Transcript InferMed - University of Sussex

Alan Montgomery, Director
[email protected]
Objective of talk
• Describe what R & D is like in a
small IT company
• … at least in my own experience
• Ways of getting funding
• Demo of a product produced
• Differences from academic research
Topics we will cover
• Starting a company via an MBO
(Management Buyout)
• Search for a growth
• Research grants for companies
• IPR issues
• Primacy of people
• Cash flow
• Marketing high-tech
• Selling a business
• Demo of Clementine
My experience
•
•
•
•
•
•
•
•
•
66-73
72-74
74-84
84-86
M Sc. D Phil (Physics) from UoS
CAP Process Control - House
ICL Compilers - Redundancy
Software Engineering Research
Poplog Product Manager, SDL
“AI Winter”
89-99 Founded Integral Solutions Ltd.
94-99 Clementine

• 00-
Sold Company 
Director, InferMed
 …?
The Management Buyout
•
•
•
•
•
•
•
•
•
“AI Winter” - Close AI Products Divn.
Main Product was Poplog!
Maybe 60 industry/government clients
Negotiation with Sussex and SD-Scicon
Ignorance of business – enterprise agency
Business Plan
Bank loan – collateral
Ownership
Thank you SDL and UoS/COGS
Where to go next?
• Poplog is a tool, used by R&D defence,
manufacturing, academics, ….
• Extend it with a toolset of other AI tools
• Still does not solve any high-value
business problem
• Have to sell to technicians – endless
technical wrangling
• Low prices
• Can we turn it into an application, or find
some new market?
ISL Research, Phase 1
Lucky break as we hit recession!
• Government IT R&D favours SMEs
• Get 50% funding for R&D
• Invited into three projects:
– UIDE - turn Poplog into UI prototyping
tool 
– Gateway – KBS Methodology 
– ISS – semiconductor wafers 
– SMART proposal FORTRAN. Failed. 
Funding research and development
•
•
•
•
•
•
Self funding from profits!
Co-operative project with customer
Loans – where’s the collateral?
Equity finance
Corporate venturing e.g. Roche
R&D Grants
Funding - Equity Finance
• Equity finance from
– Business angel £20K-100K
– First stage investor £100K-500K
– Major investor e.g. 3i, Apax etc £1M-10M
•
•
•
•
Requires realistic market plan
Have to give up part of the company
Expensive: advisors take 10-20% of cash!
Need to have an exit strategy
• The best way if you have the right idea.
Funding – R & D Grants
• Governments/EU encourage growth and
innovation; get university brains working
with business R&D.
• Common themes:
– There is a project (sometimes within broad
areas defined by government / EU)
– Business only gets part-funding (e.g. 50%)
– Often has to be a collaboration (e.g. EU)
– Business gets the IPR; pays royalties
– Monitoring by the grant provider
• Works if right project and right partners
UIDE Project
• User Interface Design Environment
– British Maritime Technology
– Integral Solutions Ltd
– University of Sussex, COGS.
• Extend Poplog to allow rapid prototyping
/ simulation of UIs
• Problems
– No end user
– COGS design too sophisticated
Keep it simple
Control 1
Data 1
Widget
1
Data 2
Widget
2
But data flows have type and arguments.
If several data flows, what order?
So we need control ports, with type, order
UIDE Project
• Technically interesting - could
prototype a grand piano!
• Research papers on strongly typed
visual programming
• Some ideas that turned up later
• No product as such
• Result?

ISS Project
• GEC-Plessey Semiconductors, ISL,
Reading University
• Better scheduling of wafers through
wafer fabrication plant.
• Big problem, huge pay-off 5%
increase in output.
• Strong user; strong product
champion
ISS Project
• After 3 years just starting to work, no
customers
• RAs all looking for new job
• Follow-on proposal possible in Europe
• … Mortgage another £90K of our house
and hire the team
• Chivvy MPs and ministers to get Euro
project!
ISS and JESSI Faw project
•
•
•
•
•
•
•
•
•
Got next customer SGS-Thomson
Won competitive tender for Intel
Successful trials at Intel, Texas
$1M order negotiated!
Intel decided ISL too small for them to
become strategically dependent.
Intel introduced ASL
Later ISS business sold to ASL
Negotiations, due diligence, IPR
$3.5M in total, partly staged.

HiP – Teaching Company Scheme
• TCAs work for 2 years at industry to
transfer university skills.
• Academic teaching “bought out”
• DTI funds 60% of costs
• IPR in industry, royalty to university
• TCAs groomed to be high-fliers in
company.
• Excellent scheme, light proposal; high
success rate; 1 to 1 collaboration.
HiP – Teaching Company Scheme
• ISL + UoS COGS (Mike Sharples)
• Follow-on to UIDE
• Hypermedia in Poplog. World’s first
intelligent multimedia system.
• One software guy, one designer.
• Success ….ready to launch …..
HiP – Teaching Company Scheme
• Success ….ready to launch …..
• Then Tim Berners-Lee introduced a new
multi-media standard.
• Too late in project, too expensive to
change – so we canned HiP.

Clementine – DTI Smart Award
•
•
•
•
•
•
DTI Competition: small companies
1. 75% of costs up to £80K?
2. 50% of costs up to £150K?
No need to collaborate
Easy proposal form
Very competitive
Clementine – DTI Smart Award
• Based on our experience of
machine learning NN and induction,
• And on Poplog, UIDE, HiP
• Make data mining accessible to
business people. (Colin Shearer)
Clementine: a graphical language!
Clementine Productization
• EU Project with DB to run on Intel
PCs and in theory on parallel
machines
• EU Project on DM Methodology
• EU Project on mining web data
Early Clementine Users
Manufacturing
Finance
•Daimler Benz
•Ford
•British Steel
•Caterpillar
Retail
•Boots
•Reuters
•Tandy
•Nationwide
Pharmaceutical
•ICL Retail
•National Westminster •Glaxo-Wellcome
•Halfords
•Citibank
•Pfizer
•Du Pont
Government
Telcos
•Unilever
•HM Customs & Excise
•AT & T
•IRS
•Vodafone Australia
•The Home Office
•Cellnet
•DERA
•Airtouch Cellular
•Singapore Telecoms
Introducing
Clementine
ISL Sales Growth
91
655.12
4000
92
93
94
95
327
1165
97
1599
501
1000
98
2728
348
376
1492.17
1607.9
3100
195972.6 466326.6 957466.6
3452
1633363
847
901
1091
1246 1 to 9.
ISL
Sales
Years
655
846.64
901.36
1091.36
1246.4
96
779
829
3000
2000
1000
DM Sales
AI Sales
97
96
95
94
93
92
91
90
0
ar
$1000
Sales
90
Ye
Year
DM Sales
AI Sales
ISS Sale
ISS Sale
Focus on the winner
• Disposal of ISS gave money to develop
Clementine
• Opened office in Philadelphia
• … and in Singapore
• Distributors worldwide
• Sales doubled each year
• Hard to justify other businesses
• But there was one other line of
business…
The Happy Ending
• Clementine recognized as best of breed
worldwide.
• NCR adopted it, and paid for Japanese
version.
• Main competitors SAS and IBM.
• Offer from SPSS to buy the business
• My wife said “Yes!”
• Long negotiations, due diligence,
contracts, guarantees, ……….
• 04.00am 1st January 1999 we signed!

The KBS bit
• Throughout ISL’s ten years I’d been
trying to build a KBS business.
• LPE, Gateway, KADS, KACTUS.
• PC-PACK Knowledge Elicitation
Tools
• RED
• MACRO, PROMPT
RED Project
•
•
•
•
•
•
UK Safety Critical Systems Project
ICRF – Prof John Fox
ISL, Lloyds Register, Masons, QMW
Use of logic to express knowledge
Discovery of generic safety rules
Proforma language to describe formally
and enact logic-based guidelines
• Demonstrators in Asthma Management
and Ship Safety Assessment
• Led ultimately to a new IT company
Roots of InferMed
1990 UK
SCS Prog.
RED
ICRF, ISL, LR, QMW ..
Language for protocols
Proforma
1994 EU
H/C Tele
Prompt
MACRO
ICRF, ICSF, IB, ..
Padcom, EORTC, ISL
MACRO
1999.
InferMed
Proforma =>>
ArezzoTM
InferMed
Basic concepts of the language
Protocols
(processes)
Data
entry
Decisions
Clinical
actions
Scheduling
InferMed Software
Vertical
applications
Core
technology
•Hightmvalue product
Opportunities:
•Bespoke
development for
RetroGram
•Licence fee
(up toto
£Roche
2M)
•license
B2C site
HIV drug
•Fully
validated system
•Customisation
•pharmaceutical
Clinical
advice
delivered
in 3 months
sponsorship
trials
•Training
Pain Genetic
•Repeat
business
•flexible
revenue
Remote
•Maintenance
analysis
control
models
data entry
•Direct sales model
Electronic
best practice
guidelines
Generic
Internet delivery
Rapid development
Market must be ready
• Long time from innovation to profitable
market (5-15 years)
• Sometimes a standard must be present
e.g. electricity supply, IBM PC, WWW.
• Market slow to start - goes critical rapidly
• Can only introduce radical ideas if:
– need is great and
– current ideas don’t work
• Missionary selling is very expensive
Technology adoption cycle
after Geoffrey Moore
• Uptake of a new product (that requires
behaviour change)
Early adopters
Innovators
Early
Late
Majority Majority
Laggards
Adoption cycle with chasm
where ISL came from!
• Beware the chasm! (Geoffrey Moore)
Early adopters
Innovators
Early
Late
Majority Majority
Laggards
85-90 KBS business dropped into the chasm:
specialist suppliers failed, majors pulled out.
Differences from academic research
• Rarely intellectual curiosity -> a paper. Seek
marketable product or process
• Industry is secretive, patents, etc
• Academic research often alone – industry
research nearly always a team
• Usually short term, exploit in 3 years
• At mercy of managers and the economy
• Competition is other companies, not quite same
as academic rivalry
• New research usually in parallel with the
development/support of current products.
Lessons to remember
•
•
•
•
•
•
Keep it simple
Work with users
It can take 10 years
Every project needs a champion
Market solutions not technology
The technology adoption curve and the
Chasm
• Real wealth from selling the company
• It is also a matter of luck!
Thank you for listening
?
Any Questions?
[email protected]