In the past two or three years the consumer market has seen the idea of the Internet of Things (IoT) go from a prediction to reality. The first wave of IoT products was largely fueled by the parallel innovation of crowdfunding, which allowed makers and early stage ideas to get off the ground without traditional funding sources. Many feel that the promised innovations from IoT have not yet been realized. Almost weekly another crowdfunded startup announces it’s closing its doors without ever shipping a product. Products that do ship often offer a poor user experience and are notoriously buggy and insecure. In fact, a recent article—Why Is My Smart Home So {omission} Dumb?—expresses many consumers’ opinions about IoT devices. Drawing on his personal experiences as a founder of the IoT company Emberlight, Kevin Rohling discusses the challenges he has encountered—from security and manufacturing to UX design for an IoT product. He will make comparisons to his past experiences with mobile and web products focusing on what he sees as IoT’s unsolved problems, such as monitoring and firmware management. Join Kevin as he looks forward at the forces that are shaping the next wave of IoT products—speech interfaces, new wireless standards—and the consolidation of IoT platforms.
2. Kevin Rohling
Boomtrain
Previously the VP of product at Emberlight, Kevin Rohling is an entrepreneur with
a strong engineering and product background. Kevin held previous positions as
an early engineer at Card.io (acquired by PayPal), CTO at Breezy, and CEO of
CISimple, which he sold in 2014. His passion is the intersection of challenging
engineering problems and intuitive user experiences. Follow Kevin on Twitter
@kevinrohling.
3. The First Wave of IoT
Blood in the Water
about me
cto @ breezy
ceo @ cisimple
vp product @ emberlight
sr pm @ boomtrain
kevin@kevinrohling.com
github.com/krohling
@kevinrohling
4. iot
IOT WHAT?
The Internet of Things is a network of physical objects that contain embedded
technology to communicate and sense or interact with their internal states or the
external environment. -Gartner IT
26 BILLION DEVICES BY 2020!
source: gartner
iot
0
400
800
1200
1600
2014 2015 2016 2017 2018 2019 2020
$1534 BILLION MARKET!
IOT DOLLARS… IN BILLIONS (CONSUMER)
source: gartner
6. iot
MONEY
TOOLS
COMMUNITY
EASIER ACCESS TO CAPITAL
Crowdfunding Platforms
‘Long Tail’ Equity Investing Platforms
Increased availability of Institutional VC Funds
source: cbinsights
FUNDING IN IOT INCREASED FROM $768M IN 2010
TO ~$2B IN 2015.
iot
MONEY
TOOLS
COMMUNITY
BETTER TOOLING AND PROTOTYPING
INVESTOR DEMO IN MINUTES!
$25$35$19
8. iot autopsy
Smart switches that don't need switching.
Overestimated machine learning & underestimated efforts in converting a prototype to
fully functional hardware product
“Hardware products sell at 4x–5x the component costs. How did we not know this?!”
“Building a prototype is the easiest part of building a hardware startup. The real
challenge comes in product design, production engineering, manufacturing, distribution
and marketing/sales.”
source: medium
9. iot autopsy
The 3D printer everyone can use.
They produced an amazing product ... The founders simply failed at building a viable
company around the product.
The company is sourcing new rounds of investment and found that investors want the
new cash to only be used as working capital and not to fulfill previous obligations.
source: techcrunch
Raised $1.5M on Kickstarter from 3500 backers.
iot autopsy
A portable party disguised as a cooler.
When Coolest Cooler was launched on Kickstarter, it cost between $165 and $225, a
price its creator Ryan Grepper said in an update to backers was far too low.
“The Coolest Cooler ultimately cost more than we expected to develop and manufacture”
source: motherboard
The 2nd largest crowd funded project in history with $13M from 62K backers.
Grepper said the company needed another $15 million to deliver on the company’s promise.
10. iot autopsy
Prototypes are EASY! But…
iot autopsy
Security is HARD!
User Experience is HARD!
Manufacturing is really HARD!
Making an actual business out of all this?
Damn near impossible.
Prototypes are EASY! But…
11. A"f Noori
CEO
Kevin Rohling
So'ware
Gordon Kwan
Hardware
Steve Arnold
Design
Kevin Wolfe
Firmware
Levi Wolfe
Cloud
Tony Lee
Marke9ng
Lexii Jaye
Community
AngelPad
our team
12. KICKSTARTER: August 2014
Funding Goal: $50K
Final Raise: $300K (6x Goal)
From: 2600 Backers
Planned Ship Date: February 2015
kickstarter
PROTOTYPING v1
Early “Works Like” prototypes were developed using electric imp.
The upside with electric imp was that it
had integrated WiFi connectivity out of
the box and plenty of IO pins.
It also has a “cloud” backend which
means you can get up and running fast.
prototyping
13. 2nd Generation of Emberlight prototypes:
PROTOTYPING v2
Custom Hardware Designs (In-House)
Custom Firmware (In-House)
Integrated WiFi and Bluetooth 4.0/BLE
Smaller 3-D Printed design
prototyping
3rd Generation of Emberlight prototypes:
PROTOTYPING v3
Switch to Marvell SoC
Injection Molded Casing
Redesigned Hardware
Rewritten Firmware
prototyping
15. timeline
May 2014 - First Functional Prototype
August 2014 - Firmware Development Started
September 2014 - Kickstarter Fundraise
October 2014 - Functional Prototype v2 (CSR Chipset)
January 2015 - Firmware Rewrite Started (Marvell Chipset)
February 2015 - EVT 1 (Electrical Validation Test)
June 2015 - EVT 2 (Electrical Validation Test)
August 2015 - EVT 3 (Electrical Validation Test)
September 2015 - DVT 3 (Design Validation Test)
November 2015 - PVT (Production Validation Test)
January 2016 - FCC/UL Testing Complete
shipping!
Emberlight just started shipping in March and will be fulfilling all orders!
17. funding
Funding? But wait… Kickstarter! right?
NO! Misconceptions about Crowdfunding:
1) Crowdfunding != Investment
2) Crowdfunding == Pre-Orders
3) Crowdfunding will not pay your salary. Or your Developer’s salary. Or
your Marketer’s salary. Or… anybody else’s for that matter.
4) Crowdfunding does not make anybody rich. In fact it is probably far
more likely to do the exact opposite.
funding
You are not building a product. You are building a business.
To build a business you will need additional funding from traditional
sources. That means pitching Angels and probably even VCs.
Just go ahead and plan on that happening. It will suck.
18. funding
My recommendation for fundraising a Consumer IoT product:
STEP 1: Raise some angel money (~$100K-$200K)
STEP 2: Pour that money into a marketing campaign to generate pre-orders.
STEP 3: Use your traction from #2 to raise traditional funds.
If STEP 2 generates less than $1M in pre-orders STEP 3 will be VERY hard.
MANUFACTURING
Time to turn that prototype into a real product! Fun!!
19. manufacturing
Selecting a manufacturer
This is a very big decision. Absolutely, get references from other startups.
Manufacturers are huge companies and they still do not know how to
work with startups.
Overseas communication
Timezones, language barriers, and Skype all conspire to make it as difficult
as possible to communicate complex ideas and dependencies to your
partners in China.
EVT - Electrical Validation Test
Are your power sources, radios and all other electronics working and safe?
DVT - Design Validation Test
Are there any issues with the physical design? Do the welding points
match up? Does it look half way decent when it’s done?
PVT - Production Validation Test
Verify that the production line is ready to make the product and all
assembly and post-assembly testing equipment is functioning properly.
validation tests
20. validation tests
Some of the issues we had at Emberlight
Failed EVT Tests:
We went through 3 EVT runs because of issues w/ the power supply and
radios. This delayed our product ship by about 4 months.
Failed DVT Tests:
We discovered that we would need to lower the supported bulb wattage
because of insufficient heat dissipation with the physical design.
certification
Don’t burn people’s houses down.
UL and FCC Certification was >$50K and took >6 months to get
completed.
21. PRODUCT DEVELOPMENT
This really is the fun part!
1) User Experience
2) Security
3) Connectivity
4) Testing
user experience
Smart Phone vs Light Switch.
22. user experience
Smart Phone vs Light Switch.
1. Find your phone
2. Turn it on
3. Unlock it (passcode)
4. Locate the “lights” app
5. Turn light on
6. Put your phone away
1. Walk up to the light switch
2. Flip the switch
3. Go back to bed
user experience
1) The smartphone is not the right answer.
2) If it’s not broken, don’t fix it.
3) New interaction models must be very low friction.
23. security
IoT devices are simply harder to secure.
-Limited hardware resources
-Intermittent or no cloud connectivity
-3rd Party Platform Integrations
-Potentially multiple RF access points (i.e. Wifi,
BLE, Zigbee, etc.)
security
Common IoT Device Vulnerabilities
1. Open BLE/LAN APIs
Interacting with a device over the LAN or BLE
should require authentication. Many devices
allow unrestricted access via the same WiFi
network or over BLE.
All access points should be secured.
24. security
Common IoT Device Vulnerabilities
Devices should be protected against physical
theft or access. Changes to the network
connection or re-assignment to another user
account should be restricted.
2. Unrestricted ‘Reset To Factory’
security
Common IoT Device Vulnerabilities
Example:
Nest devices were found to allow unsigned
firmware updates over USB once placed into
‘Reset’ mode.
2. Unrestricted ‘Reset To Factory’
25. security
Common IoT Device Vulnerabilities
Developers should assume that their firmware
code will be publicly accessible.
It is very difficult to prevent people from
dumping device memory.
It is also relatively easy to monitor UART and
other onboard communications.
3. Security/Encryption keys embedded in firmware.
security
Common IoT Device Vulnerabilities
Example:
source: http://blog.sec-consult.com/2015/11/house-of-keys-industry-wide-https.html
A recent report by SEC Consult analyzed 4000 embedded devices and found 580
private keys.
These keys include ~150 server certificates which combined account for 9% of all
HTTPS hosts on the web! (3.2M hosts!)
These keys also include ~80 shared SSH host keys used by 900M hosts!
3. Security/Encryption keys embedded in firmware.
26. security
Common IoT Device Vulnerabilities
Example:
source: http://blog.sec-consult.com/2015/11/house-of-keys-industry-wide-https.html
A recent report by SEC Consult analyzed 4000 embedded devices and found 580
private keys.
These keys include ~150 server certificates which combined account for 9% of all
HTTPS hosts on the web! (3.2M hosts!)
These keys also include ~80 shared SSH host keys used by 900M hosts!
3. Security/Encryption keys embedded in firmware.WTF!!!
connectivity
Pick a radio. Any radio.
Connectivity options are highly fragmented there are many tradeoffs to consider:
-How much of your BOM is dedicated to radios?
-How important is power consumption?
-How important is latency?
-Do you need real time communications?
-Do you need a persistent network connection?
-How will you be performing firmware updates?
-Does your device need to communicate directly with mobile phones or tablets?
27. connectivity
At Emberlight we chose both WiFi and BLE. Why?
We were able to get a combo chip from Marvell that had both radios integrated.
This reduced cost and the complexity of our PCB.
BLE allows us to have very low latency communication with mobile devices,
essentially supporting real-time control when within BLE range.
BLE also allows us to dramatically streamline the setup process because no
pairing or credentials were necessary.
WiFi allowed us to support device control from external networks and easily
push new firmware updates.
testing
Testing is very difficult with IoT products.
Mobile developers have Simulators and Emulators. Not true for embedded
developers. This makes it near impossible to meaningfully test the integration of
hardware and software changes.
IoT/embedded developers need low cost PCB/Circuit emulators, unit
testing frameworks and CI processes that integrate w/ hardware designs.
28. iot
CHARACTERIZING THE FIRST WAVE OF IOT
How do we fund IoT products/startups?
How do we design user experiences for connected hardware?
How do we connect all of our devices so that they all work together?
How do we develop secure, reliable and connected hardware systems at scale?
iot
CHARACTERIZING THE FIRST WAVE OF IOT
In getting to this point many companies have died in the funding and
manufacturing stages.
The ones that made it through have helped discover new UX patterns and
exposed poor security practices.
However, very few of these products/companies have
resulted in successful businesses.
29. IOT NEXT
1) Funding
2) User Experiences
3) Cloud Services
4) Wireless Connectivity
5) Cross Platform Tools
6) Deep Learning Systems
iot next
FUNDING
More established companies are getting
into the IoT game taking territory from
startups.
Also, as institutional funding for IoT
increases fundraising patterns will start
looking a lot more like software
companies.
30. iot next
FUNDING - IN CASE YOU’RE LOOKING
iot next
USER EXPERIENCES THAT WORK
Fewer startups will be developing their
own user experiences. Instead they will
leverage devices/services like the
Amazon Echo and the Alexa Voice
Service.
31. iot next
MATURE CLOUD SERVICES
Startups already have to develop fewer
backend systems themselves.
Connectivity to other IoT products and
automation services is already becoming
much easier.
Walled gardens will fail.
iot next
HUBS AND AGGREGATION OF WIRELESS STANDARDS
Devices like the OnHub, which have a
wide array of wireless radios will start
making the decision over wireless
standards much easier.
More importantly it will also make it
unnecessary for IoT companies to
develop a companion hub for their
product.
32. iot next
CROSS PLATFORM TOOLING
The choice of which chip to use is
becoming much more flexible.
With the release of tools like ARM’s
mbed platform and partnerships with a
large number of chip manufacturers it
will finally be possible to write firmware
once, against a consistent set of APIs and
run it on a large number of different
chips.
iot next
INTEGRATION WITH DEEP LEARNING SYSTEMS
With the potential for AI to improve voice
systems and contribute to a broader
understanding of customer preferences
and usage patterns it’s inevitable that
Deep Learning will become an important
part of IoT.
Yes, this does have the potential to
become very creepy. Let’s hope it
doesn’t go that direction.