Beta launch plan

[Berkmann18]

Here are the things we need/should look at:

1. Goals
2. Metrics
3. Other things we should do or think about
4. What next

   ---

Note: "them"/"their"/"users" refers to beta users.

Goals

Those are what we want the beta launch to accomplish.

1. How large our beta users pool will be?

   • Those in the "covid.js gang" messenger group (24 in total).
   • Perhaps some people each of us know, who would be interested in trying the app?

2. How and where will we reach out to them? Using Messenger (for those in the group chat) or Slack (using the appropriate channel(s)).

3. What exactly we want from them using the app (edit: see form questions below)?

   • Specs (namely: device model, OS, CPU/GPU, RAM)?
   • Browser(s) used?
   • Their thoughts on the app (satisfaction, their impression of the app' speed, rooms for improvements)?
   • Environment? Is the room lit? Does the app handle day and night time well?

4. How should the end-user experience be?

   • Web (MPA/SPA) (<-- we are using this right now)
   • Progressive (PWA)
   • Native (NA)

Metrics

1. What are the metrics that we'll collect around their interaction? i. How long the app took to load on their device (so essentially the Time-To-Interactive LightHouse metric) ii. How many times did they touch their face vs how many notifications they had? iii. Where the notifications done at the right time, if not how late were they? iv. How many times did they go on the app? v. Were they using the app in the foreground or background?

2. What thresholds will give us the confidence to move to general availability? i. Did the app allow them to have a significant amount of memory/CPU/GPU power left for other tasks? ii. How close were webcam/browser lags to 0? iii. Did the background executions inhibit the app from performing its duty? iv. Was the app kept alive for the whole browser session?

Other things we should do or think about

1. Our target audience is primarily is those using a desktop/laptop with mobile users coming second.
2. The releases should follow a semantic versioning approach (we currently in alpha-v.??). But we should also find (or make) a tool that automates the release process (similar to how semantic-release does with known commit conventions).
3. App monitoring to check if it's down and how it copes with users? https://github.com/misterpeddy/hands-down/issues/9#issuecomment-616240034 ~4. What should be the privacy policy and Terms of Service that we use? Some examples would be https://prinsec.com/terms-of-use or https://www.jrtapsell.co.uk/privacy.html~
4. Do we have a balanced distribution of true/false labels with enough variety of hand and face positions?

What next

• Finishing the WASM backend implementation for the Conv2D part of TFjs.
• Hysteresis / RNN implementation.
• Test suite with good coverage.
• Further optimisations in performance, app-size and such.
• Chrome extension.
• Multi-hand model(s)
• Other matters that are on https://github.com/misterpeddy/hands-down/issues

[misterpeddy]

Form questions:

• Did you feel it was too slow to load?
• Did you quickly understand how to use the app?
• Was the app sluggish?
• Were there specific configurations where the model was confused?
• Anything that surprised you?
• What other kind of notifications / experience would you want? Any options that should be provided to control the experience?

[misterpeddy]

https://forms.gle/fyBYtopNySkDdyrv7