Digital video technology is enabling the cable industry to explore new opportunities in advanced advertising. These opportunities challenge the traditional broadcasting model in which advertisements are presented to a general population of viewers with very little direct feedback or measure of success.
How has the industry moved toward addressable advertising and what are the components and interfaces needed to build and leverage such systems successfully?
The following discussion first tackles those questions, then describes an example of how the ad placement decision process can be adjusted as viewer preference data is collected, and closes with an update on current standardization efforts, which are critical given the need for flexibility and open interfaces in this arena. Advanced advertising Technology has advanced on a number of fronts that enable advanced advertising systems. The first and most fundamental is the replacement of pure broadcast delivery with digital pointcasting and narrowcasting—for example, video on demand (VOD) and switched digital video (SDV) services. These technologies allow customized viewer experiences and can allow advertising campaigns to target specific individuals and populations with greater effectiveness.
A second important technological advance is the establishment of a “back channel” or “upstream” control path from the viewer to the headend. This allows viewer choices to be solicited and monitored and provides valuable advertising feedback. Data can be collected on subscriber behaviors, such as purchase history and program viewing history, and can be used to deliver a more relevant advertising message at the next opportunity. It is well known in the advertising industry that any feedback that narrows the general population to groups that have expressed an interest in the product or related products is an extremely powerful leverage point. We refer to advertising that adapts to viewer feedback as adaptive advertising.
To support adaptive advertising, the cable industry will require flexible and standardized system architectures. Multiple servers from multiple vendors will perform the following coordinated tasks:
• Discover and characterize advertising avails that arise in broadcast, switched digital or VOD channels.
• Decide which ads to place in which avails. Changes in programming and ad effectiveness will drive these decisions.
• Insert and transmit ads to viewers. Today’s digital program insertion (DPI) systems rely on coordination between ad servers and splicers. Variations support VOD, SDV, etc.
• Collect feedback from viewers and measure preferences. Existing feedback merely verifies when ads play.
• Manage avail inventory and sales to advertisers. With good success metrics, the value of avails will be more dynamic.
• Prepare and distribute new content. Increased dynamics will drive more creation and distribution of ad content.
The SCTE Digital Video Subcommittee Working Group on Digital Advertising has generated a number of successful standards supporting digital advertising and is currently working on standards in these key areas. This work will be outlined in a later section.
Advertising content is also expected to undergo a fundamental change. In addition to video and audio tracks, MPEG content can contain data tracks. This allows the traditional 30- or 60-second “spot” to be augmented with barkers, scrolling marquees, and other graphical overlays. These dynamic overlays will serve as invitations for viewer feedback and will increase the data available when selecting ads. The capability to follow through with email, phone contacts and pointers to longer-format content will also play a key role. Delivery mechanisms To support addressable advertising, delivery mechanisms must direct specific streams to specific sets of viewers. In addition, to support adaptive advertising, systems must support and enhance the collection of viewer feedback. To discuss this subject further, we shall examine three basic advertising scenarios: linear advertising, VOD advertising, and advertising in an SDV environment.
Linear advertising: The most established, but least flexible, advertising system is referred to as linear advertising. Linear systems, by definition, insert advertisements on broadcast channels and must adhere to a precise broadcast schedule. Some addressability can be achieved by segmenting the market into geographic zones. However, this is a “blind” operation—there is no return path to measure the effectiveness or interest level of the advertisements on a zone-by-zone basis.
There are ways, however, that zone-based linear systems can support some measure of adaptive advertising. The first of these is client-based linear (CBL) advertising. CBL advertising relies on the client or set-top box to select an appropriate advertisement from a set of ads that are either stored in the client or available, in real time, on another channel. Since storage is generally not available on all clients, the alternative real-time-tuning method is attractive. However, tuning from the broadcast channel to an ad channel and tuning back again cleanly and on a precise schedule is a challenge for many clients.
A second way to support advanced advertising in a linear environment is to directly solicit input with interactive ads. Although these ads must maintain their schedule (that is, they cannot pause for viewer input), viewers can either request follow-up information (such as by mail or email) or may be offered the option to dynamically tune to a separate unicast channel containing additional information.
VOD advertising: Traditional VOD is based on a digital network’s ability to pointcast streams in response to individual requests and allows the targeting of ads based on individual subscriber preferences. The targeted delivery and frequent use of a return path presents an ideal opportunity for advanced advertising. However, many variations are possible, and the optimal systems for maximizing advertising impact are not yet known.
One issue to be resolved with VOD content is the location of avails. Because VOD does not have to adhere to a timeline, VOD content can be preceded or followed by “bookend” or “bumper” ads. This has become very common on the Internet and is “forced” on the viewer (that is, fast-forward is disabled). The example we present in the following section uses a different approach by allowing trick-play operations by tracking these viewer actions and using them strategically to adjust future ad placement decisions.
Placing advertisements as “interior” or “interstitial” ads in VOD is complicated by two factors, one business related and the other technical. The first involves the rights to interrupt content for an advertisement. These rights are not generally granted by content owners. However, some content offered as VOD, such as sports events or TV series, have known avails that can be (or need to be) replaced.
The technical issue to be resolved to support interior ads is the marking of avails. Either in-band markers can be used such as SCTE 35 cue tones, or out-of-band metadata can be delivered to point to specific locations.
Advertising in SDV: SDV also leverages the presence of a return path, allowing downstream capacity to be tailored to user channel selection preferences in real time.
The dynamics introduced by SDV, however, complicate ad placement. This is because SDV streams change from a “unicast state” when exactly one viewer is tuned to the channel to a “multicast state” when multiple viewers are tuned to a channel. Each state presents addressable advertising possibilities. When a channel is in a multicast state, ads can be selected based on service group (geography) and content. When a channel is in a unicast state, ad placement can also depend on the viewer. Note that this introduces a complication when a channel is transmitting a viewer-specific ad and a second viewer tunes to this channel.
The choices here are to either allow the second viewer to view the advertisement or to “fork” the channel into two unicast variations. If capacity is available, this has no negative impact and, in fact, allows future targeting to each viewer. However, if capacity becomes scarce, it would be useful to allow a “rejoin” of the two channels; this would necessitate a retune of at least one of the clients.
Optimizing capacity usage and ad revenue by dynamically dividing channels into broadcast, unicast and multicast streams based on a viewer tuning requests is an open technological challenge, but one that must be addressed in the coming years. The headend component at this center of this activity is sometimes called the session resource manager (SRM).
In a typical VOD system, the SRM receives session setup requests and secures the necessary quadrature amplitude modulation (QAM) signal capacity and resources from video servers, downstream multiplexers and other components. The SRM is in a position to recognize and characterize advertising avails; yet the SRM is not aware of advertising strategies or equipped to make targeted advertising decisions. Therefore, the current architectures to support VOD advertising are based on communication between an ad manager (ADM) (equivalent to or closely related to the SRM) and one or more ad decision servers (ADSs). The example in the next section is one such system.
Advertisement decisions can be based on the content selected and/or knowledge about specific viewers or populations. On the Internet, these two types of addressability are referred to as contextual-targeting and behavioral-targeting respectively. There are variations of each. For example, contextual-targeting might be dependent only on the channel or VOD category, while behavioral-targeting might be based on population groupings and aggregate statistics. Targeting of this type can be highly effective and is enabled by a return path providing detailed data on each and every interaction between each individual subscriber and each advertisement. Next steps It is likely that many experiments will be attempted in the broad area of adaptive advertising over the next few years. This will necessitate flexible architectures and active standards activities.
Further variations on adaptive advertising include:
• Allowing the user to launch nested sessions to obtain further information. This is called “telescoping”; it provides a direct feedback loop and is reminiscent of the browse dynamics of Internet surfing.
• Simply tracking subscriber trick-play operations to give an indication of user interest. An example of this type of feedback is described in greater detail in the next section.
Addressability and interactivity are the key leverage points for change in cable advertising. We will therefore interject an example before discussing additional topics. Addressable scenario The VOD advertising architecture shown in Figure 1 supports dynamic addressability. The key components of the system are:
• An ADM capable of managing sessions composed of advertisements and on-demand content. The ADM can easily modify which ads are shown in which avails and can change these associations dynamically as specified by the ADS. The ADM is also capable of collecting detailed session data and reporting this data to the ADS.
• A (possibly remote) ADS that communicates with the ADM through an early variation of a SCTE/DVS/WG5 draft protocol. The ADS is responsible for all ad selections and placement decisions.
The dynamics of this system can be grouped into three stages as shown in Figure 1. The steps labeled A1, A2 and A3 are initialization steps that occur before sessions are launched. Content (both advertisements and entertainment content) must be loaded onto video server inventories and must be communicated between ad sales organizations and the delivery system. In addition, the ADS must share initialization data with the ADM.
Session setup steps are labeled S1 through S5. These steps very closely resemble the classic VOD session-setup steps managed by an SRM. In this extended message flow, the SRM either references pre-loaded ad bindings or communicates directly with an ADS to access a set of ads to precede the viewer’s request.
The exchange labeled P1 denotes the periodic reporting of viewer data from the ADM to the ADS(s). This data includes trick-play statistics for all advertisements and can be used by the ADS(s) to intelligently select future ads.
The three principal components shown in Figure 1 closely parallel the current components seen in the field of Internet advertising. In the Internet architecture, the ad management infrastructure communicates with a Web site (as opposed to a headend), which in turn supports a browser. The interactions labeled A1, A2, A3 and P1 are directly analogous. The session setup steps S1-S5 serve the same function in cable as content delivery on the Internet.
Note that the ADS must be aware of the set of avails and advertisements available to it at all times (communication labeled A2 in Figure 1). In the early implementations, these sets can be static and downloaded manually into the ADS and VOD server. However, as these systems become more sophisticated, market dynamics will require standard application programming interfaces (APIs) for coordinating the changing sets of avails and advertisements.
In this system, subscribers have full trick-play control of the content, including the advertisements. Thus, subscribers can fast-forward through ads or rewind to watch an ad again. Each of these user actions, however, is noted, and the interaction can be measured for each advertisement. As on the Internet, these interaction metrics can then be used to adjust future ad selections for the individual.
This architecture can support advertising placement ranging from basic to advanced:
• Initial systems place bumper ads before or after user-selected content. Bumper ad selection can be entirely content-based or, with the right system capabilities, can change rapidly based on knowledge of the actual subscriber at session setup time.
• Interstitial ads are a possible future addition to this architecture. It is not clear, however, whether interstitials will be a welcome or useful advertising tool for movie content. Interstitials will be needed, however, for other types of content such as sports events.
• There are a variety of ideas and open questions for handling pause, resume and trick-play operations. When resuming a session, subscribers can be shown the same ads or updated ads. These may be displayed at the resume point or only at the original avail points. Other policy decisions include behavior after a rewind and fast-forward; for example, should the system show the same ads or replace the originals? Inventory As discussed in the previous section, there are many ways to organize addressability and two-way control traffic. As leverage points are established that show signs of a solid return on investment, a dynamic supply and demand scenario is likely to emerge. In the previous example, we encapsulated this activity in the campaign manager. However, the value of ad inventory and the overall strategic creation and pricing of ads is supported by a network of advertisers, content providers and distributors.
Thus, the ADS servers, which are a source of ad decisions for the ADM, will themselves be dependent on this campaign infrastructure.
Again, we can look to the Internet as an example of the kind of market dynamics that are possible with an effective feedback loop. Effective and near-immediate feedback on the Internet allows campaigns to be scrutinized, altered and optimized quickly, resulting in advertisements whose effectiveness increases throughout execution of the campaign. This results in a renegotiation of rates between the buyers and sellers of inventory.
Standards bodies may play a significant role in this area. Automated measuring, modifying and re-pricing of ads can all benefit from uniform data formats and APIs. Enhanced content The adaptive advertising architecture described previously is designed to be flexible in terms of the selection and placement of ads. However, it does not address the changing visual composition of TV ads.
Digital delivery of compressed video and audio also allows the delivery of supporting data streams. As envisioned when MPEG-2 was standardized, these streams can be used to send enhancements to the standard video display.
The DVB MHP and OpenCable Application Platform (OCAP) efforts have focused on new client mechanisms for enhanced ads. The OpenCable initiative has also defined the enhanced TV (ETV) specifications for use on existing, or “legacy,” boxes.
There is still a fundamental question of whether viewers will actively engage when presented with interactive options. By comparison, the Internet is based on user interaction and regarded as a “lean forward” medium. Entertainment TV may require more subtle invitations to engage with the programming. Although it is unlikely that television will grow into a lean-forward or browse-based medium, it is very likely that the Internet toolset will be used to support new usage models for entertainment TV. Standardization Given the range of possibilities for building advanced advertising systems, there is an industry need for flexible architectures, data formats and component interfaces.
The SCTE Digital Video Subcommittee has supported standardization in the area of DPI since 1998. Figure 2 is a high-level block diagram showing the data and control flows for advanced advertising systems. The subcommittee’s initial work centered on the precise structure needed at the MPEG layer to ensure clean splices (labeled A). This was followed closely by a standard (SCTE 35) to identify “splice points” (that is, avail boundaries) in a network feed (B). A realization followed, however, that digital splicer vendors were emerging separately from ad server vendors; thus, a standard API between a generic video server and splicer would ensure interoperability while allowing competition among splicer and server vendors (C). With these standards in place, a first wave of digital ad insertion products has emerged for inserting local ads in digital broadcast feeds.
To accommodate the advertising directions described in this article, new work has begun in a number of key areas. These include:
• Communication between the ADM and ADS components and between the ADS and campaign management systems to support dynamic ad decisions and the distribution
• The standardization of metadata in the network feed that identifies the content (called program identification fields). This metadata can be monitored so that ad placement can rapidly react to schedule changes. For example, a sports event might run beyond its scheduled time; with program ID metadata in place, additional high-priced sports ads can still be placed.
• Standards and technologies to support client-based advertising.
Future work may address data formats and APIs for managing the dynamic and automated pricing decisions as well as the ad composition changes discussed earlier. Moving ahead Over the next few years, cable advertising will be able to reach out and engage with consumers in a more direct and strategic manner than ever before. This is necessitated by the viewers’ increasing ability to view entertainment content from a wide variety of choices on their own schedule. Thus, adaptive advertising is a way to stay engaged with the consumer with messaging and information that is meaningful in a dynamic, information-rich culture.
Considerable experimentation, in the form of pilots and trials, will be necessary to create optimal advertising systems. However, much of the invention should be on ad selection, campaign organization, setting inventory values and other higher order concepts. The basic language of avails, cue tones, content metadata and API infrastructures can and should be standardized to whatever extent is possible without compromising innovation. Guy Cherry is principal architect, video systems, at C-COR. Reach him at firstname.lastname@example.org. This article was adapted from a paper published in the IBC 2006 Conference Publication.