People have been toying with the concept of artificial intelligence (AI) since John McCarthy first coined the term in 1956. AI can be defined as the theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.ⁱ In the digital signage industry, we are seeing a split between two categories of solutions: the low cost systems that provide simple, generic (low relevance) messaging to a wide number of locations, and the intelligent systems that provide unique (or high relevance) messages to targeted audiences based on their current situation. When you categorize the various digital signage solutions, they tend to fall at either end of the spectrum. Both models have a place in the digital signage ecosystem, but their application will depend on the business model for the network and should be justified on a total cost of ownership basis.

In the late 1970's and early 1980's, dumb terminal or thin client computing was a popular idea. In this model, the computing was done by a central computer and the terminal was simply an input/output device. Although it had the advantage of lower cost hardware, dumb terminals greatly restricted user productivity and had limited success. Today, thin client computing is having resurgence because modern CPUs are significantly more powerful than personal computers were a decade ago. The computational capabilities of modern thin clients make them a viable option for most desktop applications, such as word processing and spreadsheets, but the requirements for digital signage are quite different.

Digital signage is a graphical medium and modern displays are high definition screens that require more powerful graphic capabilities. Pixilated images and jerky motion is not acceptable to most people and can reflect poorly on the product presented on screen. If your content includes a lot of high resolution video or animation files, they will require more processing power for decoding and playback. Higher resolution screens raise the bar for the hardware handling the content, and good content is resource intensive. A powerful CPU is important but sophisticated software also takes advantage of the computer's graphical processing unit (GPU) and to date, that is a weakness in most thin client architectures. The cost difference between thin clients and ultra thin desktops is not significant and changes in the next generation of hardware and operating systems will provide much greater graphic capabilities in low cost hardware configurations. Today, the variation in hardware costs is no longer as significant when looking at dumb versus intelligent networks; the difference has more to do with the sophistication of the software.

Computers and software provide us with the tools to build smart systems that can respond to stimuli and deliver an appropriate message based on a set of rules or logic. Multiple rules can be applied to any given situation to determine the best fit between an audience and a message. Alternatively, these rules may adjust parameters within the message based on feedback from the audience. By manipulating variables, these rules can allow each display to determine the appropriate message to play at a given time to a certain audience. This model depends on smart systems that make local content decisions based on a combination of information. The local decision could change the content or individual components within that content, such as the relative size and location of an advertisement on a screen.

We can categorize inputs or variables into three areas: descriptive, circumstantial and locality. Descriptive variables are terms that characterize a fixed location or a piece of content (also known as metadata). A circumstantial variable is typically information about the viewer. This could come from facial recognition software (using a web camera), an action (touching a screen) or a peripheral (loyalty card or barcode reader). Locality information looks at variables within proximity of the screen (temperature, local inventory, GPS coordinates). Looking at combinations of variables, one can perceive the local environment and take action to display the content that maximizes the chances of a desired reaction. Adding feedback loops, such as interactivity or sales at the register, allows the system or operator to modify these rules to improve results (in AI terms, this would be analogous to a learned response). PCs allow us to place intelligence at each display while centralizing the management process. This concept of distributed intelligence is more expensive to deploy than a simple broadcast model, however, it offers more targeted messaging (which means that it is more effective) at significantly lower operating costs because less manual intervention is required per unique impression.

The idea of an intelligent player is not new, but its application has been restricted due to hardware costs and operating system limitations discussed earlier. Operating costs and flexibility are becoming more significant as networks reach critical mass and can justify investing in more sophisticated solutions.

There are other instances where a "smart player" makes sense. In an application such as wayfinding, a patron often interacts with a touch screen to find the best route to a desired destination from their current location. Typically, the digital signage system would select a map or animated image with a line indicating the best path between these points. The system would use a series of predetermined routes and select the one that matched the two end points, oblivious to current conditions; but what if the shortest path is not the best route at this time due to some condition such as congestion, lighting (avoid dark alleys at night) or maintenance (escalator temporarily out of service for maintenance)? The system should have the ability to make decisions based on inputs to determine the best path upon looking at all possible routes. If there is a fire in the building, we want to send people to the nearest exit moving away from the fire, but advancing to the predetermined exit might put you in harm’s way. If the alarm system knows where the fire is, then that information should tell the wayfinding system which routes to avoid. An emergency response system is a good example of where wayfinding technology needs to respond to current conditions and those conditions will vary at every location. One message does not fit all in this case and adding local logic makes a complex problem much easier to answer.

At first glance, digital signage at a quick service restaurant (QSR) in a sports stadium would appear to be a fairly simple application. The menu is limited, the target audience is known and the duration is short, yet it is a much more complex solution than most people would suspect. Alcoholic beverages are a popular and profitable product to sell at stadiums, however, many stadiums by law must stop selling alcohol prior to game end. In basketball, the beer ads and all related menu items are removed just before the 4th quarter starts. When the home team scores 100 points, some food outlets offer free pizza as part of a team promotion.

Similarly, when the team's top rookie player reaches a scoring threshold, promoting his sweater at selected concessions will ultimately drive more sales of that item.

A taxi service using the digital signage network at the stadium to promote their services can get much higher value for their advertising dollar by only advertising during the 4th quarter. By linking the timing of their ads to the game clock, they ensure that they are reaching their target audience when their taxi services are needed. By integrating the game clock and score into the digital signage, you can match the content to the environment providing much greater relevance.

There are a variety of concession stands offering different menus and merchandise on their digital signage, with all of the information being managed from a single point. You could devote a full time person to do this manually if you only have one type of concession but a stadium has a broad mix, so effectively doing this requires a large staff for short periods of time. Content can be tagged through metadata; these are addendums that describe assets. Metadata can also be assigned to locations, schedules, and playlists. A smart system can continuously compare metadata tags using logic to determine what should play where and when. Manually managing playlists is expensive and impractical. Targeted ads pay for the system and the more focused the ad, the more impactful for the advertiser and more profitable for the network operator.

In advertising based models, the cost of running an ad typically varies by time of day and location. Many digital signage systems display multiple assets at the same time, so a viewer could be looking at headlines or searching for a location at the same time those ads are presented. Allowing interactive content to trigger an ad or determining the size of an ad based on current conditions can impact the relevance of the ad and value to both the advertiser and the network operator. In a large network, the complexity of doing this centrally makes it impractical. However, if each player is smart enough to make changes on the fly based on local conditions, the experience is much more valuable. Having each player respond differently as opposed to displaying a standard loop played at all locations can increase viewership. If a viewer keeps seeing the same information playing as they walk around the mall, they soon learn to ignore the content since it offers nothing new. In advertising, repetition can establish familiarity for a product or a brand as long as the viewer remains engaged. Digital signage is adjunct to an activity such as shopping, exercise or commuting; it is information, not entertainment. If the content is not relevant it is ignored. That is why time of day, news and weather information is included on many signs.

As we move forward, the divide between low and high end digital signage solutions will increase. Advances in hardware and changes in operating systems will make the appliance versus PC argument redundant. Delivery options will increase as cloud computing becomes more ubiquitous, but most solutions will eventually be able to operate on premise or on the cloud and the delivery mechanism will be a configuration option for the customer. Low cost systems allow small networks to become established but as the network grows, operating costs restrict the network from delivering sophisticated content and growth. Higher priced systems exist because they offer lower operating costs and greater value through improved functionality. When a small network hits the point of critical mass it might be constricted from further growth, which would justify investing in a more sophisticated solution.

Building a network requires a significant investment, and looking at the long term requirement is just as important as the short term. Being able to offer sophisticated capabilities without a lot of manual intervention may help networks grow faster. Most suppliers offer pricing scenarios to allow customers to grow into their systems. Products are constantly evolving, so rejecting a solution because it does not include an item on your checklist can be short sighted.

In the mobile phone market, limited feature phones have higher share but smart phones are where all the action is. In the digital signage market we will see the same trend; the cost of low end systems will continue to drop and providers will need to reduce services in order to live on smaller margins, while smart systems will grow as operators demand a better return on their investment through lower operating costs and greater flexibility.

Jeff Collard is President of Omnivex Corporation (www.omnivex.com), which makes enterprise-wide software to manage all aspects of digital signage networks, including content management, real-time data acquisition and distribution, and remote device management. He can be reached at jcollard@omnivex.com.