Over the course of 9 weeks, the team went through multiple pivots to find product-market fit:
1. They started with facial recognition for smart door locks but learned their solution did not fully address parent concerns.
2. Their second pivot was to enterprise access control but customers said tailgating was a bigger problem.
3. Recognizing tailgating risks required surveillance, they pivoted again to surveillance software.
4. After failing to find problem-team fit, they did a complete restart and developed an AI assistant for streamlining similar work tasks. Customer feedback was positive and they identified a potential market of 500k ML professionals.
19. What we learned: We don't actually solve the problem for parents
“There are so many risks that I worry
about (for my son getting home safely)
that making one part of the chain safer
might not be enough.”
— Ana Ramirez, Parent
Week 1Week 1
24. What we learned: One challenge for enterprise
access control is preventing tailgating
“The number one problem I’d like to see
a solution for is tailgating.”
— Jay Kohn, Director of key card services
Stanford
Weeks 2-3 Week 2 Week 3
25. “Tailgating is the top problem or a
very important problem for me.”
Microsoft: Brian Tuskan, Chief Security Officer; Joe
Fairchild, Lead for Real Estate Technology
GRAIL: Lisa Peloquin, Facilities Coordinator; Arlito Legaspi,
Facilities Manager
Kilroy Realty: Rob Paratte, VP Business Development;
Chris Johannsen, Director of Security and Safety
Stanford: Jay Kohn, Director of Card Services; Fred
Vasquez, Building Access and Security Manager
Weeks 4-6 Week 3
30. We solved tailgating, will you buy
our product?
“No.” — multiple customers
Weeks 4-6 Week 4
What we learned:
Our solution detects tailgating, but doesn't
address the risk they represent after
getting in…
That requires surveillance
31. Perimeter Access Volume Surveillance
Weeks 4-6 Week 5
2nd Pivot: Access Control to Surveillance
47. Existing tools: Knowledge is archived (i.e., lost forever)
Surveillance
project
Archive
Week 7 Week 7
48. Existing tools: Knowledge is archived (i.e., lost forever)
Surveillance
project
Week 7 Week 7
New project started
from scratch!
Same old mistakes,
redundant work, etc.
50. New task
Week 7 Week 8
What happens:
AI returns most
relevant tasks from
history
MVP #3
51. New task
Week 7 Week 8
What happens:
AI returns most
relevant tasks from
history
MVP #3
The user gets:
1. A warm start on their
task
2. Know who to ask for
help
52. Customers want the product!
“If you can give my
engineers boilerplate
starter code, I would buy
it immediately.”
Eric Xiao, PM, Facebook
Week 8
“If you can surface
relevant documentation,
I would buy it in a
heartbeat.”
Kevin Bao, EM, Karat
“I would use this, and I
would pay for it today.”
Aref Erfani, Enterprise
Architect, DC Water
53. Customer Segments
• ML researchers
- Industry research
groups
- Academic research
groups
- PhD students
• ML engineers
- AI/DL/RL engineer
- SW engineer
• Data scientists
- Data engineers
- Data analysts
Week 9
Start where
we understand
the problems
54. Customer Segments
• ML researchers
- Industry research
groups
- Academic research
groups
- PhD students
• ML engineers
- AI/DL/RL engineer
- SW engineer
• Data scientists
- Data engineers
- Data analysts
Week 9
Value Propositions
• Dataset access:
Standardize
preparation of common
datasets
“I have 10 copies of the
MNIST dataset on my
laptop.”
Kristy Choi, CS PhD,
Stanford
“...version control for
data would be cool.
Right now I just save
different versions.”
Sherrie Wang, ICME
PhD, Stanford
55. Customer Segments
• ML researchers
- Industry research
groups
- Academic research
groups
- PhD students
• ML engineers
- AI/DL/RL engineer
- SW engineer
• Data scientists
- Data engineers
- Data analysts
Week 9
Value Propositions
• Dataset access:
Standardize
preparation of common
datasets
• Data processing:
Templates for data
pipelines
“...you feel bad that your
whole day was spent
figuring out how to load
and preprocess your
data. And it happens
every time you have a
new dataset.”
Aditya Grover, CS PhD,
Stanford
56. Customer Segments
• ML researchers
- Industry research
groups
- Academic research
groups
- PhD students
• ML engineers
- AI/DL/RL engineer
- SW engineer
• Data scientists
- Data engineers
- Data analysts
Week 9
Value Propositions
• Dataset access:
Standardize
preparation of common
datasets
• Data processing:
Templates for data
pipelines
• Improved
communication:
Increase transparency
and reduce barriers
“I would love to talk to
other researchers who
are working in the
same area, even if
they’re not well-known.”
Rui Shu, CS PhD,
Stanford
“It can be hard to ask
for help if you’re not
sure who can answer
your question.”
Geet Sethi, CS PhD,
Stanford
58. Week 9
A proven business model… freemium Saas
$100
per user/year
x
500,000
ML + data science workers
in 2024
=
$50M ARR
59. Week 9
A proven business model… freemium Saas
$100
per user/year
x
500,000
ML engineers + data
scientists in 2024
=
$50M ARR
Source: The age of analytics: Competing in a data-driven world, McKinsey Global Institute.
60. Week 9
A proven business model… freemium Saas
$100
per user/year
x
500,000
ML engineers + data
scientists in 2024
=
$50M SAM
Source: The age of analytics: Competing in a data-driven world, McKinsey Global Institute.
61. Week 9
A proven business model… freemium Saas
$100
per user/year
x
500,000
ML engineers + data
scientists in 2024
=
$50M SAM
Source: The age of analytics: Competing in a data-driven world, McKinsey Global Institute.
Eventually:
Expand to adjacent markets
like software engineering
25M+ workers =>
$2.5B market
62. “...as I start my own
lab, I need a tool like
this.”
— Lisa Wedding, Associate
Professor, Oxford University
“I would use this
today!”
— Kristy Choi, ML PhD, Stanford
University
“Let me know when I
can beta test it.”
— Neeraja Ravi, BioE PhD,
Stanford University
Week 9
“I want to be your
first customer!”
— Surabhi Sharma, Product
operations, Twitter
Lining up beta testers
65. David Zeng
PhD, Machine learning
Jesse Zhang
PhD, Machine learning
Neal Jean
PhD, Machine learning
Shing Shing Ho
PhD, Applied physics
Elaine Ng
PhD, Bioengineering
Beacons AI
66. AI for medical
diagnostics & imaging
David Zeng
PhD, Machine learning
Jesse Zhang
PhD, Machine learning
Neal Jean
PhD, Machine learning
Shing Shing Ho
PhD, Applied physics
Elaine Ng
PhD, Bioengineering
Beacons AI
67. AI for medical
diagnostics & imaging
David Zeng
PhD, Machine learning
Jesse Zhang
PhD, Machine learning
Neal Jean
PhD, Machine learning
Shing Shing Ho
PhD, Applied physics
Elaine Ng
PhD, Bioengineering
Beacons AI
68. AI for medical
diagnostics & imaging
Summer:
Beacons AI
Fall:
Stanford postdoc
David Zeng
PhD, Machine learning
Jesse Zhang
PhD, Machine learning
Neal Jean
PhD, Machine learning
Shing Shing Ho
PhD, Applied physics
Elaine Ng
PhD, Bioengineering
Beacons AI
Come up with a better slogan
Title slide must include:
Team name
Succinct description of what your company does
# of interviews done this week
# of interviews in total
Team members (names, pictures, roles)
Market type
New market
Re-segmenting existing market as low cost
Re-segmenting existing market as niche entrant
Cloning a successful business model from another country
Hacker = Engineer
Hustler = Customer Development
Designer = Product
Picker = Visionary
Ended at: 0:50
Ended at: 1:23
Ended at: 1:23
Explain our idea so that audience knows why we bought a door
Ended at: 1:39
TODO:
1. Add enterprise and Airbnb to Customer Segments.
2. Add shortened value prop for each customer segment
*Update weekly
Each customer segment needs a matching value prop. Use a different color for each customer segment.
Order of Validation:
1. Customer Segments
2. Value Propositions
3. Channels
4. Customer Relationships
5. Revenue Streams
6. Key Activities
7. Key Resources
8. Key Partners
9. Cost Structure
Ended at: 2:20
What we thought: Parents with young children would love this
Some ways we were dumb: Which of these value props is important to working parents? Hypotheses from value props and customer segments should be linked
What we thought: Parents with young children would love this
Some ways we were dumb: Which of these value props is important to working parents? Hypotheses from value props and customer segments should be linked
Ended at: 2:43
Draft
Ended at: 4:00
Ended at: 4:00
Expected start-ups to care more about security because of IP, but they actually didn’t
Middle to Large companies very much cared about tailgating for IP reasons
Ended at: 4:25
Ended at: 5:51
Tailgating requires 1. detecting tailgating and 2. deterring tailgating (which requires knowing what's going on inside)
Ended at: 5:51
Tailgating requires 1. detecting tailgating and 2. deterring tailgating (which requires knowing what's going on inside)
Ended at: 5:51
Tailgating requires 1. detecting tailgating and 2. deterring tailgating (which requires knowing what's going on inside)
Ended at: 6:05
Ended at: 7:50
Highlight our flip from tech searching for problem to thinking directly about the problems
Ended at: 8:19
Ended at: 8:19
Ended at: 8:19
Ended at: 8:42
Ended at: 10:21
TODO: Add mentor photos
Ended at: 12:35
TODO: Elaine will edit this slide - move Shing Shing to left
Ended at: 12:15
TODO: Elaine will edit this slide - move Shing Shing to left
Ended at: 12:15
TODO: Elaine will edit this slide - move Shing Shing to left
Ended at: 12:15
TODO: Elaine will edit this slide - move Shing Shing to left
Ended at: 12:15
Ended at 11:25
Video demo of Trello system
"Know it when we see it"
Ended at 9:00
Ended at 9:00
Today: A diligent assistant
Tomorrow: A powerful collaborator
TODO:
1. Add enterprise and Airbnb to Customer Segments.
2. Add shortened value prop for each customer segment
*Update weekly
Each customer segment needs a matching value prop. Use a different color for each customer segment.
Order of Validation:
1. Customer Segments
2. Value Propositions
3. Channels
4. Customer Relationships
5. Revenue Streams
6. Key Activities
7. Key Resources
8. Key Partners
9. Cost Structure
How did we get to schools? By talking to many different customer segments
Need to explain that we dropped enterprise due to privacy concerns
They way this presents is that parents are against us, but parents are
Founder-product fit was not so good
Ended at: 7:17
Ended at: 9:18
“I’m sure I’ve written the same code many times” -George
Also add quotes from our interviews!
George: don’t really understand the image
The way you explain your value props matter: we were describing it as streamlineing repeated tasts, but people didn’t respond positively even though they described - they ddidn’t like to think of themselves as doing repeatable work. People don’t like to think of themselves - it’s very easy to blame others for the problem
What should you test over the next week?
See hypothesis validation sequence in the slide note on the Business Model Canvas slide.
Rank your hypotheses by High Impact + Low Confidence/Certainty
Using this slide:
Before running the test, complete left side boxes and leave right side boxes blank.
After running the test, leave original left side boxes untouched and complete right side boxes to compare predicted to actual
Feel free to add subsequent slides with screenshots, photos, or other visuals showing your test setup and results
Stronger close - team still friends, learned a lot, excited to keep working on it
Ended at: 11:10
Great space but big, where should we focus? Start with where we have domain expertise