To help runners avoid injuries by coaching them to achieve a better running style, Sensoria developed an anklet device attached to a special running sock that users wear while running. Sensors in the bottom of the sock measure where the user’s foot makes contact with the ground and for how long. The anklet contains a CPU that analyzes a rich variety of data from the sensors and relays that data to an accompanying smartphone app. The app displays a detailed heat map of where pressure is being placed on the foot, along with detailed statistics regarding foot contact time, cadence, steps taken, stride length, and speed. This information is shown in the Sensoria app so users can adjust their movements in real time. The app also gives audio alerts in real time with an automated voice that provides feedback about any incorrect foot-strike patterns or other adjustments to prevent injury.
The Automatic Execution Connected Customer Experience
Let’s revisit the printer story one more time. This time, imagine that the printing is going well when the doorbell rings. David is surprised to see a box delivered. He doesn’t recall having ordered anything. Inside the box is a toner cartridge. How odd, he thinks. He resumes printing, and his computer alerts him that his printer is about to run out of toner. Only then does he remember that when he bought the printer, he gave the printer company permission to automatically send more toner as it ran low. David just experienced an automatic execution connected customer experience. Once the firm is authorized to take care of something, the firm automatically gathers information and fulfills the need, often before the customer had realized that the need has arisen. This is shown in figure 4-5. This diagram is almost the opposite of the respond-to-desire flow in figure 4-2. Here, almost all activities are controlled by the firm because the firm knows what the customer needs and when. This is somewhere between Big Brother and loving mother.
Delivering such a connected customer experience can be difficult. Unlike in the case of curated offering, the lack of customer involvement in the decision-making steps of the customer journey makes errors more possible. True, customers who go online and search for a book on baby names are likely to be future parents. But should a retailer send them a crib and diapers based on this information alone? Making automatic execution connected customer experiences work requires increasing the bandwidth of the information flow that moves from customers to firms.
With the increasing connectivity of objects through the Internet of Things, more and more of these automatic relationships based on continuous information flows will become possible. The printer example is real. HP has a program called Instant Ink that works exactly as we described. In this case, HP ships replacement toner to customers once their printers send out a “low ink” signal to the company. Brother has a similar program called Brother Refresh, where customers can decide whether they want the company to send the ink or leave the fulfillment to Amazon.
FIGURE 4-5
The automatic execution connected customer experience
Soon our refrigerators, sensing that the weight of the milk container is very low, will be able to reorder our milk for delivery by tomorrow morning. (Of course, the fridge also checked our calendar to make sure we were not going on vacation starting tomorrow and wouldn’t need the milk.)
Other already-existing examples of automatic execution connected customer experiences are in the realm of medical intervention. Fall-detection sensors are small medical devices for seniors who are at increased risk of falling. Early generations of this technology followed the principle of respond-to-desire. A senior in need of help could press a button on a wearable device, activating an emergency call anywhere in the apartment or house. This was clearly addressing an important unmet need, as previously, a senior who fell down the basement stairs might have needed urgent medical attention but was unable to move and reach the phone. That is why the latest fall-detection devices have switched to an automatic execution relationship. Sensors in the device detect the fall and are able to take action automatically, even without the involvement of the patient. (For another example of merging curated offering with automatic execution, see the sidebar.)
The Information Flow from Customers to the Firm
A central aspect in creating new connected customer experiences is to design the information flow between the customer and your firm. After all, it is this information that allows you to recognize your customer’s needs and to identify the optimal solution. We find it helpful to think about five dimensions to describe this information flow:
The trigger of the information flow, which could be the customer or the firm
The frequency of the information flow (episodic vs. continuous)
The richness or bandwidth of the information flow (low vs. high)
The customer effort associated with this information flow (low vs. high)
The information processing that is required to infer the right product or service solution in response to the customer need (the customer might explicitly express which product or service she wants, or the appropriate product might be inferred by the firm)
AUTOMATIC EXECUTION WITH VIDEO GAMES
Consider the following stunning statistic: if the players of the hugely popular online game World of Warcraft spent their time creating Wikipedias rather than playing the game, they could create a new Wikipedia every week! That’s a remarkable testament to how engaging and “sticky” games have become. Video game companies have come a long way when it comes to curated offering and automatic execution. It used to be that every player would purchase the same version of a game by going into a store and buying a game cartridge or CD. Now, with online gaming, game producers learn about the preferences and skills of individual gamers and create customized experiences that keep a player in a state of flow. Players are exposed to challenges that are neither too difficult (causing frustration) nor too easy (causing boredom), a process called dynamic difficulty adjustment that requires sophisticated artificial intelligence on the back end. Likewise, different players derive pleasure from different aspects of a game. There are achievers (interested in gaining levels and points), explorers (interested in understanding the nuances of the game), socializers (interested in interacting with other players), and killers (the name says it all). By understanding a particular player’s type, the video game can adjust and get the player into more enjoyable situations. No wonder one study of World of Warcraft found that 75 percent of gamers play longer than two hours per day on average, and 25 percent play longer than five hours per day.
You need to make decisions along these five dimensions before you can turn to the technical aspects of building a connected strategy, such as smarter devices or increased communication bandwidth. This is summarized by table 4-1.
At the beginning of the chapter, we introduced the buy-what-we-have customer experience, illustrated by the customer’s painful journey from the late realization of toner shortage, to a painful ordering process, and finally to the reception and usage of the product. This experience was not part of a connected relationship. The episode was triggered by the customer’s realization that the printouts were of poor quality. The only information flow took the form of a customer purchase (and even that took a detour by going through a retailer). And it was left entirely to the customer to decide that purchasing toner for the JetPro 6978 was the right solution for his printing needs.
The respond-to-desire connected customer experience was primarily one of reducing the friction associated with the transaction. This could be accomplished via a simple ordering interface (e.g., one-click shopping), conveniently located sensors (e.g., Amazon Dash buttons), or voice recognition (e.g., Google Assistant). All of this dramatically reduces the customer effort compared with ordering via cumbersome websites, telephone, or mail or making a trip to the store.
While the curated offering connected experience also relies on the customer as the trigger of the transaction, it puts the firm in a more active position of helping the customer figure out a solution to her needs. Central to this is th
e recommendation process. Based on past purchases (the customer spends two hours per day streaming World War I documentaries) or expressed preferences (“I really like Bollywood movies”), Netflix can make recommendations to the customer on how to fulfill her entertainment needs. This recommendation process requires data about the customer, increasing the need for information frequency and information richness. The more the firm knows, the better a recommendation it can make. Technical challenges for this connected relationship include a strong recommendation technology. Moreover, firms need to deal with the demand for privacy by customers, as became obvious when Google was criticized for reading through its customers’ Gmail accounts to improve targeted advertising.
Because overcoming customer inertia is one of the main benefits of the coach behavior connected relationship, we cannot rely on the customer to be the trigger for transactions. Instead, the trigger point is moved to the firm. That has consequences for the other dimensions as well. The firm needs to be receiving information from the customer all the time so that it doesn’t miss the right moment to take action. In this case, the connection to the firm is always on, and customers send information to a firm continuously and quite often autonomously. You could say the firm is automatically “hovering” over the customer. For instance, a customer’s Fitbit is continually gathering information, using the customer’s smartphone as a relay device to automatically send information to a health care provider or the customer’s personal trainer. The technical challenge for this connected relationship lies in enabling cheap and reliable two-way communication between customer and firm, reflecting the fact that communication in this case has to happen 24/7.
Finally, as we move to automatic execution, the firm takes on all responsibility for finding the right solution to the customer’s needs. There is an important difference between a fall sensor that waits for its user to press an alarm button and a device that issues a distress call based on the reading of an accelerometer. The sensor has to be read continuously and ideally transmits information in real time so that even in the rare event that the device is damaged in a fall, help can still be sent without delay. Technically, this raises the bar for the ability to correctly infer the right product or service for the customer based on the automatically transmitted information. Recommending a bad movie or reminding a male patient to make a gynecology appointment might lead to customer annoyance. Shipping a Brother cartridge to the user of an HP printer, in contrast, has more significant costs associated with it, but even these costs are small when compared with the damage done by a failure to call a much-needed ambulance.
ARTIFICIAL INTELLIGENCE AND DEEP LEARNING
As we have seen in our discussion in this chapter, there often is more to connected customer relationships than just providing customers with what they ask for. Curated offering, coach behavior, and automatic execution all rely on joint problem solving between the customer and the firm. In the past, such joint problem-solving behavior was only possible through human experts, oftentimes sales managers who would help the customer to determine which product or service was right for them and when and how to acquire it.
Thanks to advances in artificial intelligence, human skills can now be vastly augmented and more and more processes can be automated. The field of artificial intelligence is concerned with equipping machines with skills that previously could only be mastered by humans, thereby enabling them to not just execute orders but solve problems instead.
It is helpful to distinguish between two ways a computer can help in the customer journey. The first one is based on applying a (potentially large) set of rules, such as “propose paper to everybody who has purchased toner” or “reorder milk whenever the last bottle has been opened.” Such rules can be carefully audited, making sure they lead to desired recommendations, but they have an important drawback. Every aspect of the problem needs to be coded in the form of a rule, which can quickly lead to an explosion of rules (e.g., if customers leave for vacation, a query of the calendar might also be required, so that milk is not reordered in this case).
Coding everything in rules might work for low-complexity situations such as ordering toner or milk. In situations of high complexity, however, it is much harder to codify knowledge in the form of rules (what exactly makes one skin irregularity an indicator of a risk of cancer?). This is where the second way in which a computer can solve a problem comes in. Rather than defining rules for identifying skin cancer, the computer gets fed a very large number of skin images and is informed which are cancerous and which aren’t. Based on patterns the computer finds in these old images (often referred to as the training set), the computer then evaluates future images.
This approach is much more humanlike—after all, we do not tell our children what exactly makes an animal a cat; they just figure it out from observing many animals and listening to our parental classification. (It turns out that determining whether a video includes a cat has been a much-studied problem in computer science and has attracted the attention of some of the greatest minds in that field.) The technical term for this learning approach is training a neural network.
Such digital representations of neural networks have been around for many decades. Recently, a breakthrough in this area has been achieved, referred to as deep learning. Deep learning is inspired by the human brain and organizes the neural network into multiple layers, each layer using different levels of abstraction. For example, when looking for a cat in a digital image, the first level might identify the pixels on the image that constitute the edges separating one object from another. The second level might be concerned with the task of translating the edges into objects (such as the leg of the cat, the ears of the cat, or the couch under the cat). The third level might group the objects, and the fourth level might determine whether a group of objects is a cat. This approach requires lots of data and computing power, but it does not need prespecified rules.
Because of the repeat dimension in our framework (see the next chapter), firms pursuing a connected strategy have access to the data required for deep learning, making them a better partner in joint problem solving with the customer than their (unconnected) competitors.
The Different Domains for Different Connected Customer Experiences
While we are excited about the emergence and possibilities of new customer experiences related to automatic execution, we want to stress that we do not see them as the best solution to all problems or for all customers. In other words, connected relationships are not always becoming better as you move from the left to the right in table 4-1.
Customers differ in the degree to which they feel comfortable if things begin to happen automatically around them. An experience that is magical for one person might be creepy to another. One person might find it delightful that Disney sends them an automatically created picture book of their last visit to Disneyworld. Another customer might find it invasive. Not only do customers differ in what value drivers (or pain points) are particularly salient to them, but they also differ in the degree to which they are comfortable with sharing data and having the environment around them act on that data. Understanding which connected customer experience is best suited for a particular customer is as important as understanding the particular needs of this customer. Transparency and the ability of customers to opt in or opt out is key in this respect. Unless you have asked the customer whether it is OK for you to collect data, and unless you have explained very clearly how you will use this data and how this use will create value for the customer, you run the big risk of alienating rather than delighting your customers.
In sum, each of the four connected customer experiences has its own merits and works well in specific-use cases and for particular customers:
Respond-to-desire works best when customers know what they want and the firm is capable of providing it quickly. The problem is that fulfilling a random customer request, such as “I want to eat a bacon cheeseburger now, even though I am in a vegetarian restaurant and it is thr
ee o’clock in the morning,” can be costly or impossible. The firm’s essential capability is an operational one: fast delivery, flexibility, and exact execution. Customers who like to be in the driver’s seat, having full control, like respond-to-desire.
Curated offering makes sense when customers don’t know exactly what they want because they don’t know all the available options. In this setting, a firm can delight its customers by finding them a product best suited to their needs, and also gain efficiency benefits by proactively steering them toward something that can be easily provided. The key capability here is the recommendation process. Customers who like to make the final decision but still value advice benefit from curated offering.
Coach behavior is of the most value for latent needs that customers are aware of but have a hard time pursuing themselves, because of inertia or some other behavioral reason. Yes, the customer wants a bacon cheeseburger, but once reminded of his cholesterol levels, he is willing to order a salad. For this to work, the firm needs to have a deep understanding of customers’ needs. This is often based on a rich information flow from the customer to the firm via automated hovering. It also needs to balance keeping the customer engaged and loyal with being parental and restrictive. Customers who do not mind sharing personal data if they see a clear payback in terms of being able to achieve personal goals are willing to engage in a coach behavior customer experience.
Automatic execution should be the connected relationship of choice only if the firm is able to understand the user so well that it is better positioned to make purchase (or other) decisions than the user herself. It also requires a setting in which mistakes are not too consequential. Customers who are comfortable with having a continuous data stream from themselves (or their devices) to a firm and who trust that the firm uses the data to fulfill their needs at a reasonable cost will be the most open to an automatic execution customer experience.
Connected Strategy Page 8