1.
A Lean Hardware Toolbox
Marcus Gosling
Design Lead, Highway1

2.
Overview
16 weeks.
Selective, Immersive, tailored.
Goal setting
China visit
Demo day ->

3.
Some examples of Highway1 alumni:
Navdy, a heads-up display for the car you already own

4.
…..
Ringly - Smart jewelry that notifies you about what matters most

5.
Cue - perform medical diagnostics at home

6.
Drop - connected kitchen scale and recipe app

7.
Teams enter with prototype
- Validated with users
- Not designed, or engineered for manufacturing

8.
Teams exit with a product
- Highest risk areas addressed, significantly closer to manufacture
- Design - Product experience more developed, ability to communicate their vision.

9.
9
Hardware is amazing!
- Potential for wide impact - different use cases, different contexts, different users
- Multiplies the power of networked computing technology - wearables, IoT
- The hardware ‘cool factor’ - motivation, learning, excitement, attention, hiring!
- Novel business models, multiple revenue sources.
- Huge potential upside if you have a hit product,
But…

10.
Hardware is terrifying!
Supply chain management
Manufacturing costs
Fulfillment
Inventory
Returns!
(Plus you probably have a connected mobile app to develop on top of all that :)

11.
Ideal product development roadmap
This is a representation of a typical product dev cycle.
For a simple CE product this could cover 12 - 18 months.
Unlike software, where the technology used to validate, develop and deliver a product
is essentially the same.
Hardware uses very different and very specialized tools depending on your stage of development.
ID = Industrial Design
PD = Product Design Engineering
DFM = Design for Manufacture
PM = Project Management
Almost universally, these tools get exponentially slower and more expensive as you move from left to right, closer to mass production.
Now let’s look at what actually happens…

12.
Actual product development roadmap

13.
Actual product development roadmap
Lot’s of trial and error, incorrect assumptions, new discoveries.

14.
Actual product development roadmap
As we get into industrial design, we need to look at multiple approaches,
and balance those with technical considerations.

15.
Actual product development roadmap
As we get into real Design for Manufacture and start to work with vendors the design must adapt to their individual constraints and efficiencies.

16.
Actual product development roadmap
Ouch, 2 week delay with a production line standing idle. $$$

17.
Actual product development roadmap
Things start to stabilize.
Issues come less frequently
Quality and yields rise.

18.
All-too-common product development roadmap
Smooth sailing…

19.
All-too-common product development roadmap
but, we are getting panned.
1% of our customers are not happy and are writing 50% of our reviews.

20.
All-too-common product development roadmap
It’s an issue that would have been easy to find
and cheap to fix with a little more early prototyping.
It won’t be cheap to fix now.

21.
For hardware, you have to get a lot of things right to win precious reviews.
Design, Features, Customer support, Documentation, it’s all the same thing to your customers.
At the point where you hit retail it’s often too expensive or too late to fix a hardware problem.

22.
Managing a high-risk, increasingly costly process.
A more useful mental model for the product dev cycle
is a series of gateways.
You do not want to pass through each gateway before you are 100% certain you are ready.

23.
Understandthecontext
Managing a high-risk, increasingly costly process.

24.
Understandthecontext
Deliverrealvalue
Managing a high-risk, increasingly costly process.

25.
Understandthecontext
Deliverrealvalue
Delightyourcustomer
Managing a high-risk, increasingly costly process.

26.
Understandthecontext
Deliverrealvalue
Delightyourcustomer
Executeatscale
Managing a high-risk, increasingly costly process.

27.
Do I understand the context?
Does my product deliver real value?
Do I know how to delight my customer?
Can I execute brilliantly, at scale?

28.
Do I understand the context?
Does my product deliver real value?
Do I know how to delight my customer?
Can I execute brilliantly, at scale?

29.
Highway1 startups

30.
Highway1 startups

31.
Highway1 startups - ‘The Sweet Spot’
Application process:
Team, Market and Product.
Culture and Progress.
Culture:
Humility
Curiosity
Vision

32.
A Lean Hardware Toolbox
• Investigations
• Instrumentation
• Hacking products
• Manufacture
• Sequential production (1-10)
• DIY Batch production (10 - 50)
• Outsourced batch production (50 - 500)
• Pilot production [100 - 1000]
• Mass production [10 - 100K]

33.
Investigations

34.
Sereneti are building a robot chef for consumers.
It uses pre-prepared, raw ingredients and robot smarts to cook you a fresh, hot meal.

35.
The Existential Questions
• Can a robot cook?
• Will people eat food cooked by a robotic appliance?
• Will people find enough value in our solution to
choose it over existing solutions?
It’s an ambitious project, with some interesting questions.

36.
Initial investigation:
What is actually happening when a good chef cooks a great dish
and what do they do differently from a less skilled chef?

37.
The team records:
ingredient timing, temperature, and stirring patterns
They find that by carefully following this ‘script’
they can reproduce the dish consistently

38.
They explore various methods of ‘mixing’ food

39.
A series of prototypes explores different aspects of the robot.

40.
A working unit is installed in the Highway1 kitchen…

41.
…and a lunch menu is displayed.
The goal of this is twofold.
Will people like being cooked for by a robot?
Will they be able to load and empty food trays, etc?

42.
Here are the pre-prepared trays with raw ingredients.

43.
A meal and happy customer.

44.
Over 4 weeks the team is able to observe problem areas
and make modifications between lunch times.

45.
By hacking on the menu app,
by using text messaging
and modifying and hacking the hardware
They refine the experience.
Engagement is good and people are strangely drawn to this machine that prepares them a home-cooked meal.
“The way to a person’s heart is through their stomach.’

46.
sensassure

47.
7,000 km
This team is helping elder-care facilities take better care of seniors, many of whom are incontinent and need to wear diapers.
Changing too often disturbs the senior and is wasteful.
Changing not often enough can result in discomfort, dermatitis, and pressure ulcers.
They ‘got out of the building’ and took a 7,000k road trip, visiting 10 centers…

48.
…taking some initial prototypes with them.

49.
They quickly learned some important factors if their product is to be successful.
The device must not be noticeable when worn. Many patients have dementia and will remove anything unfamiliar.
Caregivers need to be able to fit the device on their customers in seconds - they are very busy people.
Minimal training should be required - caregiver turnover can be high.

50.
They worked on some quick basic non-working models
to refine the training and fitting experience
and to make sure that elders would not detect the device in place and try to remove it.

51.
A later version.
This design also incorporates the teams greater understanding of how much fluid the diaper should have before requiring a change.
This understanding had significant implications for the electronic design of the sensor unit.

52.
The team were highly embedded.
Here is one of the engineers caught dosing off on his shift!

53.
By working closely with Facility administrators and caregivers they were able to refine both the hardware and the software for managing who needed to be changed and
when.
Physically checking seniors to see if they needed a change was eliminated, resulting in less disturbed sleep and happier customers.

54.
The toolbox for investigation is wildly diverse.
Storytelling is one of the fastest ways to prototype an experience and share it.
10 minutes sketching on a whiteboard with an expert can save you years of work.
Anything goes from Foam-core board to chopped onions.

55.
Littlebits is a very quick way to prototype product behaviors, etc in minutes not hours.

56.
Instrumentation

57.
The team at Loop are building a premium, connected photo frame.
They want to enable you to send a picture,
with a few taps on your phone
to a ‘loop’ device located in the home of a family-member or friend.

58.
Their earliest prototypes included screens scavenged from tablets.

59.
They did a lot of user testing of both form and functionality.
One of their early pilots involved leaving their proto in the homes of users for extended periods of time. Users were happy and did not report any major issues,
They had been logging usage metrics to measure engagement over time and noticed that some users were powering off their units in the evening.

60.
When they asked about this they were told that the screen was too bright at night when the house was dark. They were turning it off to avoid glare.
This problem could have become far more expensive to fix the later they discovered it as it required an ambient light sensor to be incorporated into the hardware design.

61.
Habit Aware
Habit aware are making a device to help people with body-focussed repetitive behaviors, e.g. nail-biting, eyelash-pulling, etc.

62.
1st Prototype: Lo-Fi
Heavy bracelet
slides down the
wrist when hand
moves to the pull
area.
This bracelet hits a
second bracelet,
causing bells to
ring & notify the
user to move away.
This simple prototype asked the question.
Does being aware of when you are doing your habit help you stop?
Yes, it helps.

63.
2nd Prototype: Off The Shelf
The second prototype used a Metaware Dev Kit.
It asked the question, does the value of awareness decrease over time, do you get used to it and ignore it?
The prototype needed to be comfortable and possible to wear for extended periods.
No, you do not get used to it. It keeps helping.

64.
One of the co-founders of habit aware has had an eyebrow and eyelash pulling behavior all her life.
Over several months of wearing the first electronic prototype she was able to make significant progress in kicking her habit.

65.
Behavior Change by the Numbers
Number of Hair Pulls Daily
Wore Bands
Didn’t Wear Bands
By instrumenting the device they were able to provide stronger proof that it was working. Showing much lowered behaviors during periods of wearing the
device.
Their smartphone app also allowed the user to report qualitatively on successes and failures.
Behavior change is a slow process so metrics and goals are key to this experience.

66.
3rd Prototype: Custom
The 3rd prototype was a custom design.
The goal of this was to increase battery life and improve gesture recognition, increasing accuracy. This would enable field trials with multiple people over
an extended period of time.

67.
500
250
15
90
51
Total Attendance
Individuals with
Hair Pulling disorder
Unique visits to booth
Email Sign-ups
Device Orders
BFRB Conference, April 15-17 2015
68
The team attended the ‘Body focused repetitive behaviors’ conference, where they started to form their community and sold 51 pre-orders.

68.
Instrumenting hardware
• Particle Photon (WiFi)
• “Spark.publish("temperature",
String(temperature) + " °C”); delay(2000);
• 75c memory chip to store log files
• SD card for logging

69.
Hacking existing products

70.
Be a rock star
This team wants to reduce the friction and barriers to being a guitarist.

71.
Photo credit: James Fortune Photography
Dream

72.
Dream

73.
Problem
The guitar as we know it has a 300 year old UI,
dictated by the physics of vibrating strings.

74.
75
90% quit in first year
Reality
Most would-be guitarists just want to be able to play
and sing along to their favorite songs.
But 90% quit in year 1.

75.
Revolutionary UI
Musical
superpowers
Software with
musical
intelligence
A guitar anyone can play
is reinventing musical instruments
The team had a hypothesis.
If you change the UI and make the guitar smart.
It could still be an expressive musical instrument, not just a game or toy
but could be much easier to learn.
You could be playing songs in minutes not months.
But a smart guitar is a complex product, how could they validate this hypothesis with limited resources?

76.
…By using an existing product as a starting point.
This electronic teaching guitar is played in a similar way to an existing guitar, but with thin buttons instead of strings.

77.
The magic instruments team wanted to take advantage of the product hardware,
but completely replace the software.
They wanted to change the guitar’s UI so that instead of holding down sets of buttons with different fingers in complex shapes, you could use one finger to press one
button and create the very same chord.

78.
Here is the existing electronics.

79.
Here with the Magic Instruments prototype ‘brains’ added, replacing almost all of the existing functionality.

80.
Arduino 1
Arduino 2
Speaker upgrade
Amplifier
Existing
String sensors
One Arduino development board controls the fretboard.
Another plays the sample library of sounds.
They upgraded the speaker to improve the sound and volume and added a more powerful amplifier.
The only electronic component they are re-using are a set of 6 op-amps that detect strumming and finger-picking on the guitars ‘strings’

81.
This budding rockstar has never played the guitar before
but is making a pretty good first try of Adele’s ‘Someone like you’

82.
She just needs to hold down the button corresponding to the number.
Much easier than learning and forming a chord shape.
But, singing, strumming and setting a good rhythm are still needed for a good performance.
Like a guitar, mastery is still required but the learning curve is less steep and results are almost immediate.

83.
By cornering the market for these existing guitars on eBay the team was able to create a fleet of modified tester guitars and leave them with users for extended periods of
time.

84.
As feedback is coming in they are gradually replacing different parts of the guitar with custom hardware.

85.
“I would say we probably saved ourselves a year by using
an existing product as a shortcut - mostly because we were
able to build 6 functional prototypes and do a lot of user
testing quickly. Otherwise, we would probably have built
one crappy prototype, and had a much harder time getting
good user feedback.”
Brian Fan, CEO Magic Instruments

86.
Lully is an early stage startup with revenue and very happy customers.
Their product helps kids with serious recurring nightmares have a peaceful night’s sleep. Which is great for parents too.

87.
It works by identifying periods in your child’s sleep that are too deep and where extremely disturbing nightmares can occur.
By gently nudging your child out of this sleep without waking them up they can continue with a full, healthy night’s sleep

88.
And .. it works.
Average success rates are a 90% drop in night terrors.

89.
Their system requires a hardware component to deliver a gentle vibration through the child’s mattress.
They started with an existing product - an alarm clock for heavy sleepers and the hard of hearing.

90.
They rebranded the ‘vibrating puck’ and added a bluetooth control module and a sophisticated app and back end…

91.
…and a beautiful unboxing experience.

92.
Varun packing and sending out orders from Highway1

93.
Lully has sales and extremely happy customers.
These people care about results, not whether the product is custom manufactured or not.

94.
When their sales volume reaches a certain point
it will make economic sense to manufacture customer hardware.
At this point the Lully team will know exactly the product they need to make at scale.

95.
Sequential production (1 - 10)

96.
Switchmate are creating a connected home product that can be installed in seconds onto existing lightswitches.
They kicked off the project by going out to the market via Facebook Ads.
They learned about what types of people would be interested, why they wanted it and how much they would be prepared to pay.

97.
Using 3D printing and development electronics, they hand-manufactured 5 units and shipped them to their customers.

98.
During this trial they learned that many people placed their unit on the switch beside their front doorway. They often entered their homes with their hands full and
expected to be able to operate the switch without putting their bags down.
They also learned that the bluetooth radio connection between the user’s phone and the Switchmate unit was not robust enough yet.

99.
On their second round they marketed the product as a device with it’s own button that covers the existing switch.
They were able to generate more paying customers and developed a new product.

100.
Rather than take a long time to craft an elegant physical design, they spoke with these customers and made them comfortable with receiving a beta unit with a rougher
appearance than what they had purchased.

101.
The next ‘production run’ of 5 units just about to be shipped out.

102.
With this trial they learned that the magnetic attachment needed to be stronger.

103.
They also learned that people were ‘fat fingering’ the controls on their basic app
and that the purchaser wanted other family members to be able to control Switchmate as well.
They designed the new, beautiful app with very large button areas.
They enabled multi-user control.

104.
In parallel the team were doing a lot of testing on different switch brands and configurations.
Building a retro-fit product has it’s own special challenges.

105.
By the time the team went to Kickstarter they had already done 3 rounds of customer validation.

106.
Compression Kinetics are creating a wearable that acts on the body, not just measures it.

107.
Existing active compression sleeves are expensive, noisy and bulky.
These products help you recover after intense training or competition.

108.
NBAPro Cycling NFL
And they work.

109.
Compression Kinetics are developing an active compression sleeve
that is affordable and convenient.
That can be worn under clothing, while driving, while at work, etc.
This goal has driven prototype after prototype to refine the experience and technology.

110.
A video of their first working prototype.

111.
Many approaches were tried.

112.
Revision became more and more subtle until a working approach was identified.

113.
The team developed some deep new skills.

114.
Aesthetics had not been a consideration until this point.

115.
The product is in 2 parts, a compression under-sleeve and an active over-sleeve.

116.
Before

117.
After (rendering)

118.
Batch production (10 - 500)

119.
Fishbit are making a control and sensor system for aquarium enthusiasts.
To work the bugs out of their system they ran an extremely limited 30 unit beta campaign on Kickstarter.
This sold out in less than a day, providing a diverse pool of testers and some market validation, without the problems associated with a full pre-sales campaign.

120.
Their product promises to save fish’s lives and free tank owners to be more creative with their tanks.

121.
The physical product itself uses magnets and induction power to avoid running leads inside the tank.

122.
They needed to manufacture 50 units and began the task by hand-making individual units to work out the many issues.

123.
Burn-in testing was key.

124.
Producing 50 units is a lot of work for a complex product so once they had a better understanding of manufacture and assembly, etc they contracted out product
enclosures and circuit board assembly.

125.
Units arriving back from their Bay-area based partners.

126.
A stack of enclosures.

127.
The first set of batch-produced, hand-assembled, hand-tested products about to ship out.

128.
The ‘reefer’ community is very supportive of the FishBit team and their product.
They are not used to fun details and unboxing experiences with other aquarium hardware companies.

129.
Thank you.