Connected Devices Weekly Update/2016-11-10

• This week's update
  • All Engineers are working towards Dec 16 milestones in the area of Cloud, App/Web and Hardware.
  • Preliminary Rapid Risk Analysis (RRA) review with Enterprise Security completed on Friday Nov 5. Target for Formal RRA review Dec 16.
  • Smaller Team started brainstorming on wish list of product names
  • Initial discussion with legal on documenting SensorWeb Data Collection on Mobile app and Website.
  • UX provided IxD screens for App & Website. Refer to SensorWeb MVP Design doc under App & Website section

• Upcoming Tasks
  • Nov 11 - Share results from User Study #1 in Taipei
  • Mid/Late Nov - Kick off pricing sensitivity study in Taipei

• Software Schedule - Dec 16 Milestone:
  • Firmware: Collect real data from sensors, transmit to cloud

  • Cloud: SensorThings API, receive and store sensor readings

  • Web: Ability to browse collected sensor readings
• Hardware Schedule
  • Nov 22 Working demos on demo boards
  • Nov 23 Printed Circuit Board Assembly (PCBA) ready for test
  • Dec 01 Sample boards ready for Mozilla

• Voice message user test preparation in progress
  • Test runs from Nov 14th - Nov 18th

• Gate 1 preparation in progress - scheduled for Friday in Hawaii
• Software POC work - WebRTC communication between Beaglebone Black boards

Design Document

• First draft of the design document available (aka White paper)
• https://docs.google.com/document/d/1OrCRDCjIuG8OWMBhds8D5eSqJ1Fhl8yPK7U-oBk6V5I/edit#heading=h.aeqt1bhqabd7
• High level architecture, and various integration patterns to connect, secure, manage and use IoT devices. Provide solid technical backbone to claim stated differentiations for our proposed platform. (Open, secure, data ownership etc)

Platform extraction

• 1st pass data available for analysis from dialog with track 1 product teams. Under review
• https://docs.google.com/spreadsheets/d/1i9Lir3Nj65AWZq6QZaRztlPQZ_8CxQ2b6j-i-cVnf6o/edit#gid=0
• Find reusable components and patterns. Prioritize and resource specific blocks and developer enablers.

Leadership checkpoint

• Planned on 11/16

Weekly Notes:

• https://docs.google.com/document/d/1yo1AtuozukgfkZwzO745LdEsB1fdZrEWoIYr8F9eydo/edit#heading=h.qgtmzupjdx35

Update

• Tremendous progress during our engineering get-together in Berlin.

• Raw output from a 10% complete training run. (only 10% due to time constraint. full run will take 2 days)
• (source is what person says, result is what deepspeech understands)
  • - WER: 0.466667
  • - loss: 70.186752
  • - source: "so what this means this incredible freedom of choice we have with respect to work is that we have to make a decision again and again and again about whether"
  • - result: "so what this mens this incredible fred im ofthuis we have with ore spet to work is that we have to make i te susions agin in egin and egin about wither"
  • ---------------------------------
  • - WER: 0.466667
  • - loss: 70.423683
  • - source: "that is to say there were some choices but not everything was a matter of choice and the world we now live in looks like this and the question is"
  • - result: "that is to say thet were some sois is but tat every thin was a anter of jis in the world we naw ivin woks like this and e question is"
  • ------------------------------------
  • - WER: 0.500000
  • - loss: 44.143070
  • - source: "the way to do that is to maximize individual freedom the reason for this is both that freedom is in and of itself good"
  • - result: "bhe way to do that is to maize individual fredim the reasen for thisis of that fredom is in in ofetself god"

Next steps:

• Finish training and correct results based on an english language model. Language model will take care of spellchecking and other basic mistakes.
• On a good path to hit the 10% WER rate

Note: Current values on doyouunderstand.me are bogus. Will start populating beginning of next week once full training run is complete.

Concept Validation Testing

• Data is coming in from the study that was started October 28th to test Abigail product concepts: 1) Smart Family Tracker and 2) Smart Reminders for the Whole Family. Now starting to evaluate the data.
• Data from Facebook campaign, active since October 26th to test consumer interest using different product concepts, is being looked at and analyzed. Abigail product page.

• Made decision to hold off on the video. Will need to find other ways to collect data the video had been expected to capture.

Engineering

• The 5 prototype devices are prepared and ready for use in the study. Resolved wifi issue. Plan to test with a Mozillian tomorrow for a dry run.
• Working this week on a date issue and minor bug fixing.

Abigail Study with Prototypes

• Wrapping up the questions for the prototype study guide that will be used during the visits to participant homes.
• Study will include consumer interaction with the prototypes and collection of feedback regarding usage and responses to questions asked of the consumers.
• The scheduling of participant sessions is nearly complete, with 10 of the 11 that have been identified scheduled for the week of November 13th.

Schedule

• Value proposition validation - Oct. 24 - early November
• Prototype development: month of October
• Assembly of devices: October 31 - November 12
• Bug fixing: October 31 - November 12
• Testing in homes: November 13 - 18
• Data synthesis from pilot and video/landing page studies: November 21 - 29
• Gate 1 pitch: December 8

Hypotheses validation

• Leap of Faith assumption we wanted to validate in the London pilot was “The Magnet experience attracts users to try the app and return frequently”. For that we would need an attractive UX, a stable client and quality content to engage users.
• Concerns about whether the UX that we could implement in the development time we had, was going to be compelling/different enough to validate the Leap of Faith Assumption
• To mitigate this risk we have planned to use a mock-up to run a User Study for 10 people in SF office on the week commencing Nov 28th, 2016. This study will provide us information to further validate the main assumption.
• In parallel we will still run a pilot in London to learn how users interact with the developed client and a FB Landing Page campaign to drive downloads of the app and validate interest

Development

• Finished our second sprint (Sprint 7):
  • Notifications and channel subscription/unsubscription in Android and iOS finished (Notifications demo in Android and iOS)
  • POI view landed (Demo video)
• In the next and last Sprint (Sprint 8) we will focus on refining the way to notify users based on their location, implement call to actions, FTE, navigation in the app and other tasks in order to have the client in the best shape to upload it to the stores on Nov 23.

Offline update this week to the PIB and CD

User Study

• New version of landing page created (Porfirio)
• Feedback from User Research (Rina, Mihaela) integrated into survey
• Twitter ad campaign configured
• Experiment about ready to launch, waiting for feedback from Marketing (Porfirio)

Market Research

• Further market research from Strategy Analytics and Forrester helped refine value proposition and identify target market
• Irina working further with Strategy Analytics to narrow down target market

Technical Feasibility Study

• Push-button commissioning of ZigBee sensors working (Juan)
• Rust ported to ARMv5TE (Juan)
• Progress on getting video stream from camera using RTSP and FFMPEG (Tamara)
• Starting working on client updater (Fabrice)

Engineering

• Still investigating how to build a PCB to eliminate as many "loose" components as possible. Looking for an EE consultant to help with this.

• Working to improve the sensitivity of the Halls sensors after eliminating relays.
• Made the spinner wireless, but it takes a lot of battery power so investigating best battery to use.
• Submitting a P.O. to purchase a 3D printer.
• Researching other games that can be used with the prototype.
• If you haven't seen the video that shows game play using the prototype, a demo of the game board has been recorded.

Research

• Continuing research into competitor products, possible game ideas.
• Met with the UX team and began a conversation about possible consumer segment change from kids to millennials.

Consumer Studies

• Working on plan to test the prototype with Mozilla kids and parents at the Hawaii work week. Will be able to test the technology and consumer reaction to how well it performs, as well as remote play.
• Team is consulting with the research team on development of a study to test consumer response to the product.

User Studies

• Fixed remaining issues with the prototype and readied it for testing.
• First shopping trips for prototype testing were held. Photos were taken of items while in the store, with the participants' descriptions of the items recorded and stored with the images.
• Image matching session followed the shopping trips. Items purchased in the store were brought to the Santa Clara Valley Blind Center. Photos were taken of these items, and descriptions were matched correctly to the items with about a 90% success rate. Results were shared with the engineering team.
• Response to the device by the study participants was extremely positive.
• Another round of testing the prototype with an addition 4 participants is scheduled for next week.
• Product study with 25 participants at the SCVBC is scheduled for November 18th to gauge reaction to and to gain impressions of various form factors.

Engineering

• Discussed feedback from the first prototype testing sessions, held earlier this week to determine what needs to be worked on prior to next week's round of testing.
• Next steps are to focus on bug fixes, possible volume control. Discussed improvements that could be made in how the mic/speaker is installed within the casing to improve ability to hear sound. Conducting analysis of why some images failed to match.
• Would like to consult with an image processing expert. Would also like to consult again on the hardware design and form factor after the study results are in.

Schedule

• Complete development of prototype: October
• Assemble prototypes: October 31 - November 4
• Prototype pilot: November 8 - 15
• Form factor study: November 18
• Data synthesis from pilot and video/landing page studies: November 21 - 29
• Gate 1 pitch: November 30

• Air pollution India’s capital is suffering from worst pollution levels in 17 years. The concentration of PM2.5 particulate is 70 times about the WHO standards. As millions struggle with hacking coughs and burning eyes, many schools across the city have either shut down or ended all outdoor activities. The sources of the pollution are festival fireworks, garbage burns, dust from construction projects and crop fires.
• Amazon Alexa The ecosystem of Alexa skills got its own section on Amazon.com where the skills can be viewed, categorised, searched for and activated. Skill developers can now directly link to their skill’s page on Amazon, improving discoverability and users can get abetter overview of what’s available in the Amazon skill store.
• Digital assistants Business Insider did a comparison test of Siri, Alexa, Google Home and Cortana and declared Google Home as having a slight edge over the competition. In the test, Google’s solution was the smoothest to use and considered the one with the most potential. Apple’s Siri was found to be too dependent on other services, like Yahoo or Wolfram Alpha. Amazon’s Alexa was found to be excelling at simple tasks, but not be able to handle as complicated ones as Google Home could and Microsoft’s Cortana was declared “a shortcut to Bing”.
• Xiaomi Philips and Xiaomi are teaming up to create connected LED lighting for Xiaomi’s smart home platform. The co-developed products will be sold through Xiaomi’s retail channels in China.
• IFTTT IFTTT has replaced its Recipes by Applets, which are more powerful engines that can create entire scenes to add multiple devices and scenarios with “And” and “Or” possibilities. The former Recipes could only be set up with one action and one consequence. Bulleted list item