MS4 level being good citizen -imperative- (1) (1).pdf
Delphi Berkeley 2016
1.
2. Original Idea:
Helping Patent Prosecutors
write better patents by providing
a window into the development
cycle
Current Idea:
Helping in-house IP
committees’ harvest and
evaluate patentable
opportunities within an
engineer's documentation
Harvesting Patentable Innovations
100
Interviews
Later (and a
few tears)
Week 0 Week 10
4. Matthew Ko Akilesh Bapu Prashanth Ganesh Giacomo Zacchia
Business App Development Machine Learning Research Design
Strategy/
Finance
Front End
Development
Back End
Development
Strategy/ Project
Management
Harvesting Patentable Innovations
Team
7. Harvesting Patentable Innovations
Original Idea
Taher Savliwala, Director of IP at Quixey:
Delphi would serve to help integrate patent
prosecutors into the development cycle
by allowing them to view the different
iterations of the technology and detailed
analyses on these iterations.
We took ONE piece
of validation and
took off running
9. Iteration Control
Platform
Unprecedented
window into the
Development
Process
Downstream
Patent
Prosecution &
Innovation
Identification
Broader Scope of
Innovation
Stronger IP
Position and
ability to
monetize
intangible assets
Harvesting Patentable Innovations
Original Idea
11. Harvesting Patentable Innovations
Customer Discovery
“We rely on our own- we
have lots of experience”-
Chinh H. Pham, Esq
$
Learned the entire patent
process
Distilled Actionable Insights
“Only certain kinds of
software patents can go
through after Alice”-
Darryl Smith, VMware
Customers need to need your
product
12. Harvesting Patentable Innovations
Patent Process
Inventor
works on
Invention
Idea is realized
to be novel in
harvest session
Inventor
realizes novelty
Discouraged
Inventor fails to
patent
Engineer files
patent
disclosure form
Patent
Committee
evaluates
PC does not file
unworthy patent
PC files worthy
patent
PC fails to file
worthy patent
PC files
unworthy patent
No one realizes
patent potential
PC bring in
outside counsel
OC
communicates
with inventor
OC writes
overly broad
patent
OC writes good
patent
OC writes
overly specific
patent
PC miss strategic
patent oppurtunity
Discouraged
inventor now
dislikes filing
Discovery Filing
14. Harvesting Patentable Innovations
Week 3 - Target Customers
87
“We get a lot of invention disclosures, and it takes time to go through
each one and see if we want to pursue it, or justify why we are not”
- Darryl Smith
Primary Concerns
Discovery Objective Criteria Tracking
18. Harvesting Patentable Innovations
What We Learned - Integration
Dynamically scrape
documentation to
extract opportunities
Analyze submitted
Invention
Disclosures,
automatically
generate ID
Build docketing
system for
invention
disclosures
We had initially prioritized seamlessness for the engineer at the
expense of quality for the IP committee
19. Novelty,
Detectability, Non-
Obviousness, Value
to Business
Harvesting Patentable Innovations
What We Learned
Novelty,
Detectability,
Non-obviousness
Novelty, patent
similarity
Patentability
We tried to test many different value propositions, instead of
testing them one at a time
25. Harvesting Patentable Innovations
Customer Relationships - GET
Get to Website
Sign-up to see Demo Trial Run Individual (Limited)
Subscription Enterprise
Subscription
Direct
$6 CPC
50% of IP committee
members acquired
through Direct
Supplementary:WebPrimary:DirectSale
Earned
$0 CPC
50% of IP
committee
members acquired
through Earned
Salary:
($100k)
(~$48/hr)
Look at product
features and case
studies to gauge
credibility
P/F
P/F: IP committee
members view case
studies as signs of
credibility
Ex 80% of IP
committee members
on website will click
case studies before
P/F:IP committee
members view social
proof as signs of
credibility
Ex:90% of IP
committee members
will click social proof
Sign up for a demo
video of our product
as well as sample
reports from case
studies
P/F: IP committee
members are willing to
exchange their email for
demo video and a
sample results page
Ex. 60% of IP
committee members
who click social proof or
case studies will give us
their email for a demo
video and sample report
of our case study
Agree to phone call and a
personalized demo to test
their own documents
through Delphi
P/F: Users who sign up for
a demo will agree to a F2F
EX. 80% of IP committee
members agree to a phone
call
P/F: Users who requested
demos will want to test their
own documents through
Delphi
Ex:50% of IP committee
members want to test their
own documents through our
system
Pay for a limited
monthly
subscription to use
Delphi
We believe that
50% of IP
committee
members will pay
$499 for a limited
monthly
subscription
Pay for a fully
customized
Enterprise
Subscription for
entire IP committee
3% of IP committee
members want a
fully customized
enterprise
subscription for their
entire team
Set up Call
26. Harvesting Patentable Innovations
Customer Relationships - Keep & Grow
1. Interaction
a. Customer Check-in Emails
b. Customer Service Line
c. Customer Feedback Channels
2. Product Improvements
a. Software Updates (up to date
patent corpus)
b. Improve Algorithms
c. Added Functionalities
Keep Metrics:
● Changes in # of Searches
● Changes in # of Prior art Requests
● Open rates for check-in emails
● Service Interaction Count (before and
after added functionalities)
Cross-Sell
Up-Sell
Next-Sell
Allow Users reaching
subscription limits to
purchase:
Extractions Separately
Searches Separately
Prior Art Seperately
Allow Users reaching
subscription limits to
purchase unlimited
monthly subscription
Allow Users to
purchase year long
or multi year long
unlimited
subscriptions at
discounted rate
Referral Rewards Program
Users who referrals
amount to a
subscription will be
awarded
extraction/search/prio
r art credit
Grow Metrics:
● Track what type of
award customer
chooses (change
extraction/search/prior
art amounts to gauge
utility to users)
● Track % of
Subscriptions coming
from referrals
● % of users
reaching
subscription limits
but need not
enough to
purchase
unlimited
● Type of Users
needing beyond
subscription limit
● % of users
reaching
subscription
limits with need
enough to
purchase
unlimited
● Type of Users
● % of users
wanting
extended
contracts
● Type of Users
wanted
unlimited
27. Harvesting Patentable Innovations
Get to Website
Trial Run Individual (Limited)
Subscription
Enterprise
Subscription
Direct
$6 CPC
50% of customers
acquired through
Direct
WebDirectSale(InterviewsasProxy)
Earned
$0 CPC
50% of customers
acquired through
Earned
Salary:
($100k)
(~$48/hr)
$48/hour salary w/
45 min (¾ hour)
engagement
= ($48*.75)
=$36/engagement
$36/engagement with
31% of engagements
converted for trials
($36/.31)=
$116.3/converted
engagement to trial
$116.3/converted
engagement, with 50%
willing to purchase limited
subscription
($116.3/.5) =
$232.6 per individual
subscription sale
$232.6/individual
subscription, w/ 3%
conversion to enterprise sale
($232.6/.03)=
$7,753/enterprise sale
Avg. CPC (Direct +
Earned)
= ($6*50%)+($0*50%) =
$3
Loss Rate= 50%
Demo Rate= 50%
Cost Per Web
Engagement= ($3/.5/.5)
= $12
$12 premium for
Web Engagement
Lead + 45 min (¾
hour) engagement
= $12 + ($48*.75)
= $48/engagement
$48/engagement with
31% of engagements
converted.
= ($48/31%)
= $155/converted
engagement
$155/converted
engagement, with 50%
willing to purchase limited
subscription
= ($116.3/.5)
= $310 / individual
subscription sale
$310/individual subscription,
w/ 3% conversion to
enterprise sale
=($310/3%)
= $10,335/enterprise sale
Sign up to see
Demo Video Show Demo
Customer Relationships -GET
Experiment Results:
31% of engagements wanted a personalized trial
32. Harvesting Patentable Innovations
Critical Activities and Partners
Delphi
University Tech
Transfer
Offices First
Adopter ->
POC
Academic
Journal
Databases
USPTO
Amazon
AWS
IP Blogs &
Newsletters
Non-CoinOperatedCoinOperated
Delphi as a free service
Validation in the form of POC (DATA / Better Models)
Increase sales from funneling customers through prior art results
Access to non-patent literature (DATA / Better Models)
A better practice to implement (they already release best practices)
Validation in the form of best practices (DATA / Better Models)
Cash
Servers
Cash or Content
Publicity/Validation/Social Proof
33. Harvesting Patentable Innovations
Operating and Funding Timeline
2017 2018 2019 2020
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
Operating
Milestones
Industry Milestones
$ 1m
$ 1.5m
$2m
$3m
$4m
$5m
List industry we are in:
Electrical / Software
Move into Consumer
Product/Electronics
Paid POC
- Uni
USPTO Pilot
Ptnt Srch + Full
research corpus
Eval. Beta
Launch
Eval
Launch
Move into Mechanical &
Information Services
Fundraising
Begin Mining for Integration
Integration
Beta
Validation +
Data
Commercial KPI’s SELL SELL SELLPOC
Integration
Launch
Begin Mining for Extraction
Extraction
Beta
Extractio
Launch
34. Harvesting Patentable Innovations
Investment Readiness
Worst Case Best Case
Model produces prior art at a lower level than
a professional search firm.
Predict patentability as determined by USPTO
at least 90% of the time.
Very cheap platform for start-up’s
to generate prior art
Model produces prior art at the same level, if
not better that a manual search.
Predict patentability as determined by the
USPTO at least 90% of the time
Standardized metric across the patent industry
Investment Ready:
One more step to derisk before we scale
Investment Ready:
Use funding to get NPL, scale to get 5 POC
36. Week 1
Went in thinking:
our customers were outhouse counsel that wanted a better way to communicate with in house engineers for patent prosecution
Outhouse engineers write lower quality patents because they do not have the same domain experience as inhouse counsel
Lawyers would be open to talking to us
We knew what Lawyers pain points were
Found out:
We found how that we actually don't know anything about the patent processes
Misunderstood the dynamics of the patent lawyers workflow
Misunderstood the relationship between in house vs outhouse counsel
Misunderstood the pain points of the patent process (overgeneralized the pain point, communication si too broad)
made the assumption that outsourced lawyers write lower quality patents because they do not have the domain expertise to de-engineer
sophisticated inventions and identify innovation in that space.
LAWYERS WERE HORRIBLE CUSTOMERS
Moving forward:
STOP BUILDING AND LEARN ABOUT OUR CUSTOMERS (or really figure out who our customers were)
Figure out the intricacies of our value proposition
37. Week 2Went in thinking:
Our value proposition was to help engineers write better parents by increasing transparency and communication
Customer Discovery: Use our interviews to understand our customers and their pain-points instead of assert assumptions about their work
and sounding stupid
Signals from last week:
Explore in house counsel as our target customer
Focus on Prior Art instead of communication
Explore patent litigation as well as prosecution - firms are more desperate in litigation in terms of prior art
Understand patent processes within different industries
Found out:
Dont sell to out house counsel
Communication was not a huge issue - they only made a few interactions - they actually wanted this because there was a trade off between
willful infringement and less interactions
Friction through stage gates for in-house counsels,
There is a lot of effort that goes into the initial invention disclosure form
We learned a lot this week (patent prosecution cycle) with this higher understanding we were ready to make hypothesis
Make value propositions that can be disproved
Dont disprove hypothesis after one interview
We were still focusing a bit too much on lawyers... we kept asking mainly technological question... started to get an inkling that this would be useful
for in-house counsels (Steven Horowitz) -- we started nailing down the steps of the invention disclosure process -- We added additional customer
segments -- We were still shooting in the dark as far as the specific person within the company -- We kept interviewing people involved in litigation --
Moving forward we started focusing on conceptualizing of the problems in terms of stagegates and figuring out where we were going to focus in -- we
started being able to build better mental models -- Started to realize taht the disconect between patent prosecutors and engineers was rarely
significant -- made value props that couldn't be disproved (Weinstein) -- started developing multiple strands of potential value props -- we still were
not focusing in (We start focusing in when we make our first real demo) -- communication get's shit on (Chinh H. Pham) -- get an inkling that startups
and individual inventors may be a customer -- we were making bold statements after one -- started realizing that we needed to understand the
nuances of customers in terms of industry segments
38. Week 3
Went in thinking:
Made hypothesis about patent process - engineer recognition -> invention disclosure -> Patent committee evaluation -> outhouse
counsel -> outhouse/engineer interactions -> patent filing (provisional/nonprovisional -> patent
Made hypotheses about what are the current solutions patents are harvested (ID forms/ harvesting sessions)
Made hypotheses about large companies having a CAPACITY issue
Patent counsels miss opportunities
Made hypothesis about incentive systems and how well they work
Start ups may be a good customer
Quality of documentation is good enough to act as inputs for our model
Found out:
Make value propositions that can be disproved. Hypothesis about patent process was confirmed. Current solutions were confirmed
Large companies have a capacity issue (lack of depth), they have the wallet to not do as much due diligence in their patents
Small companies do not really have a large communication issue between patent agents and engineers since the firm is smaller
Small companies spend 30-45k per patent and they spend around 20 hours back and forth, A LOT MORE TIME AND ARE A LOT MORE TACTICAL
WITH PATENTS
Moving forward:
Decided we wanted to pursue in house counsel at large corporations (fix capacity issue)
Started to really see that startups were different than big companies -- there is a capacity issue (we saw capacity as not missing oppurtunities... we
were still conceptualizing of the value prop in terms of value to the business not speicific people within the patent committee) versus a finding very
important patent issue -- We start to conceptualize about the theories of searching further... confirm steven horowitz findings... the spectrum of never
search to always search -- we start to really confirm communication is not that big of an issue -- we start to hear a bit about competitors --
LECORPIO -- Darren Cooke, we really started to realize the capacity distribution -- he really liked the idea, but then was confused when we assumed
he handled more than 6 applications a month -- We started learning about the different engineer archetypes -- some people love to patent, some
people hate to patent -- Further honed in on the changing incentive systems -- Started seeing other drivers of patenting -- name recognition, internal
company politics and pride (the human aspect starts to come into play) -- Start to focus on the criteria for patenting as far as developing a score -- we
hear of another potential competitor -- INNOGRAPHY -- We start to explore the different types of documentation we are going to look at and plug in
39. Week 4
Went in thinking:
Made a strict interview doc
Made hypothesis that confined our customer segments to hardware because of patent frequency and documentation quality
Started defining what qualities were important to seeing whether a patentable opporutnitiy is pursuable or not: Value to business, novelty score,
detectability
From last week after hearing that Lecorpio and Innography, we hypothesized that patent committees used some sort of enterprise software
Made hypotheses around security (Cloud based vs behind firewall)
Made hypothesis that 25% of the engineer’s incentive could go to Delphi for discovering patentable opportunities
What we found:
Smaller start-ups bootstrap as much of the patent legwork as they can because it’s a large investment Startups and early stage companies
sometimes prefer trade secrets.
Engineer documentation is not organized in a way that is conducive to extract patentability. Patent approval rate for engineers goes up over time as
they get better at writing patents and evaluating what to patent
We were wrong in assuming that startups make breakthroughs all the time and constantly need to be on the watch to patent something. They have
one product usually and so they combine as much as possible into one patent to save costs.
Filing provisional patents faster could be helpful. Large firms use patent troll insurance companies to fill the holes in their patent coverage, however,
small companies may not have the luxury to buy into patent insurance portfolios.
There isn't as much of a push for IP efficiency or any worry in general to have the best IP as we thought there was, even at the most active patent
filers like IBM and Apple.
40. Week 5
Went in thinking:
Defined our Get, Keep and Grow funnels as well as our CAC to model our business plan
Tested hypothesis that customers (both engineers and IP committee members) would be willing to send us
their personal documentation to run through Delphi, to see what the result of the novelty, detectability, and business value
scores are
Tested hypothesis that customers are interested in using Delphi to process unstructured technical
documentation and extract patentable opportunities
Started to narrow down which qualities of the scores we hypothesized could be technologically feasible
Began exploring how we could develop trust in our algorithm with the use of these scores
What we found:
Customer Relationships
Customers find all of the different metrics we were envisioning to be valuable, but only some to be technologically feasible
Comfort with the automatic extraction varies among our customer segments, with committee members being most
accepting and engineers being least
41. Week 6
Went in thinking:
Industry segmentation- we thought that we would just go for microelectronics, computer architecture and
semiconductors
Showed the demonstration of our MVP to our interviewees
Tested pricing for our services including invention disclosure generation, prior art search, and seat
subscription
What we found:
Revenue Streams
We found it was better to segment by industry… we focused on two groups to begin… and we're going to
expand into it later --- different
We took of Biotech, but then were told to keep Biotech on the map → they might pay premium since it is their
lifeblood
43. Week 7
Went in thinking:
Thinking pilot programs
KPIs
Backtesting our machine
What we found:
Key Activities are verbs
Started finding out more about the politics of universities
Started to realize that we should organize by project not individual person
Convincing people to switch their operations and onboarding is a huge activity
NEEDED TO REALIZE THAT activities around convincing people of brand were key… but we didn’t find this until after key
resources
Enterprise sales were key activity
44. Week 8
Went in thinking:
Universities will be primary customers
What we found:
Key resources are nouns
Resources are partners
Universities are not customers
BRAND is a key resource
Next Week:
→ Needed to look at other companies that have made a brand off a number FAIR ISAAC
We need to sell the vision, not the most palatable version
45. Week 9
Went in thinking:
We thought any legitimate university would be a good partner
We thought that costs was largely a numbers game of maximizing runway and staying afloat
What we found:
Costs - funding is to get to key milestones, not to sustain employees
Universities we partner with most be similar to our actual customers
Math vs Judgement → Not everything is entirely math e.g. housing… Be near customers
We want to have POC funders help us build a product THEY would buy
46. Customer Segments
- Independent
Researchers/Inventors
Customer Segments
Hardware Engineers/Inventors at BioTech,Micro
Electronics, Digital Signal Processing and Computer
Architecture companies (over 100 patent filings a
year)
Customer Segments: General Counsels and Vice
Presidents of IP at BioTech, Micro Electronics,
Digital Signal Processing, and Computer
Architecture (over 100 patents per year)
Principle Investigators & Business Directors at
University Tech Transfer Offices (over 100 patent
filings a year) - to be further segmented
Academic and Industry Research Journals in patent
heavy fields - to be further segmented
Business Model Canvas: With Customer Segments Changed
Everything...
47. Harvesting Patentable Innovations
Original Idea
Topic Identification Understand
Derived
Innovations
See Litigation
History
See Similar
Patents
News & Trends
Gauge Drafted
Patent Strength
49. Harvesting Patentable Innovations
Week 3 - Target Customers
Understanding Pain Points of Harvesting Patents as opposed to simply
Prosecuting Patents
“The engineer can file
an ID, sometimes the
manager can push for
one..”- Steven
Horowitz
“We use Anaqua- it
works pretty well for
us”- Kevin Brown
“It is a small
company...so they
(engineers) come to us
when they think they
have a new idea”-
Krishna Mehta
Delete… Kanu didn’t
understand why it was
here, and neither did I
to be honest
50. Harvesting Patentable Innovations
Week 3 - Target Customers
3 Concerns of Patent Committees
The ability to discover
patent opportunities
without influence of
engineer’s opinion or
the time lag of harvest
sessions
The ability to sort
patent disclosure
forms by relevant
criteria (novelty, value
to business,
detectability)
The ability to have an
objective criteria for
evaluating invention
disclosures
“We get a lot of invention disclosures, and it
takes time to go through each one and see
if we want to pursue it, or justify why we are
not” - Darryl Smith
Too specific… how
you said it… after
customers and who
they were… VERY
HIGH LEVEL… there
was three things to
test
Discover opportunities,
Number, track
currently
51. Harvesting Patentable Innovations
Lawyers are TERRIBLE
customers
“We rely on our own- we
have lots of experience”-
Chinh H. Pham, Esq
Learned the
entire patent
process
Key Players in the
Patent Prosecution
process and their
main functions
Patents behave differently
in different industries
“Only certain kinds of
software patents can go
through after Alice”- Darryl
Smith, VMware
Patent Processes can be different
in different companies depending on
size, strategy, etc.
“We built an Invention Disclosure
Filing system at Yahoo”- Steven
Horowitz, Ovidian
Customer Discovery
53. Post MVP v1:
● Allowed us to add
startups to customer
segments
● Allowed us to form
hypotheses about
customer relationships
● Allowed us to narrow
down value propositions
based on what stood out
to customers in our MVP
Don’t need it - just
SAY IT -- delete
55. “Documentation...is distributed
among many formats. It contains
all kinds of extraneous
information that don’t
matter for patents”- Tanj
Bennet
“We have a great filing
system we have been
developing for years...we
are not interested in
switching- Luis Mejia
“Lots of things determine
patentability...novelty,
detectability...value to
business is important..”-
Luv Kothari
“If we don’t
know our value
to business, we
shouldn’t be
doing this”-
Darryl Smith
“Non-obviousness is very
difficult to detect, I don’t
think a computer
could do it”- Dov
Rosenfeld
Notas do Editor
On left hand side --
On left hand side --
Kanu said we can just say this… we can keep it for emphasis, but we should be cutting at this point
SNIPE: DEEP DIVE INTO THE PATENT PROCESS
The pain of patent discovery falls on the entire patent committee (CTO, GC, VP of IP, Paralegals, Patent agents, etc.), not just VP of IP.
Friction between stage gates leads to problems at big companies to the point where many companies build internal infrastructure for filing patents.
But value prop changes based on whether they file patents ad-hoc (friction reduction), or have a standardized process (safety net)
Two main ideas: value to engineer and value to IP committee member
Value to engineer progressed from dynamic scraping (documentation issues), docketing system (would get the info, wouldn’t adopt), integrate into existing system
Value to IP committee member progressed from simple patentability score (didn’t really take into account all aspects of patentability), all aspects of patentability (could not compute), to what we could compute
Two main ideas: value to engineer and value to IP committee member
Value to engineer progressed from dynamic scraping (documentation issues), docketing system (would get the info, wouldn’t adopt), integrate into existing system
Value to IP committee member progressed from simple patentability score (didn’t really take into account all aspects of patentability), all aspects of patentability (could not compute), to what we could compute
We refined our MVP and initially we got feedback like how do you come up with this score, how are you doing this?
So we started out by putting in only what we knew for sure we could do at the moment. What you see here is believable because we tell how we calculated the score and named it “Patent similarity score”
After this feedback, we refocused on what was VALUABLE to our customers instead of what was BELIEVABLE
Made a switch in story-telling from product back to company
Need to figure out how to bring the audience to realize this big shift
The pain of patent discovery falls on the entire patent committee (CTO, GC, VP of IP, Paralegals, Patent agents, etc.), not just VP of IP.
Small companies lack the bureaucracy needed for this to be seen as a problem (close integration of patent agents and engineers)
Friction between stage gates leads to problems at big companies to the point where many companies build internal infrastructure for filing patents.
But value prop changes based on whether they file patents ad-hoc (friction reduction), or have a standardized process (safety net)
The pain of patent discovery falls on the entire patent committee (CTO, GC, VP of IP, Paralegals, Patent agents, etc.), not just VP of IP.
Friction between stage gates leads to problems at big companies to the point where many companies build internal infrastructure for filing patents.
But value prop changes based on whether they file patents ad-hoc (friction reduction), or have a standardized process (safety net)
We had features like this which even highlighted exactly what our algorithm picked up to shed light on how we got to our results. After doing this our customers started losing excitement and we were told we’re thinking too much like engineers.