Retail AI Hackathon: Consumer Feedback

From concept to product in 48 hours is the unique call to action of a hackathon. Agile team management and iterative production bring an idea to life. Teams created and presented bots across a range of shopping experiences, from ordering and giving suggestions to communicating with the brand and post-purchase care. The user feedback supplied here focuses on the formation and potential of the ideas and innovations rather than its final form functionality.

The hackathon was part of OMD’s Retail Revolution research initiative, in collaboration with Goldsmiths, University of London, exploring how consumers accept, understand and trust AI at the point of purchase. Among other methods in the research, we engaged with 12 participants selected based on a number factors including age, family and location-based demographics to experiment at the intersections of AI and retail. As part of the research, they were each given the opportunity to visit the hackathon and to test the bots in a separate workshop. The users were surprised and excited by the variety of use cases and the creative thinking.

Building a bot is an agile process with repeated iterative cycles of development and feedback. We represent this journey in the feedback by first setting out a synthesis of the idea as presented by each team at 2 pm on the first day of the hackathon. At a separate workshop held two days after the hackathon, users watched videos of the final team presentations and had a chance to try each bot. We discuss the user feedback on the final product. From concept to product, this offers a fluid representation of the user feedback.

Key themes that emerged

Users expect a reasonable interplay between their needs and the bot reactions. For example, a simple request for directions or opening times should be instant and easy to access whereas users expected small time delays for difficult questions mimicking the time it would take a human to think and respond. The human pause, to read and reflect, should also be incorporated into the response. They appreciated the instant replies and the fact that they can connect with the brand directly when desired.Throughout the research, we have consistently met with some distrust of the Facebook Messenger platform and a sense that it is not commonly the go-to platform for many of our users. In light of the fast pace of a hackathon, using Facebook Messenger makes sense to produce and test a product quickly. We found that the ‘35 years+’ demographic is less inclined to use Facebook Messenger to contact a brand.

Throughout the research, we have consistently met with some distrust of the Facebook Messenger platform and a sense that it is not commonly the go-to platform for many of our users. In light of the fast pace of a hackathon, using Facebook Messenger makes sense to produce and test a product quickly. We found that the ‘35 years+’ demographic is less inclined to use Facebook Messenger to contact a brand.

Reliability and trust

Users are hesitant in the first interaction with any new platform, although knowing and feeling connected to the brand goes a long way to creating positive associations. Users consistently query transparency, wondering about the use of the data generated through the interaction: where is data stored, is it shared and how is it used. The research indicates that millennials have more trust/privacy issues and demand transparency.

At the same time, many are craving more extensive personalisation we see in the near future technological roll out creating a tension between trusting the bot and the desire for a personalised experience. While users want to know about the use of their data, many are happy to offer personal information, particularly to brands they already know and trust. They appreciate a brand’s attempt to tailor responses to unique needs beyond pre-set suggestions.

Simplicity

The interaction needs to be simple, effective at meeting the specified need and be less intrusive about seeking a transaction. Ease of use rated high among all users. The transactions will follow if the user believes that the brand provides a simple way to adaptive and simple bots. Responses should be relevant, where it can’t find the answer, a brand-specific tone response can mitigate the fail.Whether a user is a satisfier or a maximiser, looking to the bot for utility, research or to shop luxury, bots need to make life less complicated. Learn quickly, focus on preferences and do not overwhelm with long response/push notifications. One of the hackathon participants suggests, “The bot helps you to solve simple life solutions. It’s not trying to convince you to buy a product from us and it is very subtle in the way it plays that game, by simply making suggestions, “would you like me to do this for you”, it feels less intrusive, its feel more like a utility, … maybe this.

Whether a user is a satisfier or a maximiser, looking to the bot for utility, research or to shop luxury, bots need to make life less complicated. Learn quickly, focus on preferences and do not overwhelm with long response/push notifications. One of the hackathon participants suggests, “The bot helps you to solve simple life solutions. It’s not trying to convince you to buy a product from us and it is very subtle in the way it plays that game, by simply making suggestions, “would you like me to do this for you”, it feels less intrusive, its feel more like a utility, … maybe this bot, could help me to plan my weekend, and while is trying to solve that problem for you, it would introduce the idea to approach you somehow to the brand or some products”.

Personality

Personality should match the brand image, but brands have to get the balance right. This is tricky, as one user suggests, “it shouldn’t be robotic but it shouldn’t try to mimic humans either.” A chatty bot can be like an annoying friend that won’t leave you alone or a ray of sunshine on a rainy day. Users seem to like a bot that has a fun and interesting personality but they also feel like the bot should know when they are in a mood for a joke, when they are willing to play, when they need to get the job done, and when they want to end the
conversation.

In the creation of interactions on particular platforms, like Amazon Echo, the ‘skills’ have a predetermined tone of voice. If your brand has its own tone of voice that is a prominent part of its personality, this platform may lose some of the user associations. Find ways to complement and be consistent.

Summary of individual bot consumer feedback

1. Ottobot helps the user create a fun night with friends, and the chance to order snacks and drinks, play games and get fun facts.

Ottobot is a Facebook Messenger chatbot that merges utility with fun. It begins with ‘surprise and delight’ on a trajectory from building an experience through to transaction. It learns user preferences and suggests and answers questions about sports or other events. It has pre-set buttons so the user can choose to invite friends, order food, set reminders or play games.

Consumer feedback overview:

  • Fun personality, which echoes that of the brand
  • New ideas, ways to engage
  • Personalisation is a big hit

2. Beerbot offers personalised labels for your beer bottles through a fun bot in Facebook messenger. Personalised labels open a conversational portal with the brand.

This bot enters from the social experience. The Facebook Messenger chatbot replies to ‘I want to go out tonight’ by finding the best suggestions of where to go and what to do based on location and creates shareable content. It is intended to be conversational thereby instigating a longer interaction.

Consumer feedback overview:

  • The bot’s fun personality is its strongest attribute
  • Great way to connect with friends
  • Once engaged, participants wanted more functions

3. Botrista serves as a customer service portal to manage customer feedback and to provide product information.

As a Facebook Messenger chatbot focused on customer service, the team plans to use Watson’s textual emotion detection to determine the user’s needs. The bot can process complaints which is the main use case but also offer information about products.

Consumer feedback overview:

  • Emotion detection and mapping interactions are the stars of customer service
  • Private moments can be treated with respect
  • Humans are still central to the customer service equation; however, the bot increases response capability without reducing need for human touch

4. Customer care chatbot after the purchase of a bed offers help with delivery and building requirements and then suggestions for bed care, related items and reminders.

Specific focus to create a post-bed-purchase bot helper on Facebook Messenger. Push notifications to interact with consumers on the following: reminders, delivery time, delivery method, inspiration of how to set up, video of how to put it together, initial customer service, share reviews and highlight the human element of the purchase.

Consumer feedback overview:

  • Ease of use, organic process, relationship over time
  • Targets user needs directly based on a specific purchase
  • Customers value clear delivery information and interaction

5. Recipe helper chatbot that generates a variety of recipes based on image recognition of the items in the user’s refrigerator.

Dinner inspiration Facebook Messenger chat bot that answers the question “what to cook for dinner?” The user takes a photo of the contents of their fridge. Using Watson’s image recognition software, the bot identifies available ingredients and suggests possible recipes and/or offers inspiration.

Consumer feedback overview:

  • Brings creativity to the table
  • Image recognition differentiates it from other recipe generators
  • Makes life easier and comes from a trusted brand

6. Scout bot allows users can upload photos (i.e. from Instagram) and suggests how to recreate the photo’s look with the brand’s cosmetic products. Additional options include finding a shop near you, setting up an appointment with a makeup artist, tutorials on how to create the look and finding outfits that match with the look.

Create a Facebook Messenger bot for the cosmetic brand. Users share Instagram pictures and their “look” is matched up with the brand’s products. The products can then be purchased in-store or online, and the user can set up an appointment with the artist to recreate the look.

Consumer feedback overview:

  • Great for finding products and matching new styles
  • Image recognition is the best hook, booking system and store finder very useful
  • Feels like a great extension of the brand
Share.

About Author

OMD EMEA

1 Comment

  1. Pingback: Sherlock czy Watson? Czyli o sztucznej inteligencji w zestawieniu z ludzką. - IAB Polska

Leave A Reply