Economic Complexity of Tulsa

A year or so ago, just I started tilting at this particular windmill, I happened across an interesting study from Brookings on “global cities”… and to me the interesting aspect was the middle-US cities on the list, including San Antonio and Kansas City as “American Middleweights” plus Dallas, Austin,  and Denver as “Knowledge Capitals”.  Flipping through again today, I can’t help but notice Trade, Innovation, and Talent under Enablers, which resonates nicely with the innovation topic near the start of this series and the learning networks of late.  Of course trade is one of the keys to the growth goals, and we’ll talk more about exports here shortly.

If you’re interested, this study can be found here:
https://www.jpmorganchase.com/corporate/Corporate-Responsibility/document/gci-redfining-globalcities.pdf

To me, this suggests two obvious paths for our “learning networks”:

-          Connect our educational institutions and appropriate industries purposefully to the nearby knowledge centers of Austin, Dallas, and Denver.

-          Connect our business and trade infrastructure, including local and state gov’t manufacturing agencies, to Kansas City and San Antonio.

The cities in the Brooking studies already have purposeful plans to leverage the Global Cities initiatives somewhere in the Chambers of Commerce and other business groups, so outreach and liaison focusing on joint success should be the goal.  Of course, this means we need to be very particular about what each outreach area should be, as it needs to be:

-          An innovation or market adjacency for us, or both, so it’s attainable for us

-          An adjacency for the target city, so it’s attainable for the other city

-          Ideally, arranged so the highest value parts of the partnership product end up in Tulsa

Which brings us to picking product and tech niches.  We know from past discussions that two fruitful areas for innovations are:

-          One-step-up tech adjacencies – building on what we know, standing on the shoulders of giants, and so forth

-          Remix opportunities, which are adjacencies formed by putting together two separate (not necessarily closely related) technologies or products

If we reach out to pull in tech from Knowledge Capitals then there should be fertile ground for some new inventions and innovations, plus if we can find outflow channels for new products through the Middleweights then there should be novel opportunities there too.

So, we have a basic vision, hopefully some creative tension to drive action and invention, and the beginnings of a strategy to grow our people-network connections, foster innovations, and expand sales.  I would submit that this alone could provide a path to success, but I believe we can do better, if we can purposefully bias our actions toward high-value targets. 

This brings us to the point of this entire series, and what I believe represents our opportunity to tilt the balance toward success.  <drumroll, please>

In his pioneering work on global trade, and associated work since, Hidalgo built a database of the relative product complexity and economic competitiveness of a wide range of products.  He also built graphs showing correlation between products in markets, which represent both technical and market adjacencies, and the networks of people/companies/institutions which support them.  If we are to be successful, we need to climb our way up the economic complexity value ladder, purposefully seeking to bootstrap ourselves from our current success points to better points.  Much as we may want to leapfrog way up the ladder, this mostly won’t be a path to success as there is simply too much build; we must climb up one rung at a time, but always moving upward.

The first step, then, for Oklahoma or Tulsa, is to determine where we sit today.  Recall that this is the necessary anchor for the creative tension part of vision:  establish the current point, set the vision point, and the delta drives innovation.  Here is where I need help, and where the Chamber or other sources of information could probably help.  

A quick Google yields five core industries for Oklahoma:

-          Oil and Gas industries, especially production

-          Data centers for information and finance

-          Transportation and Distribution

-          Agriculture and Biosciences

-          Aerospace and Defense, especially aviation maintenance, but also munitions

The Hidalgo index helpfully yields ratings for products in each space, and many more, with a high value of about 3.0 (manufacturing of machine tools) and a low of about -3.0 (oddly, for uranium and thorium mining).  Some values:

-          Crude Oil                    -1.08

-          Lubricating Oil            1.34

-          Petroleum Gases         -1.06

-          Misc Wheat                 -0.68

-          Flour                            -1.12

-          Aircraft Part                .50

-          Fertilizer                      -.70 to -.80

-          Computer parts           1.3 to 1.9

-          Munitions                    .3

-          Sausage                       About 0

-          Chromatographs          1.79 (this is a product my company makes)

-          Instruments for measuring flow 1.53 (this is the rest of what we make)

-          Trailers            .65

-          Data processing devices   1.0 to 1.4

Of course, the Tulsa industrial and technology base are far more varied and nuanced than this simplistic list, and to make this notion work I expect we will need to dig deep and be as precise as we can.  Unfortunately the list includes only products, not services like distribution and data processing (but think back a few sessions and recall that products contain and replicate information and knowledge, which is scalable and durable, versus services which are fleeting and scale-limited).   Still, I think I see in the data above why we struggle as a state to be economically competitive, and we could make the case that to a degree we suffer from a Resource Curse of our own.  Even within major niches like Energy, we should focus our growth plans on the most valuable aspects of it. 

Enough for today – there is still more to talk about in terms of correlations and adjacencies that Hidalgo uncovered in his research, and which I believe will be pertinent for our discussions as well.

Accelerate!

For those who think that one person or a vocal handful can’t make a difference, take a look at chess in St. Louis. One person with a vision has been successful at getting St. Louis on the map for youth and professional chess competitions. Sure, I know chess is a bit geeky and a small feather in a big city’s cap, but for a city with as many social issues as St. Louis it’s had a better return on investment than many initiatives. Thanks to Sinquefield (a millionaire with a purpose), St. Louis now appears on top-10 chess cities, not just in the USA, but the world:

https://www.chess.com/article/view/the-10-best-chess-cities-in-the-world

Anyway, I say we agree on the vision, pursue some analysis (crowd-sourced, I hope -- I don't have all the answers for sure!), and then take some purposeful action. A while back I did a short talk where I said what I’d like to see as a process for individuals to get involved in making Tulsa a better place. I sketched out a simple diagram, in a circle of four elements, with “Iterate” in the middle:

- Participate – get out and do something!

- Facilitate – help get others going

- Collaborate – work with like-minded people

- Integrate – join forces to promote broader progress

- Iterate – do it again, better

And finally: Accelerate – do it faster

Yeah, I know, I was on an “-ate” kick at the time (there are a gazillion words that end with –ate, by the way – you could make all sorts of catchy lists), but I still think the notion is solid. Get out and do something, work together, get organized, then do more. The angle I’ll add this time is that clearly it would be better to have a vision and a strategy, plus a good team, and that’s what I’ve been working to better understand since. I’d be happy to discuss goals for a bit, and then next time get into the meat of fostering purposeful city-wide innovation. I believe we can do this.

Where from here?

So what to do with these insights? I submit we need to leverage what we know to make our nation, state, and city more competitive. But let’s start small – what can we do with Tulsa?

First, if I had my druthers, I’d host a dinner party with Eric Ries and Cesar Hidalgo, and maybe invite a few others like Malcolm Gladwell and Angela Duckworth. Cesar could provide insight on how to analyze Tulsa’s current capabilities and create a vision, Eric could provide the insight into small bets and experiments, Malcolm could help us connect and tip it over, and Angela could give us determination. With a team like that, how could we lose? Anybody up for arranging this for me, please?

But I’m getting ahead of myself. First let’s agree on what we want to do: I’d say we get Tulsa out of the doldrums and on a path for clear competitive differentiation, and even strive for disruptive success, compared to its peers on the US and world stages. I want to see:

- Growing high-tech and high-value employment for a sustainable white-collar workforce, including self-employed contractors and start-ups

- Intentional development of a strategically valuable skill base that provides long-term advantage

- A solid employment base for blue-collar workers, including trades and craftsmen, with a viable path through the coming AI/robotics/automation revolution

- Expansion of a solid local tax base for great schools, strong public infrastructure, and vibrant communities

- Solid opportunities for students, artists, and entrepreneurs of various sorts

Does anybody have better goals? Other ways of stating them?

Probably we should also be clear on the current state of our city, complete with weaknesses and shortcomings, and we saw early on that it’s the gap between current reality and future vision than helps generate creative energy. From there, I envision an intentional plan with clear progress metrics, yielding an approach that starts strong and gains momentum from there.

Some will say that all cities want something like this, and I would agree; however, most of them don’t have a pragmatic and reasonable plan. Most have some good ideas, some idealistic notions, some idiotic but well-funded concepts, and a bunch of tax dollars chasing the latest big company that dangles jobs. To me, this seems like a recipe for a haphazard drunkard’s-walk toward success, and it just doesn’t seem all that difficult to have a better focus and expect better results. This would be especially true if we do it based on the concepts I’ve discussed earlier here, leveraging these and other similar ideas and research by really smart people. We don’t have to be brilliant: we just need to listen and think like entrepreneurs, and work together to push in more or less the same direction. 

Networks of People

Today we’ll talk more about networks of people. 

Networks of people grow around the purposes of the entities that hire the people – no big surprise there.  Common type are quite familiar – governmental/regulatory agencies, educational institutions, and companies.  Of course there are also churches, charitable organizations, and social organization like Masons and commerce chambers, and on another tangent you find political parties, action groups, sports leagues, and hobby or special-interest groups for anything you can imagine.  One really good organization, with a network of teams, and which gets lots of scrutiny, is the US military; they are REALLY motivated to be good at adapting and remaining capable.  Note that each person typically is part of multiple networks, and the interconnections between networks is important to how organization effectively work. 
Each network is also a system, with its unique purpose, resources, and goals, and like all systems they attempt to maintain some sense of equilibrium and to adapt to their environment.  For many networks of people – teams – learning and adapting are important capabilities, and yet it’s often not easy.  Good teams are generally good at adapting and learning, and a team can gain new capabilities in several ways:

-          Add new skills to the team by adding new people

-          Add new skills by forming new connections to other teams

-          Add new skills by improving skills new members

-          Change the environment so adaptation isn’t needed after all

As part of the learning/changing process, teams often do trials, and Ries would advocate for purposeful, modest experiments with clear intentions, solid analysis of results, and further adaptation.

So where is this going?  That’s the next point:  our economies are an amalgamation of lots of teams, and the capability of any country to be competitive on the world stage is a function of the effectiveness of its companies and other institutions.   The competitiveness of the products largely determines income opportunities for the country, and Hidalgo’s key insight is that the embedded complexity of the products impacts the ability of a country to make them.

Way back at the beginning of this series we talked about products embodying knowledge, and enabling the leverage of the skill of their builders by their users.  For all but the simplest products the embodied knowledge is the output of a team, or teams of teams, and not just individuals.  The level of embodied knowledge is thus a function of the skill of the team that builds it, and the ability of a country to build products is limited to the sorts of skills they have available.  This means that the long-standing tenet of economics, that manufacturing of products moves to the area of lowest cost, is not accurate, or at least is not complete:  manufacturing will only go to locations where the necessary skills are available, else the local cost must also reflect the cost of gaining the skills, and the time to do so.  If we revisit Metcalfe’s law, a large network of resources will have more capability and value than a small network, and I’ll add a corollary that a network of better resources will offer still more value, and a second one that a network that grows in scope and in strength of resources will expand in value exponentially.

What follows is that complex economies must have complex, differentiated skills, and associated teams of workers/scientists/regulators/etc.  Simple economies can only make simple products, and this goes a long way toward explaining the “curse of oil” and other “resource curses” that have plagued third-world countries over the centuries.  This pattern is not merely notional, as Hidalgo performed quite extensive regression analysis of global exports and national GDP at the granularity of common product codes, and obtained solid statistical results. 

Hidalgo then went further, projecting which nations should have GDP upside based on the capability of their economies, and which are already maxing out their potential.  In the years since, others have followed these projections and the natural evolution of economies, and the theory seems to hold.

Next round we’ll chat about how we might leverage these ideas, and move from talking about the past to talking about the future.

The Power of Networks

The past week we’ve mostly talked about relatively mainline business and systems engineering concepts. Many people may not be familiar with them (the world is a complicated place after all, and we all have our little islands of knowledge), but the recent topics of this thread are pretty well circulated and have a fairly broad base of adherents, and probably some critics and detractors as well.

Today, we’ll stretch things a little, into a world of cross-overs between network theory, human psychology, systems engineering, and maybe economics. Networking is a pretty basic concept, and the term is broadly used, to the point that the meaning is no longer precise. There are telecommunications networks that transmit digital data between a vast variety of servers, switches, data stores, your house, and you (at least your phone, your headset, your watch, and maybe your flesh-embedded monitors – but not quite your brain directly….yet). I like this sort of network because (a) it’s nice and logical, (b) it paid rather nicely for much of my professional life.

There are also networks of people, and these are a LOT messier, but interesting as well. Gladwell talks about networks of people made of three sorts of individuals: mavens (experts on a topic, but generally not well-connected), salespeople (good at advocacy and driving change), and connectors (people who know a lot of people). These theoretical people are ideals, and reality is more of a continuum, so none of us would fit such categories perfectly even presuming his categories are correct. Still, the notions of “connectors” rings true to me, and I have a couple of friends here on FB who I peg squarely in that category. The same is true for mavens, and for salespeople.

A sharp guy named Metcalfe devised a theory that became Metcalfe’s Law, which says that the value of a network is proportional to its size. With a little thought you can convince yourself that this is true. Remember fax machines? What value has a network with one fax machine? Yup…zero. How about two, or ten? Not very much. How about 10,000 or a million? Lots of value. The same became true for e-mail a generation after faxes, and more recently for social media like FB too. Bigger networks have bigger value, and quantity is a quality all its own.

What about networks of people? A person, even an expert, has modest value by themselves, and they may struggle to find problems to address, and people with problems need to find them, and most hard problems requires a group of specialists working together to solve them. Now we’re back to talking about teams – and what is a team but a group of people with varied skills working together for a common goal? This means it’s also a system, by the way - a group of diverse components functioning together to achieve a common goal. And it’s a network, given the interconnection of individuals. So, a team is just a network of people that functions as a system.

As it turns out, teams of people are reasonably complex systems, and such systems pretty much always manifest emergent behaviors, meaning that they do things and have characteristics that weren’t intended. Any manager will recognize the truth in the system engineer’s adage, “All systems exhibit emergent behaviors, most of which will be undesirable.” One of the good characteristics, though, is the ability to learn. Networks of people – teams – learn new skills and new ways to solve problems a bit differently than individuals alone, because they have broader connectivity to other people, a broader collective experience base, and can leverage broader resources. It is this ability to stretch, grow, experiment, and learn to achieve problems that provides a team with value.

So if Metcalfe’s Law says that bigger networks have more value, and people interconnect as networks, why not have huge teams? Today, people are individually and collectively more well connected than ever before, and it is not clear what this will mean to the composure and effectiveness of future teams. It is clear, though, that huge teams have different dynamics and require different structures and processes than little ones, and part of system engineering is to break down big problems into smaller ones that can be solved by modest individual teams. Silicon Valley, (and Ries’ startup culture) embraces Amazon Bezo’s the “two pizza team”, even though most larger companies may have many such teams.

Of course a network may have sub-networks, systems may have subsystems, and companies may have a hierarchy of teams – that’s what an org chart displays, after all. What determines the optimal sizes? Well, one guy a while time ago took a crack at it, and his name was Dunbar. His theory is that people can keep inter-relationships straight for about 150 people, and closer relationships with smaller group sizes. Many of you who have worked at varying sizes of companies, gone to small or big churches, or interacted with small or big vendors, will likely agree at least in principle that size does indeed matter. A few companies, notably the tech company WL Gore (inventor of Gore-Tex) embraced the Dunbar Number explicitly by keeping their plants sized no larger than about 150 people; if a company grows too big, they simply split it up. So far, it's working for them.

Tomorrow we’ll pick up with types of people networks, and how those interrelate in the real world. We might start straying into economics a bit, too.

Power of Compound Growth - for Innovation

Lean is a popular approach followed in manufacturing to drive out waste and improve performance.  Lean got its start with Deming in Japan, as The Toyota Way and Six Sigma proved their worth.  It’s good approach for just about any business or group, not just manufacturing firms and teams.  If you manage a group and don’t know Lean, then learn about it.   Eric Ries took many of the same ideas and applied them to running startup companies, wrote a book and built a consultancy on it, and that’s the hot approach today from the denizens of Stanford and Harvard, much of Silicon Valley, and recently GE with their FastWorks initiative.

Eric’s book The Lean Startup is worth reading if you’re entrepreneurial at any level, but in a nutshell his approach is to embrace an experimental approach to running your startup, devising meaningful but small tests to hone your assumptions, refine your approach, polish technological solutions, and court your customer base.  By using small tests – placing small bets – and learning rapidly, you can make rapid tiny improvements that over time fuels significant growth.  The book Small Bets suggests much the same approach.

The rapid experiment-feedback approach results in a learning organization, and this is a critically important point.  If we think back to effective teams, and consider what great teams do best, you’ll see that they learn and adapt, and find ways to succeed.  Great teams collectively do better than great individuals can do individually, and all a startup really is another team with lofty goals.  What better way to run your great team than to have it think like a startup, with that heady mix of high expectations, significant risk, tight constraints, time pressures, and (hopefully) large rewards? 

The Startup Way and Small Bets both discuss the value of compounded growth, but it’s Hidalgo’s How Information Grows that adds a theoretical context:  when systems iterate with repeated decisions or choice at each cycle (termed preferential attachment), the result is a power-law distribution, with winner-take-all scenarios like Facebook and Google as an outcome.   As you might intuit, companies (or any system) that can either cycle faster, or make more improvements per cycle, will win out over time, even if they start from behind.  If you’re my age you’ll recall using both established heavy-weights Alta-Vista and Yahoo when searching for news about that fledgling start-up called Google.

Next time we’ll shift a bit from talking about teams and companies, and chat a bit about broader interconnected networks of people, and the understated and underappreciated power of such networks.   Until then, here’s a some Cliff’s Notes on the current realm of business books:  Devise purposeful experiments in your market, make small bets with direct customer feedback, fail forward by learning from wins and losses, leverage meaningful metrics to account your progress, iterate quickly, and have fun!

Optimization Algorithms

When discussing optimization algorithms, many fall into the general class of hill-climbing algorithms:  if you want to get to the highest elevation, start where you’re at, look around, and take a step uphill.  Then see again – lather, rinse, repeat.   This is exactly the approach we talked about yesterday, making many small, incremental improvements.  It’s easy, it’s robust, and mostly it works.

The trouble with hill-climbing algorithms is that once you get to the top of your mountain, there you stay.  If you started on the slopes of the biggest mountain, then great, you’re at the highest summit!  But what if you started in the foothills?  You climb the mountain, and stand proudly on the top of a modest hill, with no way to get higher.

This is where a different class of algorithms comes into play, and this is why small, disruptive companies can get ahead of big companies with their teams of plodding hill-climbers topping their little foothills.  Actually, there are a bunch of algorithms that allow you seek further afield for higher peaks, such as jumping randomly to any point on the map and seeing what the elevation is, or overlaying a grid and walking from intersection to intersection and mapping the elevation.  The trouble with these algorithms is you don’t know when you’re done, and you can spend a lot of time jumping and checking more points.  If you’re short on time, and want to be pretty sure you get to the top of a relatively high mountain, what do you do?  Hold onto that thought….

Curiously, certain type of organisms called protists exhibit an interesting reproductive strategy, in that they change from asexual reproduction to sexual reproduction depending on the environment.  When the environment is benign, the organism reproduces asexually – it’s fast and efficient, and the chance of the new offspring to be well-adapted to the environment is high, and the population of identical genes expands with a small rate of change caused by random mutations here and there.  But what if the environment changes and becomes hostile?  In that case the organism switches modes, and uses sexual reproduction, which mixes up the genetic makeup of the offspring, increasing the chance of finding new good matches (this is called the Red Queen Hypothesis).  Switching modes is the useful insight here, and that’s the behavior to emulate.  Genetic algorithms in software today do this, with a mix of random mating combinations that drive large changes and random minor mutations that create small changes. 

A few decades back a Russian engineer working on optimization of microprocessor circuit routing hit upon a similar approach after talking with a mathematician who was modeling glass annealing behavior.  In glass annealing, reheating and slow cooling is used to heal micro-fractures in the glass and reduce entrained stress, with higher temps healing larger cracks but causing more stress, and lower temps reduce stress but don’t resolve cracks as well.   The engineer devised a novel approach now called simulated annealing, where a computer performs a number of simulations, first with “higher temperatures” – large changes in component layouts, and then cycles of “lower temperature” – small changes and lower stress.  For each run, the computer performed a quick routing of interconnections, and weighed the result against previous tries, keeping the best. 

For both types of algorithms there is a necessary intermediate consequence:  when there are large changes, most new offspring and most simulation runs yield really poor results, and only a few yield promising results, so there must be a culling step where only the best survive.  Of course, for genetics (and genetic algorithms) this is the Darwinian step, and for simulated annealing it’s a fitness function weighting step.  We will need to emulate this behavior as well.

Now let’s go back to our hill-climbing problem, and emulate the behavior we’ve just seen:  perhaps we can cast a rough grid or take a few random hops to sample the uncertain topography, and of those data points, pick the best one or top few.  Around those, take smaller hops and repeat.  Throw out the valleys, take the biggest peak so far, and shift modes to a hill climb.  Voila!  You know when you’re done, you’re at a high peak (probably the highest), and you didn’t spend a lot of time hopping around.

What’s the lesson for us, at work or as individuals?  If you want to make big changes, you have to accept that failure will be ever-present as the environment is hostile to your experimentations, and thus your goal should be come up with an approach that allows you to experiment quickly and cheaply, creating and culling a lot of permutations, and the learning by taking the best and refining again.  Throwing many samples away, sometimes taking steps backward, and having to survive many improvement cycles are all hard for us to stomach, but that’s what we need to do. 
This is exactly the general approach espoused by the Lean Startup, and Little Bets, and earlier by Peter Drucker in the Effective Executive, though they do a more thorough job of explaining (and made a lot more money) than I just did. 

If you or your business is struggling in a harsh environment, maybe you should raise your annealing temperature a bit.  This is a thought we’ll revisit, but first we’ll take a little deeper look at Lean, and the precepts of Lean Startup.  Until then, ponder on the thought that few people and few companies really try to take chances at all, and fewer still do so with a clear vision and a decent plan.