Commentary

Examining customer experience challenges in retail voice technology

Ken Sutton, CEO of Yobe, offers insight on the challenges inherent in voice shopping: the number of voice shopping enabled devices and the quality of the customer experience when trying to shop with voice.

Photo by iStock.com

March 19, 2019

By Ken Sutton, CEO of Yobe

Voice tech is uniquely positioned to take over the retail market. However, customer experience challenges remain that directly affect full adoption.

The voice shopping market is projected to grow to over $40 billion by 2022, up from $2 billion today. Increasingly, retailers are looking for ways to create customer experiences that are engaging and one-of-a-kind, and while voice shopping exists already in devices such as the Amazon Echo and Google Assistant, there are still some difficult technology challenges that developers must overcome to create convenient and reliable voice shopping experiences.

In an environment where seamless customer experience is everything, the voice shopping market will have to make some important breakthroughs in personalization and voice recognition in order to truly leverage voice shopping - or else customers will be left frustrated and disappointed.

Voice shopping getting ready to take off

We are still in the early stages of voice shopping. Only 2 percent of Amazon Echo owners used the Alexa voice assistant to attempt a purchase in 2018, and 90 percent of those users never tried again. A recent Forbes article points to two factors inhibiting voice shopping: the number of voice shopping enabled devices out in the wild, and the quality of the customer experience when trying to shop with voice.

At the same time Amazon, Google, Apple, Alibaba and Xiaomi are investing aggressively in voice assistants with shopping capabilities. Last May, Google overtook Amazon for the first time in voice assistant sales with a 483 percent increase in sales.

Voice purchasing: Tricky to Implement in the real world

The proliferation of connected devices with microphones offers a huge opportunity for retailers to create a convenient bridge from customer desire, to vocalization of a purchase wish, to a completed purchase. However, voice interfaces are much more susceptible to environmental conditions than touch screens or keyboards.

Shopping interface aside, two environmental factors dramatically affect the quality of voice recognition: background noise, and movement. A home voice assistant must contend with the possibility of cluttered sound environments: music, televisions, conversations, and appliances can create muddled soundscapes where it is difficult for the voice assistant to pick out a specific voice and understand the intended shopping request.

Even more challenging is when the sound sources in the environment are moving relative to the device's microphones. The speaker might be walking through the room, facing toward the voice assistant at one moment and away from it at another. There may be others walking through the room speaking simultaneously, or the device itself might be moving if the user is trying to make a voice purchase using a phone or tablet while walking through a space.

To successfully complete a purchase, the voice assistant must be able to identify the target voice among a forest of interfering sounds and voices, and be able to track that voice as it and surrounding sounds move between the near and far field, and even change positions with each other. A device that is only designed for stationary voice recognition with minimal interference will fail in many environments, tainting the customer experience and discouraging those customers form using the tool again in the future.

Personal shopping experiences require voice ID capabilities

In January 2017, a six-year-old girl ordered a $170 dollhouse and cookies using an Amazon Echo without her parent's consent or supervision. In October 2018, an Amazon Echo owner had to scramble to cancel an order when a television commercial triggered Alexa to begin ordering cat food. Amazon maintains that without a confirmation command, Alexa would not have completed the cat food order, but there are many situations like that of the dollhouse where an improper purchase could be made. In addition to the ability to consistently and correctly understand shopping interactions in challenging sound environments, successful voice assistants will need a biometric or identification component to prevent frustrating and embarrassing accidental purchase.

Voice assistants will need more than voice recognition to succeed

The real world is a complicated place. Voice assistants will need to be able to parse the complex, moving soundscapes of the real world and track their target speakers. They will also need to be able to discern whether the speaker is authorized to make purchases and accept or reject purchase commands appropriately. The voice assistants that will succeed will need more than speech recognition and user accounts. They will need to be backed up by a new kind of voice intelligence capable of navigating real world situations.