How to stop worrying and learn to love probability

Veronica Silva
December 20, 2017

Canada has a constitution opposed to probability. Our founding principles are “peace, order and good government.” Probability tramples on order, mocks cause and effect. One plus one equals two, mostly. Sometimes it equals zero… sometimes several billion. And so, we hate it. Probability now drives the knowledge economy. So, if we don’t want to lose our civil society we must embrace and master probability. Here’s how we can surf the probability tsunami and remain true to ourselves.

Where probability comes from

In the Renaissance, astronomy emerged as a science and with it, measurement uncertainty. To manage the uncertainty probability emerged from mathematics. Even then, scientists argued that probability is a waste -- all you need is one good measurement.

Carl Friedrich Gauss showed in the early 19th century that most measurements cluster around a very narrow range of outcomes. And so the idea of “standard deviation” entered our thinking. Yes, there are outliers. But they are very rare. Measurements map out in beautiful bell-shaped curves. The standard deviation measurements form the bell, the outliers are pushed to the edges.

But the bell curve began to crack in the late 19th century when Vilfredo Pareto showed that some events are more impactful than others. In everything from garden pea pods to land ownership, 20% of the events produced 80% of the results. In the 1940s, quality management evangelist Joseph Juran dubbed this Pareto’s Law.

By the early 21st century Pareto curves overlie Gaussian curves. Still a bell-ish curve. There are fewer events in the standard deviation, so the bell is stumpier. Its thicker edges reflect the fact that outliers are far more frequent than Gauss predicted.

Risk managers found that their data on extreme storm events and the most powerful earthquakes followed a Pareto distribution rather than a Gauss curve.

Where probability is going

For some years venture capitalists believed that there was a glass-half-full side to Pareto’s Law. Sure, bad outlier events could be really bad … but there must be inconceivably good outcomes too.

Then came Facebook. Founded in 2004, in 12 years its profit topped US$1 billion. In 2005, Paul Graham founded Y Combinator , a Mountain View CA-based seed accelerator. It was Graham’s experiment to test his hypothesis that outlier events could be incubated.

When you map the results of YC companies against other startups there is no bell-like curve. Nearly all YC companies cluster at the top left side of the graph. Few in number with extreme growth in sales, profit and valuation, then results plunge down to all the other startups. They form a long tail that dwindles slowly off to the right, with comparatively small returns.

This is now referred to as a Power Law distribution. In 2004 there were about 10,000 startups and Facebook’s results top the total of all the others. Every year, there’s another one of these unicorns.

Should we chase after unicorns like everybody else? No. Strategy 101 says: “if you want to knock off the market leader using his methods you will need a three-to-one advantage in material and resources.” At least. Besides, the Silicon Valley model is deeply rooted in US culture and the structure of its economy.

A different model of innovation

The German model is a better fit for Canada and Canadians. Germans practice commercialization as a joint private-public partnership. Sometimes the public sector funds the capital costs, and industry and academe carry the operating costs. This is how the ZAL aerospace hub in Hamburg works. Sometimes the private and public partners split the core funding, and private companies fund commercialization projects derived from the core-funded work. This is how the Fraunhofer Institutes are structured.

Early on the Germans recognized that commercialization involves a lot of uncertainty and the timelines are often longer than those of governments and companies. Governments must face voters every few years, and companies have to give shareholders an annual return. Neither are suited to commercialization, so they share the effort of sustaining it.  Germans don’t try to hit grand slams. They play tortoise to America’s hare.

Canadian startups often complain about access to capital, and it’s tough, but it obscures the real issue. Go-to-market costs are far higher than research costs. It’s expensive to build brand and distribution. Without them, you can’t sell.

Enterprise companies that are willing to share brand and distribution reduce the capital requirements of startups and help to win crucial first customers in overseas markets. Often, enterprise scale companies provide the exit for the founders. Germany’s multinationals play a key role in its model.

Beyond Superclusters

Consider Canada’s dismal record of translating health science discoveries into commercial products. The root cause is the lack of domestic multinationals and the brain drain of go-to-market talent. When a foreign multinational buys a Canadian life sciences startup the loss of product management, marketing, sales, and executive leadership is far worse than any loss of tech talent and intellectual property. Most go-to-market skills are learned by doing, not in a classroom. The new parent company HQ vacuums up the best talent and the company stops hiring these skill sets in Canada.

Over time and repeated buyout cycles, the ability of the ecosystem to bring promising tech to market at scale is crippled. This has been going on for over a century. Don’t believe me? Read Ideas in Exile. The Centre for Commercializing Antibodies and Biologicals was created expressly to close this gap. Several other national centres of excellence focus on commercialization in other pharma segments. This is good.

The expertise to design, finance, set up and run medical device production lines and global sales and marketing is very different from that required for pharma. There is only one centre focused on medical device manufacturing. And there’s ReMAP, to cover all other aspects of advanced manufacturing.

Medical devices are also different from automotive, the backbone of our manufacturing expertise. Production runs are much shorter, tolerances are higher and factories have to be much cleaner. And unlike automotive, there is no Linamar, Husky or Magna. For medical device startups, access to global brand and distribution and patient capital is much more challenging.

A modest proposal

Here’s what I recommend:

  • Study our manufacturing winners and stop obsessing about the losers.
  • Build our global brand around the winners.
  • Get much better at adapting the German model

Geoff Foulds advises tech companies on go-to-market strategies and tactics. He works with Global Advantage to map innovation ecosystems. He was a founder of Input Technologies in the 1990s, which commercialized Centres of Excellence research into computer supported collaborative work. Currently, he is at work on a new project to make music more accessible to people with dementia.

Other News

Events For Leaders in
Science, Tech, Innovation, and Policy

Discuss and learn from those in the know at our virtual and in-person events.

See Upcoming Events

You have 1 free article remaining.
Don't miss out - start your free trial today.

Start your FREE trial    Already a member? Log in


By using this website, you agree to our use of cookies. We use cookies to provide you with a great experience and to help our website run effectively in accordance with our Privacy Policy and Terms of Service.