectus cites IDC in stating that "from now until 2020, the digital universe is expected to double every two years [a cumulative factor of 300 vs. 2005, reaching 40,000 exabytes or 40 zettabytes]." There are myriad categories within Big Data from which to choose, e.g. auto, travel, sports, and an ever changing zeitgeist of consumption patterns for advertisers and publishers to evaluate.
Established AI-based operators like Rocket Fuel may learn more quickly if their managers acquire into these smaller pockets of RTB capability while their brand equity continues to reflect an attractive cost of capital. Synergistic combinations of programmatic vendors and content (video) platforms can effectuate superior monetization of requisite ad inventory via in-network traffic that is available at a discount to outsourced acquisition costs. Meanwhile, smaller early-stage operators who develop an established niche and/or robust end-to-end solution may scale appreciably on their own with the right execution.
One early-stage public company in the RTB space is Adaptive Medias, Inc. which acquired Ember, Inc. in December to power the SSP's inventory in mobile, video and online display advertising. The Irvine, California-based company partners with mobile app developers, video content providers and website owners who have a supply of ad inventory and puts this inventory into its RTB marketplace to be matched with advertisers. Adaptive reported Q3 2013 revenues of $373,737, up nearly 14x from the second quarter. Its market capitalization huddles around $12MM, a level which we believe is a very compelling entry point for potential new investors.
RTB emerged out of a desire to improve performance in ad campaigns, but the buck does not stop there. We tend to agree with Andrew Tu's comments in a recent ClickZ interview that the "Holy Grail behind all of this is that the brand dollars will shift to programmatic faster for video than display." As always, correlation does not imply causation, but we suspect that RTB and video are a great fit for the immediate future.