Connected Devices Weekly Update/2016-08-11

• Learnings this week center around Technical Feasibility Validation
• Experiment:

Does the hardware that we received work better or worse than initial tests with a RPi and a Mic?

• Prototype building w/ Nascent Objects
  • Software build complete
  • HW under tests for mic & speaker

Nascent speaker max volume test (Conversation is 65dB @ 1m):

Preliminary microphone test results (Baseline WER 15.84%):

• Right now, the technology we started with is working better than the hardware that was made. Our current plan is a follows
  • Use a RPi and Jabra to prove our theory
  • Work in parallel to see if we can get better looking hardware but we will not be waiting for it to move forward with testing

An always-on voice-enabled smart screen device to keep track of the activities, events, and to-do lists of family members. Initial focus for the prototype is on Reminders only.

MVP1 Hypotheses

• Creating a voice enabled solution to create reminders, lists, coordinate schedules, we will reduce the time taken to perform these tasks by 20%. (Proved)
• Offering a solution that provides an alternative way to create and manage reminders, lists & schedules using a smart display, families would be able to accomplish their tasks 20% more times than before. (Pending smart display device to test)

Next Steps

• Increase the number of participants to 10.
• Create a working smart display device with reminder functionality to hand off to 10 busy family members to gather input.
• Evaluate which tablet to use for the study. Look into charging solutions, stands, microphones.
• Based on results from last study, want to test additional improvements, functionality.
  • Improve speech patterns

MVP2 Hypothesis

• Creating a product that leverages voice interaction & chatbot to coordinate small gatherings* will result in these families seeing
  • 50% reduction in the time required to coordinate such events using traditional methods of email, sms, phone calls
  • 25% quicker response rates from recipients

Next Steps

• Conducting a Wizard of Oz style study to gain insights into how people coordinate simple events. In the process of collecting feedback.

Identifying Interest

• Working on identifying potential customers who can be used in focus groups and/or for testing. Goal of signing up at least 50 users for updates.
  • Creating a targeted Facebook campaign to drive traffic to a landing page
  • Creating a polished landing page for Cue that will host a Cue video
  • Creating a video to host on the landing page that will clearly articulate the vision for Cue
• Working on ability to setup sign up/recruitment tables in areas where members of the targeted market segment can be found. Under consideration are:
  • Farmer's markets
  • Elementary schools
  • Soccer Schools
  • Social networking

Experiment

Survey sent to 25 girls in Camden, NJ aged 12-15 years. First we asked them to list all the people they communicated with. Then they grouped those people into “Most Important” and “Important.”

Second we asked them to imagine they could press one button to send a “Sorry, I’m busy right now” message. Then we asked them to write a list of other messages they would send with one button.

Assumptions

If people organized their friends and family into "most important" and "important", they would have 5 to 6 "most important" people.

• Finding: Average was 5.

If people have an easy to way to auto-reply when they are busy, the number of desired auto-replies is small.

• Finding: The girls provided a wide range of desired auto-replies. At the category level, however, there were only 3 types of auto-replies: Reasons (Sorry, I'm eating now), Actions (Txt you back later), and Acknowledgements (Got it, OK).
• These categories suggest that 3 generic auto-replies would work for people. Will test further in future experiments.

Next steps

• Designing a series of experiments that move us incrementally from digital to physical prototype.
• Interviewing potential vendors for form factor prototyping.
• Working with Braden/Legal on vendor paperwork.

• Experiments:
  • #1 (finishing up by Aug10)(https://trello.com/c/fdvHl50c)
    • Hypothesis:If we posted dozens of posters with QR codes linking to real-time PM2.5 data for 3 weeks in Taipei city, there would be more than 100 sessions. If we at the same time provided the notification mechanism, more than 3% among would subscribe for further notifications by providing emails. If we at the same time provided the sharing link through Facebook/LINE, more than 3% among would share.
  • #2 (to begin once we get legal green light) (https://trello.com/c/st2pQBxc)
    • Hypothesis: If we provided the website link to through MozillaTW newsletter mailing list and fan page, there would be 400 sessions. Conversion from landing page to relevant sensor page would be more than 10%. Homepage average concentration subscription rate 3% Homepage share rate 3% Single sensor subscription through the website would be more than 1%. Single sensor sharing through Facebook/LINE would be more than 1%.
  • #3 (to begin together with #2)
    • If we A/B test to compare the conversation rate between simplified and original subscription form, 15% more people would sign up for the simplified one.

• The results of the Project Tablet user study were presented to the board on Wednesday last week
• The board decided to discontinue Project Tablet
• The reason given was that we have limited resources in Connected Devices and there are still major concerns around:
  • Apps
  • A lack of support from the platform team
  • WhatsApp

• Feedback from PIB was to explore "Upstream" integrations and Technical Partnerships
• Upstream Integrations
  • Experiment #6 - Concierge Prototype to investigate if an upstream integration can be used to design a meaningful service from a user perspective
  • Question being answered: Can an upstream integration with a grocery store or delivery service provide sufficient information to create a service that 1) is useful to people and 2) has differentiator qualities
    • Collect users purchases and enter into Database
    • Use existing Bot MVP to request meals and recipes
    • Pre-interview/Post-interview
    • Recruit participants that shop weekly from internal Mozilla and Facebook friends/family
    • Duration 3-weeks of user trial (want to capture two weekends of shopping). Targeting results by 9/9
    • Market research of other apps in this area.
    • Legal draft consent available by EOD.
  • We will pause on contacting Grocery/Delivery services with a proposal until we have results from Experiment #6

* Image recognition device

• • New training added today (carrots and ground beef)
  • 2 users using the device currently. Third to start today.
  • Metrics in progress

* New Survey designed to dig deeper into the existing problem and explore new problem areas.

• Apps passed initial app-store review (android and ios), but not publicly visible yet.
• Will be submitting app for second review following large telemetry update shortly.
• Preparing content for View-Source conference experiment in September.

• Seek a deeper understanding of how people think of being at “home” in order to better understand how technology can best fit in and how displays can be touch points into the experience
• Half way into the 8 week program (now onto ideation stage)
• 5 themes emerged from the 4 week research
  • Facilitate Shared Experiences (Better us)
  • Help me be Better (Better me)
  • Embrace a Budding System
  • Create an Inviting Home
  • Keep me Present
• Explore into the better us / better me areas
  • Better routines/habits in children
  • Enhance family communications (in home or outside of home)

• Skype unveiled new chatbots for travel booking and controlling IoT devices through IFTTT. "To call Microsoft bullish on bots may be an understatement. On multiple occasions, CEO Satya Nadella has stated that conversational computing is the new interface and it will “fundamentally revolutionize how computing is experienced by everybody".
• The Alexa Skills Kit got support for templates for building voice-based games, alongside the existing templates for crafting command-and-response skills and controlling Smart Home devices. Currently, 20% of Alexa's top skills are games.
• This article describes Amazon's plans of increasing their investments in 2 strategic areas: original video content and cloud services (AWS). AWS was the growth engine for the company in Q2. Voice recognition is another area of investment mentioned.
• In this article MIT shows a way to mitigate potential privacy concerns about IoT for energy management in the Smart Home through doing processing of the data locally on the client or within the home network, when possible.