Companions to Marketing & Customer Segmentation
All customers are now well-aware of the fact that we follow them on their journey through our websites. We want to see what they did; we want to look at their choices when faced with a decision branch; what attracted them; what did they avoid; what convinced them to make a purchase; and, more importantly, what caused them to leave.
Up until now, we have used KPIs (Key Performance Indicators) to help us figure that out. One of the oldest, yet still most popular, is the NPS (Net Promoter Score) system developed by Fred Reichheld, Bain & Co., along with the firm Satmetrix Systems ‘way back in 2003. Do your customers “promote” you to their friends and associates?
NPS is designed as a “two-part, single question” survey system that is often presented similarly to these example images. The user is asked to rate their experience as they are departing the site, or after a financial transaction.
They are also asked for a typed comment, which can be a useful aid for research. Always give them a Close button (⌧) of some description, so they don’t feel trapped by your survey.
Is it still useful?
Once you collect sufficient responses, people that scored 1-6 are totaled as a percentage, all the 7-8 people are ignored, and all the 9-10 people are calculated as a percentage.
If 10% were in the red zone. 30% in the yellow zone and 60% were in the green zone; we perform the following calculation:
60%-10%= NPS of +50. We ignore the yellow 30% since they don’t affect public perception. Yellow people neither actively support nor detract from your reputation. As you can see the total score can range from -100 to +100, and +50 is looked upon as a very good score.
This feedback is potentially very useful, but it is a “reactive system.” If the customer is not specific in the comment, using something like this requires you to “guess” where things went wrong. You can “chase” the problem forever without ever solving it.
Artificial Intelligence (AI) & Machine Learning (ML)
AI and ML are changing how we follow our customers through our sites, and these techniques are increasing their engagement. As sophistication improves, an AI (using ML) will be able to detect a long hesitation on the part of a visitor. It will be able to recognize (from prior observations) that customers who hesitate at one particular point often leave the site, or occasionally, make a limited number of selections, some of which lead to successful conclusions (ordering, signing up, etc.), so now the AI can take appropriate action.
Recognizing the situation, the AI can immediately fire off a pop-up which might say “Are you looking for [link1], [link2], or [link3]? Just click to be taken there immediately”. And now the potentially frustrated customer has an immediate solution and doesn’t close the page.
The Friendly Helper
Alternatively, an AI could steer them to an onsite ChatBot. Generally text-based, they can answer specific questions (“Do you have this sweater in size 4, and does it come in teal blue?”) before directing them more comprehensively to whatever resource they are looking for.
Customers voted overwhelmingly in favor of available ChatBots because of their instant responses. Waiting for “a human” isn’t a primary goal anymore; being “on hold” is a vestige of the past, and completely unacceptable. The biggest driver nowadays is obtaining the information required as fast as possible.
As an example, consider the local Electrical Utility, responding to a widespread power outage. People call, worried that it will be a long event that will cost them hundreds of dollars in thawed, ruined food, or non-functional furnaces in wintertime, or broken air conditioners in the middle of a heatwave.
Customers want fast answers, but even if that business has a PBX (Private Branch Exchange) so they can answer hundreds of calls simultaneously, they only have a fixed number of customer service representatives. People end up “on hold,” and the CSRs seldom have adequate information to provide immediately. Long waits and unsatisfactory answers are a terrible combination.
AI Answering the Call
A ChatBot, on the other hand, can answer all the calls at once, conducting hundreds of simultaneous chats. We’re now seeing the arrival of systems that can listen to words, understand them, and respond in an audible conversational voice, although they are still rare. Most of us are content with online text-based ChatBots that don’t care how many times you’ve called for an update, and always have the very latest information.
AI & ML can now read a human face to determine emotional state with about 80% accuracy (and getting better). It can do the same for voices, interpreting anger, fear, desperation, happiness, and so much more. They’re somewhat more limited in text-chat, focused on word choices—it’s a work-in-progress. Nevertheless, customers are now thoroughly delighted with the “almost read my mind” intuitiveness, which is just Machine Learning in action.
Slight variations in customer behavior trigger the AI to present answers that most closely suit the individual. It’s like a personal assistant helping them reach their desired goals.
Some people still don’t like the idea, however. While the tinfoil hat crowd may not usually form a part of your customer base, it is still a good idea to present the concept to your customers as something genial, innocuous, and friendly.
If you are ever having trouble locating something, please feel free to use our Electronic Concierge by pressing the floating button located on the right side of the screen. He can help you to find anything you are looking for, or answer any questions you may have.
Understanding the Customer
Upselling is a fundamental part of the retailer’s toolkit, but even the best salesperson cannot know every item in stock, or what is available to order. AIs can understand all of these things and know that part “X” goes with the equipment “Y” that the customer bought from you last year.
Not only does everybody have a public profile on the internet, but it is also freely available for you to use. Anything in the public database is fair-game, and adding it to your Big Data collection for your AI to peruse at need, means you are more in touch with both current and future customers’ needs. You can serve them better when you have even a superficial understanding of their lives.
Maybe they’re campers & hikers, which present many selling opportunities (lightweight cookware, solar power chargers, etc.), or they rent a new car every three years, or their home & car insurance is expiring soon. Maybe their neighborhood has elevated crime, and they need an alarm system. Perhaps their Facebook page reveals that they are new parents or grandparents, opening a whole new area of potential sales.
How does an AI Learn?
Using what is called a Neural Network (NN), an AI can emulate human thought. It can automatically collect facts and correlate them. Granted, an AI “thinks” more slowly than an intuitive human, but it can consider thousands (or millions) of ideas simultaneously and reach smart decisions faster than a person can. Yes, they are nowhere near as smart as us, but they are blindingly fast, and we can use that to our advantage.
Thus a customer can ask “Will that 8-inch sieve work with my 12-inch Lagostina Stew pot?” and receive the correct answer from an AI in just almost instantly. A human might need to physically measure such a thing to know if it is appropriate, but in milliseconds an AI could find an image of the pot and superimpose the sieve, determining that it would not fall inside, and was suitable for the task.
Medical Imaging is now often analyzed by an AI/ML system because it can spot subtleties that might go unseen by a human. X-rays, of course, but also MRIs, PET, and CAT scans, Doppler, Ultrasounds, or anything you can imagine. Machines are naturally better than human eyes in finding relationships.
Research shows that when customers tell a friend “Oh if you need X, you need to go to OUR-Company.com” that it almost inevitably means a sale. People trust their friends with real-world experience much more than they will ever trust advertising. And, nowadays, those customers have a much broader influence than ever before, from posts on social media, all the way up to personal blogs with dozens, thousands, or even millions of followers.
There is nothing intrinsically wrong with using NPS as part of your data-building. You may be accustomed to it, and with interpreting its results. Knowing that your customers are happy with you is like money in the bank, even if it doesn’t provide a lot of actionable feedback.
However, the truth is that NPS is a 15-year-old technology in a fast-changing world. AI provides so much more information, and can even self-suggest ways that you can use it effectively. And it continually builds on itself, making better information available constantly.
Maybe it’s time to switch to pure AI/ML and be proactive instead of reactive. Are you ready for the Future? We are!
Please feel free to start a conversation with us here. Our plug and play platform is the ideal tool for end-to-end implementation of AI in your business. Give us a call today…we’d love to hear from you!