3. 1. about me
2. about Ballard (where I work)
3. various scenarios
4. P A G E 4
about me
UBC chemical engineer
(1992 to, um, 1999)
13 years at Ballard
occasional playwright
avid dogsledder
5. P A G E 5
“fuel cells are like batteries, with an external fuel pack”
“an electrochemical analogue to the combustion engine”
chemical, not electrochemical
6. P A G E 6
about Ballard
Burnaby-based fuel cell maker
we used to do cars
• late 90’s – Daimler and Ford bought stakes
• late 00’s – automotive half spun out
now: everything except cars
7. P A G E 7
about Ballard
…a stack for every size and setting…
System
Integrators /
OEMs
Downstream
Customers
Backup Power Supplemental
Power
Material
Handling
Bus Distributed
Generation
1–10 kW 5–25 kW 100 kW MWOutput:
8. P A G E 8
yes, Ballard was once
a tech bubble darling
(more on that in a few slides)
“I’m the king of the world!!”
9. P A G E 9
why I’m there
New mining companies
usually go through:
- a speculative phase
(stock peaks)
- a development phase
(stock sags)
- a production phase
(stock recovers)
10. P A G E 10
why I’m there
The same applies for tech companies.
New mining companies
usually go through:
- a speculative phase
(stock peaks)
- a development phase
(stock sags)
- a production phase
(stock recovers)
11. P A G E 11
why I’m there
Ballard had its
speculative phase.
We’re finishing our
development phase.
(It took a long time)
Production is gradually
ramping up. (At last!)
12. P A G E 12
why I’m there
not many people get to work
at a leading company,
in an industry,
at this inflection point
the fuel cell sector is big enough
that it won’t disappear – and
small enough that I can still
make my mark
admittedly, I’m biased –
I’ve been working on this for 13 years
13. P A G E 13
Scenarios: UDo Research
Q: what are the right metrics to measure?
[MK - I don’t know, but the wrong ones can hurt you]
You’ll hit
what you aim for,
but what yo u aim fo r
might not be
what yo u want!
14. P A G E 14
Scenarios: UDo Research
Real-world examples of bad metrics:
Topic Bad Metric Effect
Call Centers time-per-call company reps won’t spend time
to resolve customer problems
Dep’t Store sign-up quota
for store card
customers get annoyed
CEO bonus stock options CEO pumps up stock price but
weakens company, then leaves
You’ll hit
what you aim for,
but what yo u aim fo r
might not be
what yo u want!
15. P A G E 15
Scenarios: UDo Research
Real-world examples of better metrics:
Topic Better Metrics Effect
Call Centers rings before
pickup
prompt service, happy clients
(Southwest Airlines: 3 rings)
Dep’t Store $ sales per ft2
per department
enlarge departments which
bring in the most sales
(Wal-Mart)
CEO bonus # options based
on operating
targets
CEO focuses on operations,
not wild & woolly schemes
(CN: $ cost per tonne-km)
You’ll hit
what you aim for,
but what yo u aim fo r
might not be
what yo u want!
16. P A G E 16
Scenarios: UDo Research
What metrics would be best for UDo?
…I don’t pretend to know
Some possibilities:
total membership size of community
user time-on-site level of engagement
% of heavy users target market
heavy user time-on-site target market satisfaction
$ revenue per user path to profit, with scale-up
% who upgrade “freemium” models
17. P A G E 17
Scenarios: HMI
Q: how to verify the product design is adequate?
Lab ≠ real world !
18. P A G E 18
Scenarios: HMI
Lab ≠ real world !
Paraphrasing a friend working on oil sands tailings cleanup:
“you can make anything work at room temperature, indoors.
But when it’s -40°C and snowing…”
19. P A G E 19
Scenarios: HMI
Lab ≠ real world !
In the 1980’s a Japanese carmaker had gear shifter problems in the US
(but not Japan). Crumbs from burgers would jam the shifter – but since
the Japanese didn’t eat in their cars back then, they didn’t test for this!
20. P A G E 20
Scenarios: HMI
Lab ≠ real world !
Q: how to verify the product design is adequate?
…here, I have some ideas
Could test voice-recognition against:
different accents a few years ago the Telus virtual
assistant couldn’t understand east
Asian accents many complaints,
e.g. from my wife!
people chewing gum
background noise e.g. near construction
multiple voices e.g. kids talking in back seat
& others…
21. P A G E 21
Scenarios: Cryotonics
Q: how to maximize learnings from first product run?
he who learns fastest,
often wins
Toyota?
BMW?
GM?
22. P A G E 22
Scenarios: Cryotonics
Q: how to maximize learnings from first product run?
some suggestions…
he who learns
fastest, often wins
Check tolerances (what can you get away with?)
IKEA furniture is cheap, because it’s
made of sawdust, glue, and a topcoat.
(Genius!)
They probably test different glue/sawdust
ratios to see what range is acceptable
(40-60% glue? 37-82% glue?)
Note: ratios might be different,
depending on the type of sawdust!
(oak, pine…)
23. P A G E 23
Scenarios: Cryotonics
Q: how to maximize learnings from first product run?
some suggestions…
he who learns
fastest, often wins
The more variability
you can tolerate in
incoming materials,
the better.
But reject parts are
expensive! Find where
this boundary is, add a
margin of safety, and
then avoid it!
24. P A G E 24
Scenarios: Cryotonics
Q: how to maximize learnings from first product run?
some suggestions…
he who learns
fastest, often wins
Could Cryotonics perhaps test…?
RhInSb ratios how exact does the ratio have to be?
RhInSb batches how repeatable is the alloying process?
wound core quality can Cryotonics products work with
cheap wound cores, or are expensive
ones needed?
in-house testing varying the above (or other factors)
measure how well each sensor works,
to see how much effect each factor has,
on functionality.
25. P A G E 25
Scenarios: Bioplastics
Q: what went wrong with first scale-up run?
[MK – I was running out of ideas… ]
26. P A G E 26
Scenarios: Bioplastics
Q: what went wrong with first scale-up run?
conditions always about
the same, everywhere!
temperature gradients and uneven mixing
can really affect chemical processes!
I’m told DuPont had a lot of trouble scaling up Kevlar™…
27. P A G E 27
Scenarios: Bioplastics
Q: what went wrong with first scale-up run?
…may be unclear to,
or interpreted differently by,
production technicians.
Instructions that seem clear and comprehensive
to experts in a particular field…
28. P A G E 28
Scenarios: Bioplastics
Q: what went wrong with first scale-up run?
some example tools…
Ishikawa (“fishbone”)
diagram is a useful way
to problem-solve until
reaching root cause of
an issue.
29. P A G E 29
Scenarios: Bioplastics
Q: what went wrong with first scale-up run?
some example tools…
5 Why’s is self-explanatory:
Why was the product off?
(Incomplete polymerization)
Why was polymerization
incomplete?
(Instructions not followed)
Why not?
(Instructions unclear)
Why unclear?
(Writer never checked clarity
with operator)
Why not?
(Didn’t think he had to)
30. P A G E 30
Scenarios: Bioplastics
Q: what went wrong with first scale-up run?
some example tools…
P-diagram (parametric)
captures what affects the
product.
Noise factors =
variation in incoming material,
between operators, etc.
Control factors =
what you monitor, to keep the
finished product OK
31. P A G E 31
Scenarios: Bioplastics
Q: what went wrong with first scale-up run?
some possible ideas…
Could Bioplastics look at…?
lab-scale variability could the big batch just be a low run
within the same distribution curve?
prior runs Bioplastics probably did test runs in
the scaled-up process; how do those
compare?
confirming “knowns” as with the big reactor case, are we
making inappropriate assumptions?
(e.g. local temps ≠ mean temps!)
32. P A G E 32
Scenarios: PNA
Q: to outsource, or not to outsource, that is the question…
33. P A G E 33
Scenarios: PNA
Q: to outsource, or not to outsource, that is the question…
Generic advantages to outsourcing / JV’ing:
you can focus on your core competency
many companies “deworsify” by trying to do R&D and
manufacturing and sales and other stuff
manages cash
no need to buy equipment, hire manufacturing folks, lease more
space, etc.
34. P A G E 34
Scenarios: PNA
Q: to outsource, or not to outsource, that is the question…
Generic advantages to going it alone:
slow the spread of trade secrets / know-how
contractors sometimes become competitors!
self-reliant on quality
better integration (hopefully!) from R&D to Production
35. P A G E 35
Scenarios: PNA
Q: oh wait, this was about batch sizing, wasn’t it?
36. P A G E 36
Scenarios: PNA
Q: oh wait, this was about batch sizing, wasn’t it?
Advantages to small batches:
stats come fast can quickly do several runs to learn
variability of process
losses smaller each bad batch is less expensive
(can be important, early on!)
redundancies equipment malfunctions / breakdowns
less catastrophic
37. P A G E 37
Scenarios: PNA
Q: oh wait, this was about batch sizing, wasn’t it?
Advantages to big batches:
much, much cheaper …as long as you’re not troubleshooting
scale-up all the time!
go-forward flexibility increases throughput capacity faster
(so you don’t turn down orders)
38. P A G E 38
Scenarios: PNA
Q: oh wait, this was about batch sizing, wasn’t it?
some possible ideas…
Perhaps PNA could consider:
how similar is this to other industry processes?
if similar processes have been scaled up before, maybe PNA
goes big and trust that consultants can help with issues
could process time be reduced?
shortening your longest step (often separation) allows you to
increase throughput without spending capital
how comfy is PNA with the existing process anyways?
PNA may want to wait until it has proven to itself that it has
fully mastered the small batch size, before scaling up
39. P A G E 39
End.
Matthew Klippenstein
matthew.klippenstein@ballard.com
http://ca.linkedin.com/in/matthewklippenstein