Looking back going forward - LONG TERM GLOBAL GROWTH April 2020 - Baillie Gifford
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Looking back going forward LONG TERM GLOBAL GROWTH April 2020 THIS MAGAZINE IS INTENDED SOLELY FOR THE USE OF PROFESSIONAL INVESTORS AND SHOULD NOT BE RELIED UPON BY ANY OTHER PERSON. IT IS NOT INTENDED FOR USE BY RETAIL CLIENTS.
RISK FACTORS In this issue:
The views expressed in this magazine are those of the Welcome to the latest edition of Looking Back
Long Term Global Growth team and should not be Going Forward.
considered as advice or a recommendation to buy,
sell or hold a particular investment. They reflect personal As well as introducing new topics, from flying taxis
opinion and should not be taken as statements of fact to protein-targeting therapeutics, this edition picks up
LTGG • Contents
nor should any reliance be placed on them when making where we left off in LTGG XV, our recent celebration
investment decisions. of our first fifteen years.
Our lead article ‘Work in progress’ shares some
This communication was produced and approved on the
thoughts on how we deepen our understanding of the
stated date and has not been updated subsequently. It companies we invest in. We pay close attention to
represents views held at the time of writing and may not their corporate character with a particular focus on
reflect current thinking. how they balance labour issues with the demands of
superfast growth.
Potential for Profit and Loss In ‘Long-term talent’ we introduce you to Gemma
Barkhuizen and Robert Wilson, two recent additions
All investment strategies have the potential for profit
to the LTGG team.
and loss, your or your clients’ capital may be at risk.
Past performance is not a guide to future returns. In ‘AI: Learning on the job’ we examine the
transformational effects of self-teaching computers
on data-rich companies, from the West Coast of
Stock Examples
the US to the East Coast of China.
Any stock examples and images used in this article are ‘Lessons from the Sonoran Desert’ explores further the
not intended to represent recommendations to buy or sell, pioneering work of Professor Hendrik Bessembinder
neither is it implied that they will prove profitable in the at Arizona State University, whose heroic feat of
future. It is not known whether they will feature in any global number-crunching revealed extraordinary
future portfolio produced by us. Any individual examples things about the nature of company returns.
will represent only a small part of the overall portfolio and
are inserted purely to help illustrate our investment style. Finally, we share our conversations with Professor
Mike Berners-Lee who is helping us to reassess how
This magazine contains information on investments which we calculate the environmental impact of the holdings
does not constitute independent research. Accordingly, it within the LTGG portfolio.
is not subject to the protections afforded to independent We hope that you enjoy the magazine and, as ever,
research and Baillie Gifford and its staff may have dealt in would welcome any feedback. If you’d like to hear
the investments concerned. All information is sourced from more from the Long Term Global Growth team, please
Baillie Gifford & Co and is current unless otherwise stated. visit ltgg.bailliegifford.com
The images used in this magazine are for illustrative
purposes only.purposes only.
Contents
02 Work in
progress 28 Lessons from the
Sonoran Desert
Looking back going forward
Why high-growth So much from so few: what
companies are racing Hendrik Bessembinder
to do the right thing taught us
12 Long-term
talent 34 Footprints
and first steps
Swapping academia for Assessing a company’s
investment: two team carbon footprint means
members share their stories seeing the whole picture
18 Ringing
the changes 36 First principles
of flying machines
Picturing the LTGG story: The answer to urban
ten years of transformation congestion could lie in
the sky
20 AI: Learning
on the job 38 What’s on
our minds
How everything changed You read it here first:
when computers got leaves from our investors’
creative notebooks
1
Work in progress
Good labour practice is often a casualty of superfast growth. LTGG looks for evidence that a company is doing the
right thing, both for its workers and for its own long-term success.
The turbocharged performance sought by
Long Term Global Growth demands a lot
from the companies we invest in. To merit
LTGG • Work in progress
a place in the portfolio, their operational
progress must be extraordinary. But the ... We need to know if our
sheer scale and speed of this progress holdings can walk fast and
can give rise to less savoury by-products
of rapid growth. These include strained chew gum.
corporate cultures, unhappy workers and
the unwelcome attention of regulators and
media.
These and countless other datapoints
and pieces of news tell the same story: Why this broader outlook? After all, LTGG However, we know there is no perfect
however exciting the growth opportunities is not an environmental impact or ethical company. The best-case scenario is often
presented by the companies in the LTGG investing strategy. Shouldn’t we stick to a story of innovation, mistakes, lessons
portfolio, they are not immune from obsessing about growth? and adaptation.
workforce problems.
But investing over decades means that the When we visit companies, we ask what’s
Can these companies continue to improve ability (or inability) of portfolio companies special about their culture, whether it’s
business operations while addressing these to behave in a responsible way can make adaptable, and how it contributes to
critical issues? We need to know if our or break their growth and longevity. For society. Such questions force us to look far
holdings can walk fast and chew gum. long-term investors, societal and customer beyond company accounts and business
backlash can damage investment returns. models to inform ourselves about a less
All the stocks in the LTGG portfolio have tangible aspect: the people.
been subjected to months – often years – of Expectations are high, and rightly so.
careful research as we seek the world’s top LTGG holdings – many of which are large
long-term growth companies. We pay close companies in the public eye – can’t
attention to the fundamental character of afford to get stuck in the middle of the
companies, their attitudes and actions, pack and must define themselves as
going far beyond the Milton Friedman inspiring leaders.
‘shareholder only’ point of view.
2
After all, if we are to invest in companies From an LTGG research perspective, commissioning inquisitive research and
that delight customers with their products we recognise that no two stocks in your immersing ourselves in different global
and services over the long term, we need to portfolio are truly alike and our approach markets. As we hold stocks for many years
know about the individuals and the teams is thus necessarily stock-specific. The (10 years on average), we can get to know
behind those products and services – and process may consist of conversations with companies thoroughly.
the conditions in which they work. management and employees, site visits,
Looking back going forward
Tesla Lyft Google Amazon Meituan-Dianping
© Getty Images North America © Bloomberg/Getty Images © Geoff Robinson/Shutterstock © Shutterstock
Tesla’s workforce, while In late 2019, a decade In 2018, over 20,000 Over the past ten years, In China, online services
relatively small, has after the term ‘gig Google employees in Amazon’s workforce has company Meituan-
multiplied 60 times in economy’ was coined, 50 cities around the grown from under 25,000 Dianping, despite being
under 10 years and in Californian lawmakers world staged a mass to over 750,000 full-time only 10 years old, is
recent years has faced passed a bill paving walkout to protest and part-time employees. wrestling with strikes
accusations of health the way for ride-hailing against the company’s Now one of the 10 largest among its 2.7 million-
and safety concerns in drivers and food-delivery handling of alleged employers in the world, strong army of delivery
its factories. couriers to be treated as sexual harassment Amazon has faced drivers.
employees entitled to cases. intensifying criticism of
fuller labour rights. It was working conditions
a warning shot against in fulfilment centres.
Lyft and others.
3
>>>
In the three stock examples that follow: Amazon ( first purchased in 2004); Facebook (purchased in
2012); and Tesla (purchased in 2013), potential labour issues are leading us to take a closer look.
Amazon
Founder CEO Jeff Bezos Amazon’s robotic automation
acknowledges that criticism and now performs more of the tedious
regulatory attention are inevitable fulfilment centre tasks. Employees
given the company’s prominence reportedly now lift less and walk less.
LTGG • Work in progress
and growing significance. While There are now 3,000 robots for every
this public focus is not new, it is 10,000 Amazon employees.
intensifying.
Going further still, Amazon pays
In 2019, for example, Amazon was 95 per cent of tuition fees to help
named and shamed in the US National employees gain new skills and
Council for Occupational Safety and qualifications. The company has also
Health’s “Dirty Dozen” list of the committed to spending a further $700
most dangerous employers, following million on retraining its 100,000
multiple accusations of gruelling US employees in skills such as
working conditions. engineering and IT support.
So what has Amazon been doing? Despite signs of more proactive
At a minimum, it has emphasised its attitude, there is of course room for
support for the Core Conventions of Amazon to go much further. LTGG
the International Labour Organization has raised the issue of working
(ILO), the ILO Declaration on conditions in various meetings with
Fundamental Principles and Rights Amazon over the years, particularly
at Work, and the UN Universal those of lower-skilled and temporary
Declaration of Human Rights. In workers. More specifically, as the
October 2018, it went much further. It company is not an accredited UK
not only announced a $15 minimum Living Wage employer, we intend
wage for all full-time, part-time, to monitor pay levels and encourage
temporary, and seasonal employees the company to commit to fair pay.
in the US (and did similar in the UK), Further to a recent discussion with
but challenged other companies to do one of Amazon’s senior independent
better and urged the US Government directors, we are also encouraging
to raise the federal minimum wage. public disclosure of incident and
injury rate statistics for its workforce,
a step we believe can drive progress.
4
CASE STUDY
Looking back going forward
Notes
From a visit to the Amazon
fulfilment centre in Dunfermline
November 2019
The primary objective of this visit was to
better understand the realities of working
at one of Amazon’s fulfilment centres.
The tour walked us through the
end-to-end fulfilment process, from the
moment an order is received to completion
and dispatch.
What did we learn?
Employees work the same four-days-on,
three-days-off every week. Permanent
employees can state their preferred
work pattern and can request changes.
Systems are in place to prevent staff
from completing consecutive shifts, for
example a night-shift worker rolling into a
daytime shift. Nevertheless, this remains
a very demanding job.
From 1 November 2018, staff have been
receiving a minimum wage of
£9.50 per hour following Amazon’s
voluntary decision to raise wages in the
US and UK – representing a 36 per cent
uplift from its previous minimum wage.
Staff are free to join a union, although
there is no single dominant organisation
for the Dunfermline employees.
Permanent employees can access several
non-financial benefits, such as private
healthcare, subsidised travel, and up to
£2,000 p.a. over four years to complete
further education in any field. However,
such benefits are not afforded 5
to temporary workers.
© DANIEL LEAL-OLIVAS/AFP/ Getty Images
>>>
Facebook
Facebook’s workforce has nearly intelligence is helping to filter out
tripled since 2016. Of roughly much of it. Facebook is establishing
45,000-plus full-time workers today, an independent global oversight
over 30,000 are responsible for board to take leadership on complex
LTGG • Work in progress
online safety and security across the cases. As content moderation at this
Facebook, Instagram, Messenger and scale is entirely new and presents
WhatsApp ‘family’. The scale reflects unique challenges, it is important that
the fact that over 100 billion pieces Facebook learns and improves.
of content are shared on Facebook
every day by over two billion daily Rapid hiring and working conditions
users. The goal of Facebook’s content affecting content moderators are
moderators is to protect this immense aspects of Facebook’s business
online community from those who operations that have gained
seek to pollute it with violent, hateful prominence in recent times. We
or sordid content. Shielding roughly question how material these factors
a quarter of the world’s population may be for our long-term investment
every day from humanity’s darkest case. Are such working practices
impulses is an unprecedented task and sustainable? Is a user backlash
can be traumatic for those involved. likely? What about increased
regulatory intervention? Ongoing
Facebook is trying to ensure that engagement with the company is
content moderators receive training essential in considering our answers
and psychological support, as well to such questions.
as technological help to limit their
exposure to graphic content. Artificial
6
CASE STUDY
Looking back going forward
Notes
From a meeting with Facebook
March 2019
During our conversations with founder
CEO Mark Zuckerberg and vice president
of global policy management Monika
Bickert, the challenges Facebook faces
when it comes to policing content were
discussed at length.
We were told that it is impossible to create
a single set of rules to determine which
content is offensive and which isn’t, and
that this will always be the case.
For example, it would seem blindingly
obvious that the Facebook ‘family’ of apps
should be devoid of pictures containing
child nudity. But what if the picture in
question were the iconic scene of the girl
fleeing a napalm strike during the Vietnam
War, and the now middle-aged subject
of the photo wanted it to be publicly
available?
Many challenging questions must be
resolved. This is the reason so many
thousands of content moderators are
needed. While artificial intelligence
flags potentially offensive content, the
moderators review it in a very human,
very manual way.
We heard that Facebook has turned
the corner in terms of online safety and
security. It has emerged as an industry
leader in terms of publishing detailed and
transparent reports of what is has removed.
Nevertheless, it is still early days for the
company in this effort and we want to
learn more. 7
>>>
Tesla
It wasn’t so long ago that Tesla health facility has shifted from basic
was wrestling with the production triage and first aid to a specialised
of its Model 3 vehicle, difficulties clinic staffed by three full-time
that threatened the company’s very doctors providing assessments and
LTGG • Work in progress
survival. Little wonder that founder on-site care. Tesla also introduced
Elon Musk – weary from countless an early symptom intervention
nights sleeping on-site at the programme to identify and address
Fremont factory – termed this period potential sources of injury in the
‘production hell’. But while Model production process.
3 production expanded dramatically,
so too did media reports of factory Because of such measures, Shelby
workers being overworked, poorly noted that injury frequency rates had
treated and exposed to injury risk. declined to auto industry average.
Were working conditions being Characteristic of Tesla’s focus on
sacrificed to rapid growth? continual improvement, the Model 3
production line has been designed to
Eager to deepen our understanding, be safer than the previous S and
we met Laurie Shelby, Tesla’s head X lines and boasts the lowest injury
of environmental, health and safety record at the factory. Consequently,
(EHS). It was, she informed us, her the S and X lines have been retrofitted
first meeting with shareholders. with improved ergonomics.
Shelby is responsible for Tesla’s
drive to run the safest car factory Recognising the positive correlation
in the world. We learned about the between injury rates and ramp-ups
bi-weekly meetings of Tesla’s over in production, our ongoing
200 EHS professionals, the creation engagement with Tesla seeks to
of an EHS centre of excellence, and understand how the company will
the company’s all-in-one reporting tackle this challenge, particularly
tools which record all incidents and as production expands at home
enable staff to flag issues and suggest and abroad at the Shanghai and
improvements. The Fremont factory’s forthcoming Berlin factories.
8CASE STUDY
Looking back going forward
Notes
From a visit to Telsa’s Fremont factory
September 2018
The objective of this visit was to learn more
about the company’s approach.
What did we learn?
The Fremont factory operates 24/7,
with employees on alternating work
schedules consisting of three days on
and two days off. Individual departments
have an input when setting their work
patterns, with employees able to vote for
scheduling options.
New employees receive several hours of
health and safety training in their first two
days, and over 20 hours in their first year.
The company has a ‘Find it – Fix it’
process, whereby employees can flag all
manner of issues and make improvement
suggestions. Thousands are flagged
each year.
Our impression from the tour was that
safety guidance and notices were clearly
visible. While all areas were generally tidy,
we felt this could have been better
in places.
The Model 3 production tent is impressive,
despite being completed in just 18 days.
We were told employee retention rates
in the tent are higher compared with the
factory. We noticed the tent is brighter,
more spacious and benefits from a
constant flow of fresh air.
9
© Corbis News/ Getty Images
>>>The cases above illustrate some of the ways in which we seek to engage LTGG holdings where working conditions may materially
undermine our investment theses. However, we also look to support and encourage holdings whose workforce characteristics appear
to be driving their long-term success, thereby enhancing (rather than detracting from) our ‘blue sky’ investment scenarios, as the
examples below illustrate.
LTGG • Work in progress
The Tencent Academy runs 8,000 online Salesforce has appointed a chief equality It takes 15 to 20 hours of manual labour
and offline training courses a year, in officer to ensure that 50 per cent of its to produce a Hermès handbag. The
addition to over 8,200 live classes. They US workforce will be composed company has developed a bespoke
have been accessed over a million times of under-represented groups by 2023. health training programme to prevent
by Tencent’s 40,000 employees. In the spirit of transparency, it publishes its skilled artisans from incurring
its equality data annually. repetitive strain injuries.
Recognising that embracing failure is an Netflix takes the view that it can bring
essential part of a high-paced innovation out the best in its employees by giving
culture, Spotify encourages its teams to them the freedom to manage their
hold project retrospectives when work-life balance themselves.
things go wrong. These are called Employees are encouraged to
‘fail-fikas’ (fika is Swedish for a take as much vacation and parental
‘get together over a coffee’). Founder leave as they need.
CEO Daniel Ek sets the tone
10 by sharing his own setbacks
and lessons.Conclusion
Looking back going forward
As we deepen our understanding of the
labour issues facing LTGG holdings, we can
build a more complete and helpful picture
of their corporate character – attitudes and
actions that can determine their long-term
success. And while the direction of travel is
encouraging in many cases, the examples
cited here illustrate the ways in which
exciting growth companies must always
adapt. Mistakes will be made. Lessons will
be learned. Patient engagement has never
mattered more.
11Long-term talent
South Africa-born Gemma Barkhuizen and Robert Wilson, who comes from Northern Ireland,
joined the Long Term Global Growth (LTGG) team as part of Baillie Gifford’s graduate training
programme. They are, respectively, postgraduates in history and philosophy, and were poised
for careers in academia before our recruiters helped them change their minds and switch to
investment. Here they share their experiences, talk about what’s important to them, and give
their view of LTGG’s way of looking at the world.
LTGG • Long-term talent
GEMMA ROBERT
How did you come to join the LTGG team? You studied philosophy at Cambridge and
I joined Baillie Gifford on the asset managers. I responded out Yale. Is there a link between philosophy
Investment Research Graduate of curiosity and was pleasantly and investment?
Scheme just over two years surprised to discover that the
There is a similar process involved. What I liked about
ago. After studying at Rhodes job was really about figuring
philosophy was finding a topic, reading a lot about it,
University in South Africa, I out how the world works.
writing about it, arguing about it and developing it. It’s
completed a master’s in modern
After spending a year on the the same structure as I have now, so there’s no
history at Durham University.
European Equities team, I big change.
I had this romantic idea of a
life dedicated to research and joined LTGG in September
Some topics within philosophy are immediately
stimulating debate within a 2018. The idea behind the
relevant to what we do here, like assessing what is
collegiate atmosphere and was rotation of trainees is that you
the appropriate amount of empirical evidence needed
set to do a PhD. But I found sample different investment
before something becomes knowledge, how we can
I was becoming increasingly styles and figure out what you
think about what is likely to happen, and what is the
frustrated with knowing more like. LTGG has the kind of
nature of explanation. I had thought a lot about
and more about less and less. approach to investment that
these topics.
Just as I was rethinking what you either really like or you
I wanted to do with my life, really don’t. Happily, I found Interestingly, these issues do come up in my work.
Baillie Gifford’s recruitment quite early that my fit with We do use a philosophical vocabulary. I’m so
process was targeting LTGG was just great and I was convinced about the relevance of philosophy that
university departments that able to stay on. I’ve started a series of philosophy seminars, bringing
wouldn’t ordinarily produce philosophers into the office from Cambridge and
Edinburgh universities. So far, we’ve had 20 or 30
people attending them to discuss the bedrock of
how philosophers think of a subject, in the hope of
12
embedding that language and analytical approach into
what we do.Fitness is a market that
has remained stubbornly
fragmented and immune
to digital distribution.
© Peloton Interactive, Inc.
GEMMA
Looking back going forward
Can you give me an example of a stock you’ve
brought to the portfolio and explain why you
favoured it?
There’s Peloton, a digital fitness company. It markets
gym equipment, but also sells subscriptions for
© Future Publishing/Getty Images
membership of an online library of digital fitness
classes. It’s like a Netflix for fitness, but of course the ROBERT
economics are much better: creating a few exercise
classes compared to having to make an entire expensive Tell us about your interest in gaming.
TV series. What excites me about Peloton, and why I How did it come about?
think it demonstrates the dynamics of the kind of growth
outlier we are looking for, is that it’s trying to disrupt I had a PlayStation when I was six, and I’ve played
an enormous and growing gym market, not just in the games ever since. I can see a lot of change in the way
US, where it started, but now in the UK, Canada and I’ve played and in the role that games play in people’s
Germany as well. lives more generally.
Fitness is a market that has remained stubbornly Gaming interests me because it has compelling
fragmented and immune to digital distribution. It has economics but also major challenges. It’s a hard sector
come up against the banal fact that people tend to to analyse because it’s so hit-or-miss. What puts off
go to the gym or the spin studio that happens to be investors is that you can publish one great hit title and
nearest their office or home, a constraint that limits the you can take 80-90 per cent margin and then the next
addressable opportunity of any single fitness company. game doesn’t even earn back the cost of development.
With Peloton, that’s no longer true. If I’m using Peloton
Digital, for example, I’m doing a fitness class with an Game companies are getting better at monetisation,
instructor based in New York who is one of the best in but there are risks associated with that, for example,
the world. I would never have been able to get a class with the ‘loot box’ model. This is considered
from that instructor, and he or she would never have exploitative and is illegal in some parts of Europe,
been able to address a community of more than a million but it’s the most common way in which games are
people, if it had been limited to a particular studio. monetised in Japan. In this model, games are free to
play, but there’s a gambling component within them.
It’s rational for Peloton to pay for the very best There’s still a lot of experimentation with ways to
instructors because the company can spread those make money out of gaming in different parts of
costs over a global subscriber base in a way that no the world.
other fitness company has been able to do. That’s
my hypothesis, to which I would add its first-mover Tencent is one of the most valuable companies
advantage, the brand it has carefully built up and the in the world and a very large portion of its cash
motivation of the management team. All of these make flow comes from intelligently monetised game
me think it might be exceptional, not just another content. You can already see how these are much more
humdrum fitness company. attractive businesses than they have been historically.
The investment community hasn’t really caught up 13
with that yet, probably because it is still largely made
up of older men who don’t play video games. They’re
not necessarily picking up change.
>>>GEMMA
There’s been lots of negative press coverage about
unlisted companies taking too much time to reach
profitability. Does that matter?
The market tends to treat Peloton has very low attrition
all currently unprofitable rates, lower even than
companies the same, without Spotify or Netflix, which are
any regard to the strength or companies we love. It’s even
weakness of their underlying more remarkable considering
business models or to to the the difficulty of remaining loyal
credibility of their path to to any fitness regime. Investors
LTGG • Long-term talent
profitability. These things should not think of this spend
How does LTGG work
matter profoundly. Frankly, as a loss, but rather as a
it’s just laziness. There are worthwhile investment. with Baillie Gifford’s
big differences between these Unlisted Equities team?
companies. The market tends to punish
companies for not being A bit of history might be
Let’s go back to Peloton, sufficiently short term, and not helpful here. Peter Singlehurst,
which listed on public being able to hit a quarterly who heads the Unlisted team,
markets towards the end of earnings target because they’ve started to look at more private
last year. This is a company been investing in a long-term companies while he was a
that isn’t yet profitable, opportunity. We like that. We member of the LTGG team. At
but its accounting loss is a think that’s what they should that time, a growing number
result of it spending more be doing, but it’s harder to do of companies were deciding to
than a third of its revenues when you’re a publicly listed stay private for longer because
on sales and marketing, company. It’s important to they were able to access capital
ploughing these revenues into think of the structure of the that was previously only
customer acquisition. That’s industry that the company is available on public markets.
a worthwhile investment, in and the defensibility of the The dedicated Unlisted
because the return that it earns earnings it’s trying to create. Equities team was spun out of
on each individual acquired The fact that a company isn’t this effort. At the same time,
is very high and it has a lot of profitable today doesn’t tell you Baillie Gifford’s Scottish
scope to improve that because anything about that. You need Mortgage Investment Trust
of the loyalty of its customers. to dig deeper. was becoming more interested
in unlisted companies and its
managers, James Anderson and
Tom Slater, are LTGG team
members. So it is natural that
researching private businesses
is embedded in our way of
thinking. The Unlisted Equities
desk is right next to ours. Its
team often comes to our stock
discussions. We’ll float ideas
by them if we come across
interesting private companies
and they will flag up things
about a company when they
know it’s going to be listing
soon.
14 © Peloton Interactive, Inc.ROBERT
You recently returned from China.
What were your impressions?
When you go to China and engage with companies,
it becomes clear that the emerging-market-versus-
developed-market distinction is not real. There are
many ways in which Chinese companies and the
Chinese economy are more sophisticated than those
of Europe or the US. This is not a more developed
economy in aggregate, but in areas such as mobile
payment, social media and digital entertainment,
Looking back going forward
China is moving quicker than the West.
The Chinese retail sector is probably more
sophisticated. There’s more use of data, and vending
machines seem to sell almost everything. The
categorisation into ‘developed’ and ‘undeveloped’
is a weird distinction. It’s as if we in the West have
achieved a steady state and they’re going to converge © Getty Images North America
towards us. The whole language around this just
seems increasingly meaningless.
How does LTGG work with Baillie Gifford’s
I don’t think western investors have an easy
newly opened Shanghai research office?
time understanding how competitive the Chinese
environment actually is. The culture of super-apps There’s a certain amount of physical exchange in that
means the platforms compete on new products all the the people from the Shanghai and Edinburgh offices
time, while smaller, disruptive firms at the margin go back and forth. John MacDougall, our Shanghai-
battle it out in a way that makes Silicon Valley look based partner and member of the LTGG team, comes
pretty polite. The companies that survive have come back to Edinburgh frequently. Then there’s also the
through that environment and it shapes their cultures fact that we share virtual meetings: Mark Urquhart,
distinctively. for example, has just published a note, and John is
marked as ‘in’ the meeting, even though it was held
We sold the search engine Baidu because we were in South Korea. In terms of our investment process,
increasingly unconvinced by its ability to stay I would emphasise that little has changed, although
adaptable. Platforms like Toutiao and Douyin (TikTok our discussion slots are scheduled at nine in the
outside of China) were springing up in terrain that was morning which is four in the afternoon in China. We
a little too close for comfort and hiring Baidu staff hold most of our discussions by Zoom teleconference
who were frustrated by its lack of experimentation. – another stock we have discussed recently. The
ByteDance, which owns Toutiao and Douyin, has Shanghai-based investors are pretty much there with
since become an important competitor, and although us in everything but a physical sense.
still an unlisted business, we’ve been following it
closely.
All of these businesses have distinct cultures. It’s
misleading to call Alibaba ‘the Chinese Amazon’, or
to see ByteDance in terms of a Western counterpart.
15
>>>GEMMA
LTGG has just celebrated its 15th
anniversary. What will it look like
in 15 years’ time?
If you look back at the initial LTGG portfolio 15 years
ago, although the team was trying to do something
different, rock-solid standard benchmark stocks were
still well represented. Since then, the portfolio has
incrementally changed and what we’ve ended up
with bears very little resemblance to what we started
off with, even though in theory it followed the same
LTGG • Long-term talent
investment philosophy.
The pursuit of transformational growth and what we
think might be the very best growth companies in the
world has become increasingly extreme over time. If
I had joined LTGG 15 years ago, I would have had
a very different experience than now. I hope that if I
am still working on LTGG in 15 years there will have
been continuous pushing of the boundaries, moving
further and further away from following industry
practice and convention.
I don’t know where that will take us, but one of the
things that I think might prove to be very important
for the strategy is the Shanghai office. This region is
going to matter structurally. We need to get to know
these Chinese companies and these consumers.
That might be one of the things that we’ll look back
on 15 years from now and think, ‘Can you imagine
that we used not to have a China office? How were
we able to weed out all of the misperceptions of
companies in that region?’
16Looking back going forward
ROBERT
What do you find most enjoyable and
rewarding about your job?
I like the stimulation. I struggle to imagine any other
private sector work that is as interesting as what we
do in terms of learning a lot about a lot of different
things in the most expedient way. If, like me, you
have an academic disposition, that’s really rewarding,
especially when the firm supports your research
financially. You can go and do what you think is useful
because we believe that this is of great value to us and
our clients.
Autonomy is important as well, and that’s an
especially nice feature of LTGG. You get a lot of space
to pursue anything you think is a good idea, though of
course you can’t just do what you want without any
hope of it ever being useful. But, even if 80 per cent
of it is not useful and 20 per cent is that 20 per cent
makes the approach more valuable than it would be
if it were more prescriptive.
17Ringing the changes
In the late 2000s, when we came to assess risk, we were finding Instead, we looked to group our holdings by what we believe to
that the traditional approach – grouping stocks by traits such as be the single most important underlying driver of outcomes.
country and sector, and analysing historical correlation – was
not helpful. In 2009, we first used an Euler diagram to depict the portfolio’s
range of growth drivers and opportunities. Regular revisions of
Firstly, it rests on an assumption that correlations are stable this simple diagram would depict our evolving views.
LTGG • Ringing the changes
over time. Secondly, such classifications are myopic. A country
classification may tell us the location of stock market listing but This Euler diagram (named after an 18th century Swiss
less about the risk associated with where the company earns the mathematician) helped us to understand our portfolio and
bulk of its revenues. Traditional sector classification can also be the results of our decisions.
obscure. For example, both New Oriental, a Chinese education
provider, and Hermès, a French luxury goods manufacturer,
inhabit the same consumer discretionary sector.
Banco Santander Apple
Canon
FINANCIAL
RECOVERY/
2009
PRECIPICE
UBS
TECHNOLOGY/
OBSOLESCENCE
Amazon
eBay Nintendo
SAP
E-COMMERCE/
INTERNET
Google
Gazprom
Teva
ENERGY Walgreen
Lukoil HEALTHCARE
Petrobras
Straumann
CVRD
China Mobile
New Oriental
COMMODITY
DEMAND/
NEW
PRICING First Solar
CONSUMER
Iron Mountain Q-Cells
Atlas Copco
Novozymes ALTERNATIVE Hermès L’Oréal
Sandvik
ENERGY PPR Porsche
UNCORRELATED INDUSTRIAL Vestas
SPENDING/ WESTERN
VCA Antech INFRASTRUCTURE Deere CONSUMER
Pool
Zhejiang ABB AGRICULTURE
Whole Foods
18
2009 thematic buckets which
do not have a 2020 equivalentAs shown in the diagrams below, our classifications shifted in line
with changes in society and understanding. For example, the ‘Internet/
ecommerce’ group was not descriptive enough when we realised that
‘an advertising company like Google is actually pretty different from
a retailer like Amazon’.
Euler diagrams continue to provide a useful framework for review and
Looking back going forward
discussion. The groupings reflect our thought processes in putting the
portfolio together and they help us to abstract from the details of individual
investment cases to make sure the portfolio is sufficiently diversified.
ASML Pinduoduo
IMMERSIVE
COMPUTING Shopify
NVIDIA
RETAIL
REVOLUTION
Illumina
Salesforce
Amazon
EFFICIENCIES IN Alibaba
Ionis Pharmaceuticals
HEALTHCARE FUTURE OF
Tencent
ENTERPRISE
NOVEL SOFTWARE
TREATMENTS FRICTIONLESS
Dexcom
FINANCE Zoom
Atlassian
Adyen Workday
Pinduoduo
L’Oréal
Delivery Hero
Inditex
FUTURE
OF FOOD
FASHION &
Meituan Dianping IDENTITY
Tesla Netflix
Trip.com Hermès
FUTURE OF Kering
MOBILITY AIA
Alphabet
NIO NEW AFFLUENCE
2020
CHANGING
MEDIA HABITS Tencent
TAL Education Group
Spotify
Peloton
HDFC
FUTURE Facebook
OF LEISURE
NetEase
19AI:
Learning
LTGG • Al: Learning on the job
on the job
The great data gold rush, plus faster and smarter
computer processing, is giving companies
superhuman powers of self-improvement
Unless you’re a computer geek, can you really
comprehend the potential impact Artificial Intelligence
(AI) could have on our lives?
It’s likely to be huge. PwC suggests it could add $15
trillion to the world economy by 2030. Andrew Ng,
co-founder of Google Brain, former chief scientist at
Baidu, and teacher of machine learning at Stanford
University, calls AI “the new electricity”.
AI has taken significant leaps in recent years, supported
by the march of computing processing power. AI-driven
algorithms now defeat world champions at games such
as chess and its even more complex Asian equivalent,
go. The next challenge for the artificial mind is to tackle
some of the world’s biggest problems in healthcare,
climate change and energy efficiency.
20Why now?
How come we’re all suddenly talking An ML algorithm would describe this
about AI when it’s been around for over challenge as the ‘target’. What factors
50 years? Three reasons: data, computing might influence my decision? Perhaps the
power and algorithms. day of the week, whether my best friend is
out, or whether Game of Thrones is on TV.
Data is the fuel of AI and it’s now Having defined these so-called ‘features’,
gushing forth freely. We now generate the the algorithm needs data points to ‘train’
equivalent of all the data created in 2002 with – this is the machine learning bit. The
every week. No fewer than 190 million algorithm starts dumb but learns fast:
Looking back going forward
emails are sent every minute, 300 million
Google searches are conducted per hour Data point 1: Monday, friend staying in,
and an estimated 100 billion-plus of GoT is on. Outcome: Stay in.
TikTok’s mini-videos are viewed per day.
In China alone, mobile data consumption Data point 2: Saturday, friend is out,
trebled in 2019. GoT is not on. Outcome: Go out.
Computer power has charted a similar … and so on and so on.
path, growing exponentially for decades
As more data points are automatically fed
following the predictions of Moore’s Law.
in, the algorithm adjusts the weighting
It is now incorporated into a myriad of
and importance of the features. Do this
previously analogue devices, and even into
a million times in quick succession and
the human body.
you end up with a model that does an
This in turn, has boosted an application incredible job of mimicking the real-
of AI known as machine learning (ML), world decision-making process. Add in
where algorithms use what they’ve learned another datapoint into the model without
in the past and apply it to new problems. an outcome, and the algorithm will predict
your reaction with high accuracy.
Algorithms were previously simple sets
of instructions – think of a recipe to bake The key attraction of algorithms is their
a cake – but ML algorithms changed flexibility. With the right data it’s easy
the rules. Lacking access to the recipe, to set the target to ‘what music does this
they can still achieve the same outcome. person want to hear next?’ or ‘what ads
Instead, the computer is given the key data should I show to tempt them to buy?’
points and is ‘trained’ to work out how to
Long Term Global Growth searches for
produce the cake by itself. In short, ML
exceptional businesses that experiment
algorithms dispense with the chore of
and adapt to incorporate evolving
writing endless lines of computer code to
technologies. So naturally, most companies
programme a specific outcome.
in the portfolio already employ machine
Therein lies the opportunity. Algorithms intelligence at scale.
can be used to predict virtually anything.
Suppose you wanted to predict whether I
would venture out of my house tonight.
21
>>>The enablers
Two distinct eras of computer usage in training AI systems
Petaflop/s-days
1e + 4
AlphaGoZero
1e + 2 Neural Machine
Translation
TI7 Data 1v1
1e + 0
LTGG • AI: Learning on the job
VGG
ResNets
1e - 2 AlexNet
3-4 month doubling
1e - 4 Deep Belief Nets and
layer-wise pretraining
DQN
1e - 6
TD– Gammon v2.1
BiLSTM for Speech
1e - 8 LeNet-5
NETtalk RNN for Speech
ALVINN
1e -10
1e -12 2 – year doubling (Moore’s Law)
1e -14 Perceptron First Era Modern Era
1960 1970 1980 1990 2000 2010 2020
Source: OpenAI
A small number of ‘enabler’ companies is vital
to the AI supply chain.
ASML, held in the portfolio since 2017, could be US firm NVIDIA is another enabler, held since 2016.
the most important company in the world you’ve Its importance in the industry cannot be overstated.
never heard of. Without the Dutch manufacturer’s Its graphics processing units (GPUs) have evolved
machines etching intricate designs on silicon into a computerised brain, straddling the exciting
wafers, the technological revolution would soon intersection of virtual reality, high performance
stall. ASML makes the machines that produce computing and artificial intelligence. GPUs are
the ‘brains’ of electronic devices, able to handle the single most important items in developing AI
the AI workloads as data volumes continue to applications and they are in demand across the globe.
explode. That’s lucky, as the computing power Tencent’s cloud gaming service will soon be powered
required to train state-of-the-art AI models has by NVIDIA chips, meaning complex graphics can be
grown over 300,000 times since 2012, shooting rendered in real-time via an internet connection. The
past what Moore’s Law predicted. need for gaming consoles will soon disappear.
Alibaba’s and Baidu’s recommendation engines run
on NVIDIA chips as well, and Alibaba has recently
lauded their success. Click-through rates improved
by 10 per cent through use of their chips, bringing
instant revenue benefits. China aims to become an AI
22 superpower in the next decade. However, it will rely
largely on the technology of two foreign companies to
make that a reality.The usual suspects
Not surprisingly, portfolio holdings The breadwinner for these firms is still What’s next? Well, how about AI within
Amazon, Alphabet, Netflix, and the recommendation engine. Respectively robots themselves? Amazon held a
Facebook have been using AI for years. 61 per cent and 76 per cent of the AI robot-versus-human ‘picking and placing’
They all use algorithms on their core workloads of Google and Facebook challenge back in 2015. The humans won,
platforms in generally the same way, as come from search and newsfeed naturally, managing to process around
recommendation engines. However, their recommendations. Those weightings tell 15 times more items per hour.
uses of AI are broadening. us where most of their revenues come
from. However, different types of AI, such Fast forward to 2018, and that difference
Facebook now uses machine vision to as natural language processing (voice and has narrowed to just twice as many per
Looking back going forward
take down nefarious imagery from its translation) and machine vision (images) hour. Like most AI systems, picking
platform. Alphabet said in a recent financial will only grow in importance. robots are improving at phenomenal
report that “machine learning and artificial speed. Covariant, a Berkeley-based robot
intelligence (AI) are increasingly driving For example, the number of brands start-up, focusing on warehouse logistics
many of our latest innovations, from partnering with Amazon’s voice assistant technologies, improved robot accuracy
YouTube recommendations to driverless Alexa is growing. In India, KFC now from 15 per cent to 95 per cent in only five
cars to healthcare diagnostics”. Amazon offers a hands-free, cash on delivery, voice months. It’s only a matter of time before
wants to put machine learning capability ordering service. A novelty for some, but human capabilities are superseded.
in the hands of every developer and data in a country where illiteracy still runs rife,
scientist across the globe. Its end-to-end voice creates a channel to reach potentially
machine learning service called SageMaker, millions of dormant consumers.
handily available on Amazon Web Services,
is doing just that.
23
>>>Tesla
The long-term opportunity for Tesla has
broadened since LTGG’s first investment
in 2013. Elon Musk’s ‘Master Plan’,
LTGG • AI: Learning on the job
penned in 2006, was to create a low-
volume car, use that money to develop
a medium-volume car at a lower price,
and then use that money to create an
affordable, high-volume car. With the last
step of this plan in train, the focus has
shifted to creating self-driving capability
for their entire fleet through machine
learning.
The self-driving opportunity for Tesla may
be seriously underestimated. This is natural
as it has little to do with the operational
aspects of car production on which most
analysts focus. Developing self-driving is
mainly a data problem. In simple terms,
if enough visual data is collected from the
eight ‘surround cameras’ fitted on every © Bloomberg/Getty Images
Tesla vehicle, the company’s algorithms
will eventually be able to perceive the
world as we do in real time, and drive
from point A to B safely. Tesla now has
data from over three billion miles driven
Tesla now has data
using its ‘Autopilot’ system. It took them from over three billion
four years to get to one billion, and less
than a year to double that number. Progress miles driven using its
is rapid. ‘Autopilot’ system. It took
An annual subscription to a fleet of Tesla them four years to get to
self-driving vehicles is likely to be a
compelling offer. Tesla’s margins would one billion, and less than
look more like those of a software business
than of a traditional car company should
a year to double that
this come to fruition. Tesla’s long-term number.
success has as much, if not more, to do
with AI advances than with the mechanics
of car production.
24China’s AI superpowers
China is committed to becoming the world leader in AI by 2030. With an online population of over 800 million, three times that of the
US, large-scale data collection is effortless. This is not surprising in a country where citizens worry less about privacy and censorship.
Less than a decade ago, China and the US were developing AI capabilities at similar rates. That is now a distant memory.
In November 2017, China’s Ministry of Science and Technology announced that the nation’s first wave of open AI platforms will rely
on Alibaba for Smart Cities technology and Tencent for medical imaging and diagnostics.
Rapid development in China is nothing new. It’s part of the reason why Baillie Gifford recently opened a research office in Shanghai.
Many products and services in China now have no US analogue, with some Chinese-born ideas now going global.
Looking back going forward
Alibaba
Alibaba’s ‘City Brain’ crunches
data from cameras, sensors,
social media feeds, and
government data. Algorithms
are then used to predict
outcomes across healthcare,
urban planning, traffic
management, and more. Clearly
Alibaba is more than just a
leading ecommerce platform.
Tencent
Tencent aspires to becoming a
leader in personalised medicine
using AI. With around 40,000
medical institutions on its
© Visual China Group/Getty Images
messaging service WeChat, Employees work at Pinduoduo headquarters on July 25, 2018 in Shanghai, China.
© Visual China Group/Getty Images
as well as several thousand
that accept WeChat payments,
Tencent has access to a
treasure trove of consumer data Pinduoduo Meituan-Dianping, TikTok
to help train its algorithms. Its Pinduoduo (PDD), only five Meituan-Dianping, the food TikTok, the viral video app
aspiration to become a digital years old but already China’s delivery behemoth, delivers owned by ByteDance, is
assistant to all industries may second largest ecommerce more than 30 million meals an example of how fast
not be so outlandish given the company with over 500 million per day. It now has over 400 development can happen.
firm’s laser focus on developing active users, is using AI to million users on its platform, TikTok supersedes the
AI capabilities. It’s YouTu lab, a help farmers meet consumer regularly ordering hot meals. traditional feed-and-follow
leader in machine learning, aims demand. What was a complex It couldn’t do it without its AI model popularised by Facebook
to help them achieve this goal. supply chain of warehouses, ‘Super Brain’ which integrates and Instagram. With TikTok,
distributors and retailers, has real-time computation, AI comes first. Videos go viral
been disrupted and simplified offline data processing and on the platform with ease due
by PDD to give better terms machine learning to perform to large-scale deep learning
to the farmer. PDD set up ‘deep sensing’ and build its algorithms pushing content to
Duo Duo Farms to help it understanding of the world. interested users. The platform
gain the necessary skills to All of this results in a better is now a global hit, with over a
sell directly on the platform, customer experience. Average billion users, all from an app not
without having to rely on layers delivery times have reduced yet five years old.
of intermediaries. Pinduoduo from an hour to 30 minutes 25
neatly connects farmers in a few years. Not surprising
(the first mile) directly with when an abundance of data
consumers (the last mile). is collected related to delivery
times, pricing and logistics
network design.
>>>LTGG • AI: Learning on the job 26
The road
ahead
With AI becoming intrinsic to the
strategy and operations of so many
LTGG holdings, we should be optimistic
about the potential benefits it can offer
companies, but also wary of the associated
tensions and biases that could creep in
along the way. That’s why we find our
Looking back going forward
partnership with Cambridge University’s
Leverhulme Centre for the Future of
Intelligence so valuable. It aims to explore
the opportunities and challenges of this
potentially epoch-making technology, in
the short and long term. We look forward
to exploring their thoughts on points
of tension, such as that between use of
personal data to improve services versus
respect for privacy and freedom of choice.
These issues affect all companies. Solving
them is vital to navigating the obstacles
that AI could throw up.
27Lessons from the
Sonoran Desert
LTGG • Lessons from the Sonoran Desert
What an Arizona-based academic taught us about where growth springs from
The initial evidence
Back in 2014, we conducted some detailed empirical work on the pattern of returns in
equities, using the US market as our dataset. One key observation was that the top five
per cent of stocks in the US equity markets tend to be ‘five baggers’, investments that
earn five times their purchase price, over rolling five-year periods.
For many years, our focus had been on finding the stocks that could grow many times
over, but this elegant empirical finding was a helpful step towards establishing the
five-bagger baseline for our growth hurdle. It also showed the importance of the large
outliers. We didn’t need to find many to drive strong client returns.
But a couple of niggling questions remained.
How could we be sure that this performance wasn’t a temporary phenomenon?
Was there any independent evidence to back up our observations?
The cost of
a handful of
inevitable clangers
is dwarfed by the
heights of a few
big outliers
28The
independent
proof
In 2017, these nagging questions were addressed
in a paper that was published without fanfare
by a modest Swedish academic called Hendrik
Looking back going forward
Bessembinder. Based at Arizona State University, he
had analysed over 25,000 stocks between the years
1926 and 2016.
Collectively those stocks had generated net returns
of around $35 trillion over and above US Treasury
bills, but when Prof Bessembinder ranked them by
return he found that:
Another 38% of A mere 4% of them
58% of them had them had made up had collectively driven
destroyed value, for that value the entire net return,
collectively posting destruction, collectively collectively delivering
a return around minus posting a return of around $35 trillion
$6 trillion around $6 trillion between them
We viewed these observations as probably the most
important findings we had ever encountered in equity
investing. They were the first independent proof of the
persistently extreme skew in US equity market returns
over long periods of time. Just four per cent of stocks
drove all returns, a fact completely overlooked by
most of the investment community.
The paper was important and exciting, but there was
another question: How could we be sure that this
wasn’t a US-centric phenomenon? We asked Hendrik
to explore this important question further on behalf of
Baillie Gifford.
29
>>>Further
rocket fuel
With our support, Prof So the overriding observation
Bessembinder embarked on was that the extreme skew of
a heroic feat of data collation. returns applies globally, not just
61% of them
He built an enormous dataset at US level. Indeed, at a global
had destroyed value,
containing the returns of over level, the extremes are even
collectively posting more pronounced.
62,000 companies, delivered
LTGG • Lessons from the Sonoran Desert
a return of
between 1990 and 2018. He
around-$22 trillion At this stage, it seemed sensible
then spent months diligently
to explore whether that special
crunching the numbers and in
one percent of companies
mid-2019 was ready to share Just 1% of them had anything in common. We
these conclusions: had driven the had demonstrated an ability
Collectively the 62,000 entire net return, to identify these outliers
companies had generated net collectively delivering historically, but how could we
returns of around $45 trillion around $45 trillion Another 38% of maximise our chances of doing
over and above T-bills, but between them them had made up for so in the future? We thought
when ranked by return, it was that value destruction, that understanding shared
found that: collectively posting characteristics would help.
a return of
around $22 trillion
Wealth
Created
$44.7
Trillion
1.3% (811 Firms)
One Month
Treasury Bill
100%
-$21.8
Trillion % of Firms out of 61,100
30
60.9% (37,195 Firms)
37.8% (23,905 Firms)You can also read
