1. The Internet of Things:
Signal, Noise,
& Business Models
Dalton Caldwell
Founder/CEO App.net
2. Good news!
• The “The Internet of Things”-era has begun
• Polished, useful IoT products released daily
• Amount of information sensed/stored by
IoT will increase exponentially
3. (Some) bad news.
• IoT is going to make things a lot “noisier”
• Making sense of the noise is harder than
building the machines which create it
• Think your email inbox and social feeds are
overwhelming? Just wait.
• ie how much data will Google Glasses generate?
4. What can be learn from
running a server farm?
• Event notifications must follow Goldilocks
principle: all the way on or off is bad
• What do you do with 10TB of logfiles?
• How manageable are SMS/Email alerts
from: 1 server? 10 servers? 100 servers? 1000
servers?
5. Noise management on
the social web
• Think about services that auto-tweet or post
Open Graph actions
• Watching a video, reading an article, listening to a
song... these actions generate noise
• How are consumers supposed to manage this?
• On an unfiltered stream: unfollow the person, or ask
them to stop posting so much “spam”
• On a filtered stream: let the algorithm do the work for
you, also you can give the algorithm hints
6. Filtered Streams vs
Unfiltered Streams
• Consuming an unfiltered stream is increasingly
difficult... human limits on information consumption
• Algorithmically filtered streams make arbitrarily
large amounts of input data feel manageable
• FB newsfeed: a great filtered stream implementation
• Surfaces important events
• Eliminates redundancy
• Optimizes on time and engagement metrics
7. How will people manage
their personal IoT?
• Raw IoT data is specialized, but event/alert
notifications will likely utilize:
• Email, SMS or Social streams
• A stream is especially useful because it
allows sophisticated noise filtering
8. Some business model
challenges...
• If you are an advertising-supported stream
company, you there is a good chance you:
• Don’t want 3rd-party UIs displaying the stream
• Don’t want 3rd-party filtering algorithms
• Don’t want your “content” leaking out
• Don’t want to host home automation-type data(?)
• Optimize for user engagement on your stream, not
utility. Those are not the same thing.
9. What if we ignored those
business model limits?
• Unbundle the stream.
• Pluggable APIs which write to unified stream
• Pluggable filters of unified stream
• Pluggable UIs + methods of stream consumption
• Ownership/exportability of all data
10. Examples of existing
unbundled services
• Unbundled read/write APIs? IFTTT
• Unbundled UIs? 3rd-party Twitter clients
• Unbundled stream filters? Flipboard
• (Worth noting these examples have all
faced recent “platform risk” setbacks)
11. App.net:
Built to be unbundled
• Pluggable APIs? Nearly 100 apps in the
directory, dozens of available libraries/
programming languages
• Pluggable filters? 3rd party notification
frameworks and threshold filters.
• Pluggable UIs? ~90% of usage already occurs in
3rd party UI, exactly as intended
• Users own data. Data export since day 1.
12. App.net private
messages API
• Person will be able to use their account to
create a private, unbundled stream
• A single paid account can be written to from
any number of endpoints, including IoT
• The stream can be filtered/consumed in any
way 3rd party devs can dream up
• App.net’s paid, “services” business model
supports this kind of use-case
13. Wrapping up...
• Digital information is already overwhelming, but
IoT will make this far problem far worse
• Making data useful: more important than collection
• Existing unified streams have business models
which discourage unbundling of filter components
• “Unbundling the stream” allows focused innovation
in UIs, filters, and notification frameworks