When will technology investments show up in productivity numbers?
The world’s mature economies are highly productive, but the growth of productivity has slowed significantly in the past 15 years. And, except for an exceptional period between 1995 and 2004, productivity growth has been anemic since the 1970s, when compared to the 100 years between 1870 and 1970. Economists have studied this problem in depth. A significant school of thought is that all of the great technologies that have been created in the past 30 years pale in comparison with the great technologies of the hundred years between 1870 and 1970, namely electricity, central heating, running water, indoor plumbing, telephone, and motor vehicles. This is chronicled in detail in Robert Gordon’s book “The Rise and Fall of American Growth.” (I highly recommend this book).
People who ascribe to this school of thought are sometimes called “techno-pessimists,” because they believe all the computer technologies of the past fifty years just aren’t that great when compared to the great technologies introduced in the late 19th and early 20th centuries, at least when it comes to improving productivity. The productivity data supporting this position are indisputable.
On the other hand, I personally don’t know too many people who are not working significantly harder than they were ten years ago. And, when I look at supply chain management professionals, they seem to be more harried than ever. Furthermore, companies are investing in software at increasing rates. If people are working harder and companies are investing more in technology, then what is all of this incremental work and investment going towards? Let’s investigate a potential answer that is not picked up by the data.
Productivity 101
Productivity is simply output per unit of work per unit of time (economists distinguish between labor productivity and total factor productivity, which I won’t get into here). If last year I produced 100 units with 10 units of work and I do the same this year, then my productivity improvement is zero percent. But what if last year my 100 units was comprised of five product variants and this year it is comprised of ten product variants? Did my productivity improve or stay the same? What if last year I provided two ways by which the consumer could attain my product (when, where, and how) and this year I provided four ways? For sure, the amount of work and skill involved increased. The units of variety have doubled. And, the units of variety are not limited to products – there are lots more variety in services – how, where, and when the product is delivered to the consumer. However, according to economists, all of these efforts and investment amount to zero in terms of productivity because my unit output is still 100. This means that everyone is working much harder with much more complex supply chain decisions to provide more consumer choice, but choice does not factor into productivity equations.
Orange Juice Theory
Last year, I published an article on consumer choice and managing supply chains. I randomly chose orange juice as a sort of mundane everyday product to see what kind of variety is offered. At that time, which was November, 2018, a search on “orange juice” on the Kroger web site yielded 110 results; on Walmart 271 results; and on Amazon 684 results. Today, in December 2019, the same search yields 88 results on Kroger, 341 on Walmart, and, incredibly, greater than 7000 on Amazon. I can further filter my orange juice choices based on brand, size, nutrition content, calorie content, special diet needs (gluten-free, organic, fat-free, kosher, etc.), number of pieces, and delivery requirements. Some of this choice is tangible, in that it is product-based, but a lot of it is intangible, particularly when compared to thirty years ago when there were only a handful of choices.
The theory here is that a lot of enterprise investment in the past twenty years has gone towards providing this choice and that while all this choice has value to the consumer, none of it shows up in economic output numbers. Most of the incremental investment by companies has gone towards delighting the customer in the form of choice, but choice has no value when it comes to economic measurements. Let’s assume that consumers value this choice. So, if consumers value something, and companies are working feverishly to provide it, but economists don’t value it, then there is always going to be a disconnect.
Amazon spent $28B on shipping costs in 2018 and is on track to spend more than $35B in 2019. A significant part of the incremental investment in 2019 is focused on increasing its geographic coverage of 1-day delivery (the incremental investment for 1-day delivery is planned to be $1.5B in Q4, 2019 alone). But moving to one day delivery from two days (3-7 days ten years ago) does not count towards output or GDP. Should it? If Amazon based its investments solely on improving productivity, then it probably would never offer one-day delivery.
In fact, providing 1-day delivery is significantly less efficient than providing 2-day delivery. It requires more investment in assets and labor. Therefore, in economic terms, for the same output (a product delivered to my doorstep), we have become less productive. And, as Amazon continues to ratchet up the dial, everyone will become less productive, until some sort of saturation inflection point (discussed below). But I am sure that the workers in Amazon’s distribution and sortation centers don’t feel like they are less productive. The Amazon drivers I talk to deliver between 150 and 300 packages per day. A peak day might be something like 350 packages delivered to 250 locations.
Orange Juice and Microprocessors
Having fifty varieties of orange juice does not increase output versus having one variety. But consumers place some value on this choice. In the Hi-tech industry, economists are able to determine that production of 1 unit of microprocessors using 1 unit of input this year is significantly more productive than two years ago because the processing power of the microprocessor has doubled. However, they are not able to determine that producing 100 units of orange juice with 10 varieties while consuming 1 unit of input this year is more productive than making the same number of units across 5 varieties two years ago. The incremental doubling of variety in this case does not provide quantifiable value in the same way as the incremental doubling of microprocessor speed. It’s an intangible that presumably provides value to consumers but cannot be quantified (or no attempt has been made to quantify it).
This is despite the fact that the complexity of producing, distributing, and managing inventory across this doubled product variety has gone up substantially. I contend that it’s actually remarkable that productivity has improved at all in the past fifteen years, given the increasing permutations that have to be managed. But variety is not like microprocessor speed in the eyes of measurement. If speed doubles, then value doubles, but if variety doubles, then presumably value stays the same. It simply cannot be that all of this effort on the part of companies, and all this desire on the part of consumers, has no economic value.
Until all of this effort to enlighten and serve consumers can be valued, productivity growth is not likely to change anytime soon. The reason for this is that a large percentage of the input investments being made by virtually every company will continue to go towards consumer delight, which is not part of any output calculations. Having said that, there is a future scenario in which traditional measures of productivity will start to rise rapidly again. This is discussed below under “choice saturation theory.”
Intangibles
This year I read the excellent book “Capital Without Capitalism,” by Jonathan Haskel and Stian Westlake. It discusses the rise of the “intangible economy.” “Intangible” is a term used in finance to describe an asset that does not have physical substance. For example, a machine tool is a tangible asset because it has physical presence that can be seen and weighed. Software, on the other hand, is an intangible asset because it has no physical attributes: you cannot see it, weigh it, or touch it. Mature economies such as the US and the UK now invest more in intangible assets than they do in physical, tangible assets. “Capital Without Capitalism” investigates the economic implications of this changing investment landscape.
The book focuses largely on the investment, or input side of the economic equation. Investment counts towards output, or GDP, so they postulate that if intangible investments are under-counted, then so too is output. However, their analysis shows that this under-counting does not mathematically amount to much incremental output or productivity growth. But what about the output side of the equation? What if all the previously-mentioned extra work and investment is going toward creating intangibles (the variety equivalent of microprocessor speed) that are not counted as economic output? The book does not go into this side of the equation.
Let’s go back to the great century of productivity between 1870 and 1970. If there were a sort of Maslow’s hierarchy of needs for US citizens, then almost everyone in 1870 would have been operating at Level 1 – no electricity, no indoor plumbing, no central heating, and so on. By 1970, almost everyone was operating at Level 3. It is indisputable that the productivity impact of the step function from Level 1 to Level 3 was large, ubiquitous, and thus far non-repeatable.
Today we live in the age of the enlightened consumer (a sort of consumer version of Maslow’s self-actualization), where almost everyone can get a tailored experience that was once reserved for only the very rich. And, this tailored experience is becoming more and more finely tuned to individual desires. Much of the incremental investment that companies have made in the past twenty years is focused on providing this finely-tuned experience. Some of this tailored experience is in the form of tangible products – specific product variants that meet the needs of individual tastes. One only has to look at the more than half a billion products for sale on Amazon to understand choice available to consumers. But a lot of what is delivered to the consumer is also in the form of intangibles – delivery and returns when and where it’s convenient. Furthermore, these demands are not limited to consumers; businesses are increasingly demanding consumer-like experiences from their business customers.
Turning the Question Upside Down
The question is not why productivity growth has not improved after all the investment in technologies in the past 15 years. Given the dizzying array of consumer choices available today, the more relevant question is: why has productivity growth not been even lower, or negative? Reversing the question creates an entirely different perspective for discussion.
Let’s say I want to produce a single variant of orange juice. I invest $200M in my plant and another $100M in my distribution operations. Then, let’s say five years later, I want to invest in producing another variant of orange juice. Do I need to invest another $300M in production and distribution? Back in the old days, that is exactly what I would do. In general, products and product-variants were single-threaded through production and distribution. Need another variant? Build another plant or build another line.
Today, in almost all cases, I would leverage the same infrastructure and perhaps make a small incremental investment. One of the key enablers that allows me to do this is software. Essentially what software allows you to do is to create multiple virtual supply chains across the same physical infrastructure. It allows you to forecast all your product variants down to individual demographics, plan and schedule them across the production equipment, manage inventories, and plan and schedule their physical distribution to the customer. And today, software allows you to do this with efficiency at scale across millions of product variant-customer-geography combinations. But remember, according to economists, this multi-million variant output is no more productive than its single variant grandfather.
If we were able to compare the efficiency of today’s multi-million variant output with producing the same output using the historical single-threaded approach, we would find today’s approach to be orders of magnitude more efficient. In other words, it would be orders of magnitude more productive.
Choice Saturation Theory
But what happens when the consumer if fully satiated? In other words, what happens when choice reaches its saturation point or law of diminishing returns? The answer is that the traditional productivity growth rate increases.
But when does saturation happen? Perhaps one day we will get to the point where each individual will be able to get their own unique individualized orange juice. Amazon is working on being able to deliver anything on its website in the same amount of time it takes to deliver a pizza. The medical field is working on personalized medicine based on an individual’s DNA profile. Perhaps we will one day have DNA-level individualized choice in the consumer world, something that would take decades to play out.
That said, it’s also reasonable to think that we will reach a point of choice saturation earlier, say in the next ten years or so. Once we have reached choice saturation, I predict that the traditional productivity growth rate will start to rise. Why? The reason for this is that once again companies will stop chasing additional choice and start focusing most of their attention on the input side of the equation. The race to the bottom in terms of providing additional choices of all kinds will finally be at or near the bottom. We will start to see the cost of providing all of this choice come down substantially as companies leverage technology to discover more efficient ways to provide product and delivery variety. For the past twenty years and for probably another ten years, we have been and will continue to tread water, at least when it comes to traditional measures of productivity. This does not mean that companies and their workers are not in fact much more productive; they are just not much more productive using historical measures.