Companies started using artificial intelligence and machine learning about five to seven years ago, but those early efforts weren’t targeted nearly enough.
That is finally starting to change as marketers are turning to the technology to solve very specific issues, like refining their customer retention efforts, targeting competitor’s customers, or creating profiles of their ideal prospects or customers.
Wilson Raj, global director of customer intelligence at SAS, says the technology can help marketers do the following:
- Refine segmentation for better personalization.
- Enable timelier and more relevant customer experiences by recognizing past patterns, current engagements, and predicted behaviors and then surface in-moment offers based on those insights.
- Boost revenue through next-best-action recommendations. Machine learning can help spot patterns or changes in customer behavior more swiftly, enabling marketing to respond in real time by adjusting offers.
The first step in using AI/ML for competitive marketing is to understand what the technology can and cannot do today and how it is evolving, says Christian Wettre, general manager of Sugar Sell and Sugar Market at SugarCRM.
AI FOR MARKETING IS STILL IN ITS EARLY STAGES
The use of the technology for competitive marketing is in the early stages, according to Wettre. But as more companies have success with it, the pace of adoption will quicken. So Wettre and others expect the technology to penetrate the mass market in the next couple of years.
Though the terms are often used interchangeably, there is a distinction to be made between artificial intelligence and machine learning. ML is an advanced subset of artificial intelligence, enabling CRM systems to learn to find insights without being told exactly what to look for, Raj explains.
Rohan Chandran, chief product officer at Data Axle (formerly Infogroup), agrees. While extremely basic AI has been around for some time, ML and deep learning have become industry buzzwords only in the past few years. AI performs the “grunt work,” like triggering email campaigns. ML drives more advanced use of AI, such as lead qualification scoring.
“You have this training data system and the feedback loops that come in from what actually happens as you use the data; then that system recursively learns and evolves and gets better and better,” Chandran explains.
Before deploying machine learning, determine if it will add value to the process, Chandran adds.
Sugar has been careful about how it approached AI and ML, Wettre says. “We’ve taken a walk-first approach. As we’ve applied the science to the marketing and CRM universe, everything has to do with the ideal customer profile.”
Once the ideal customer is identified, the company uses that knowledge to attempt to attract and convert prospects to customers in a more efficient manner, Wettre continues.
SugarCRM and other companies use AI/ML to continually refine their respective ideal customer profiles (ICPs), according to Wettre. “Building an ICP model is often done on a customer data platform.”
In a traditional scenario, companies will wait until prospects or customers fill out information on web pages before they react, Wettre explains. But with ICPs created with AI/ML, they can move forward with very little information, perhaps even just a web view, and combine that with information from other sources to score customers or prospects on how similar they are to the ideal profile and use that to determine marketing and sales efforts.
Raj says the technology can also help marketing with offer and click optimization on the web or mobile apps. It can, for example, dynamically tailor web content based on visitors’ past search history or website or mobile app interactions. It can also help forecast potential profitability, finding patterns in past behavior to predict lifetime value of prospects or customers at the beginning of their life cycles. That can then be used for improving resource allocation and campaign management and calculating the ROI of marketing investment.
AI UNCOVERS SOCIAL MEDIA INSIGHT FOR MARKETERS
Chandran points to social media commentary as a way for companies to learn which of their competitors’ customers might be prepared to make a switch, adding that AI- and ML-powered sentiment analysis can analyze the commentary to determine the most dissatisfied customers who would be the best targets for marketing outreach.
Using AI and ML enables marketers to not just scour social media feedback but also information from other digital touchpoints and in-store visits to help determine customers’ emotional attitudes, Raj says. ML can analyze “all that stuff at scale” and then “go beyond traditional segmentation approaches to almost give you a fuller view of that consumer.”
“The airline industry is a great example of this in action,” Chandran says, noting that people take to Twitter quickly to complain about very specific aspects of their travel experience.
If a customer of one airline complains on Twitter about an experience with that airline, a competitor can use the information to offer a discount on a future flight between the same airports. Longer term, airlines can use customer social media information to determine how to attack competitors’ weaknesses, Chandran says. “This is where you can highlight not just what your own unique positioning is but specifically what customers are reacting and responding to and tailor your marketing to that.”
Similarly, an airline can better gauge what competitors are doing right and determine if they want to duplicate those efforts.
But it’s not just the travel industry that stands to benefit. In the fast-food industry, Wendy’s monitors the social media and more traditional marketing of Burger King and McDonald’s to help it in its social media marketing campaigns, according to Chandran.
Successful use of AI and ML in marketing comes down to having comprehensive data, Wettre says. “The more behavioral patterns you have about a customer demographically—how the customer has behaved and reacted over time previously—the richer you can create those models and the smarter your AI models are going to be.”
Companies that perform best can collect as much data as possible and then use their own data scientists or technology to make sense of it, according to Wettre. “Understanding the data science isn’t easy. It’s hard to apply it correctly. It’s hard to get statistically meaningful information out of these models.”
Combining competitive intelligence with “look-alike” modeling to determine which prospects are most like current customers, marketers can better target the prospects most likely to convert to customers, Raj adds.
“You can get much more accuracy if you know that this person or unknown [website] visitor is acting very much like a current customer with these same attributes,” he explains. “Now we can treat this unknown user like a known user and dynamically offer content or interactions in a more powerful way. If the unknown user responds, then you get deeper; if the user doesn’t, then try another look-alike model. In the past, we just served up some general offers, but now we can get crisper in terms of the content, media, kinds of products or services, and pricing based on that look-alike model.”
AI and ML enables companies to take this approach on a micro-segmentation level, being extremely precise in the types of content, offers, or interactions they offer prospects, Raj adds. A company can offer a young prospect in the Northeast a very different offer from an older prospect in a different area of the country or even for another young prospect in a nearby state or nearby city.
Raj identifies McCormick, the spice company, as one firm that has done an excellent job with this, developing different content for several hundred flavor profiles that it can serve up to prospects visiting its website.
“It’s like a fingerprint for your profile,” Raj says. “You’re using machine learning and hyper-segmentation for the next best actions and recommendations.”
While large companies like airlines or major fast-food vendors have the resources to invest in these tools, smaller companies have unique challenges: They need AI and ML solutions that are affordable but also easy to transition to or work with their legacy marketing systems, Chandran says. “That is when it will hit the mass market.”
But even some smaller enterprises have already had success with AI and ML technology, Raj says. He points to Raiffeisenbank in Belgrade, Serbia, which used the technology in a successful campaign for a credit product. Using machine learning combined with historical customer data, risk scores, and details on timely bill payments to third parties, the bank greatly refined the customers it targeted for a credit offer, generating a 14 percent success rate, compared to just a 1 percent success rate for previous campaigns.
AI AND ML FOR MARKETING HAVE SOME LIMITATIONS
Sometimes companies expect too much from AI and ML, Wettre says. “The worst practice is when someone falls in love with the idea of a computer answering the questions—that you can turn [AI/ML] loose and it will tell you what to do. It’s not that easy. If you don’t back it up with other investments or with the right vendor, you’re not going to get the results you are looking for,” he suggests.
Chandran adds further that AI and ML solutions aren’t one size fits all. A solution that works for Data Axle likely won’t be suitable for a much smaller company in a different industry, he says.
To maximize the benefits from AI and ML, companies need to continually train the technology, and then monitor it to ensure it is learning as expected, Chandran says, pointing to the Tay Microsoft chatbot, which some Twitter users “attacked” shortly after it launched in 2016. As it learned from previous utterances, the chatbot was soon swearing and spewing racist terminology and eventually had to be shut down.
Wettre recommends starting small, adjusting the use of the technology until successes start occurring, then expanding from there. SugarCRM has followed that concept on its own use of the technology.
Some of SugarCRM’s customers hoped for a quick win with the technology but have learned to use a more structured, step-by-step approach to expanding use of AI and ML.
“Focus on who is a win-win,” Wettre says. While the initial win might be for the company, if the customer sees the relationship as a win, the customer will continue to return.
“Make sure you have the right resources,” Wettre adds. “If you’re going to do generalized AI, you have to invest in very, very skilled people. It’s not a trivial thing to do. These are expensive employees. And it’s a fairly costly thing to do.”
“One of the things that people forget for machine learning is there has to be an objective,” Raj says. “For example, if I want to be able to score [profiles] so that I can acquire these kinds of customers, the first thing I need is to establish a goal for what I want to do with machine learning.”
And then, too, keep in mind that while the technology is excellent at automating tasks and offering predictive modeling, even with sentiment analysis, it still falls short in terms of understanding emotions, Raj adds.
AI AND ML WILL EVOLVE WITH MORE DATA POINTS
However, experts agree that whatever limitations AI and ML have now, the technologies will continue to improve as they gather more data points. AI and ML will become even more important as the COVID-19 pandemic wanes and companies look to rebuild their client bases and reacquire wayward customers, according to Chandran.
“By 2022, hopefully, micro-segmentation will be mainstream,” Raj says. “Beyond just personalization, I can see machine learning helping with other complex activities. For example, making the necessary budget adjustments in real time in broad campaigns or more specific campaigns; maybe doing a quick ROI analysis on campaign results, and then authorizing changes in resourcing and planning in real time.”
Raj adds that those changes would be based on shifts in demand using prospect and customer data, behavioral data, and information from suppliers.
Phillip Britt is a freelance writer based in the Chicago area. He can be reached at email@example.com.
This article is written by Phillip Britt and originally published here