Imagine if a retailer could promote a product as personal as perfume and match it to your preferences before you've ever smelled it.
It's a tantalizing possibility for retailers, and it could be reality in the coming decades. Computer science experts believe they are edging ever closer to the day when these kinds of marketing techniques can be used to attract new consumers.
Marketers and advertisers are developing artificial intelligence applications that can home in on a consumer's desires, determine which sales pitches work best and expand a company's market base using customized, targeted ads.
But they are a long way from capitalizing on AI-based applications. Some may be wary of technology they don't understand, while others fail to turn rich results from data-mining programs into marketing strategies. At the same time, many AI applications are a long way from being implemented at the retail level.
Artificial intelligence, the field of computer science dedicated to making "intelligent" machines, has long been used by Internet companies to make search engines increasingly accurate. The military uses it to conduct surveillance, and financial institutions use it to detect fraud. Computers use complex algorithms, or detailed lists of instructions, to make decisions or come to conclusions that give the impression they are thinking or learning.
The marketing applications of AI hinge largely on one of the industry's most basic tenets: know your customer. By focusing on customer profiles, preferences, desires and dislikes, marketers can attempt to entice them.
The problem is that using traditional methods, customer profiles are often woefully incomplete or difficult to turn into strategies that actually work. For instance, consumers may be forgetful or misleading when recalling which items they buy most often at the grocery store. Data may tell retailers which consumers bought items during a recent sale, but the store has no way of knowing why they responded to that promotion and failed to show up at the last one.
With AI, however, those problems are becoming less onerous. Powerful computing systems can collect information about customers and their habits through a loyalty card system, for instance. Algorithms can identify patterns or trends in buying behaviour that can help managers determine which products might appeal to which customers, or when an in-store promotion may be most effective. A drugstore could use AI applications to determine which products people living in downtown Vancouver purchase most often and on what day, and create a tailored promotion in hopes of boosting sales on those days.
But it's not a perfect science. Many consumers may simply ignore or opt out of receiving e-mail advertisements from their loyalty card program, for instance.
AI logic can also be used to target certain customers. A person who does a lot of Web searches regarding European travel may receive ads from travel companies. That type of system is already used by Google to generate income.
Experts predict such targeted ads will become increasingly sophisticated. Imagine, for instance, walking past a clothing store and receiving an alert on your smart phone that a brand of jeans you have purchased in the past will be available to you at a discount if you walk in and buy another pair right now. The alert would then guide you to the product in the store.
"Things like that are entirely possible," said Stan Matwin, computer science professor and a director at the School of Information Technology and Engineering at the University of Ottawa.
Another innovative development is what's known as customer segmentation. Traditionally, retailers might divide their customers into distinct but very broad groups such as working mothers or single businessmen. This approach overlooks all nuance, said Monica Casabayo, associate professor of marketing at ESADE Business School in Barcelona.
Now, complex AI applications are being created to catch the details.
This is accomplished using "fuzzy logic," which Prof. Casabayo described as categorizing people in terms of "greys" instead of black and white. It moves away from statistics and instead uses massive amounts of customer information gathered through point-of-sale data, loyalty card demographic information, customer surveys and social media activity. It determines the degrees to which consumers feel loyal to a particular brand or are fixated on low-cost shopping.
"It's trying to classify your customers, to segment your customers, according to their real behaviour," Prof. Casabayo said. "We need some techniques that help understand the market at this point.
Four categories where artificial intelligence can have a significant impact on marketing practices, according to Monica Casabayo, a professor at ESADE Business School in Barcelona:
- Product innovation: Companies can use AI-based algorithms to generate variations on existing products - a new shade of paint, for instance. Prof. Casabayo said this technology is still in an exploratory phase.
- Recommendations:Amazon and Netflix increase sales by using AI applications that recommend additional products based on a consumer's purchase history. Experts predict these recommendations will become much more sophisticated and accurate.
- Segmentation: Customer profiles often lack details about a person's interests or background. New systems will help companies avoid pigeonholing consumers and instead focus on the likelihood they will respond to promotions or switch to another brand.
- Customer service: More companies are using AI-based applications to provide customer service online. These programs may answer consumer questions, provide technical support and solve other issues in a quicker, more efficient way than humans could.