Noise vs Signal in Tech Investments

Each investor has a different style. Some are macro-thinking and studying business cycles, while others are product-driven and focusing on founding teams. Between the two opposite styles, Jinjian Zhang of Trustbridge Partners has created one refined and unique individual signature. As an intelligent scholar in sociology and demographics, and as someone who’s also very street-smart in reading people, he is indeed one of the most impressive investors and systematic thinkers I’ve seen in years.

More importantly, he’s well-positioned to be one of the leading VC investors from China in the next 20 years. Trustbridge’s portfolio companies include such well-known names as Meituan and 360, and two of the investments that Jinjian led — LAIX and NEW have gone public on NYSE in 2018. Here, I’d like to share with you a recent speech by Jinjian, which well demonstrates his philosophy in investments.

We are living in an information explosion era. It is getting harder and harder for people to extract essential information quickly and effectively. But if you look at a longer period, say 5-year, you’ll see the full picture of the life cycle of many things clearly.

Human beings are able to achieve more and more, not because we consumed all of the massive and noisy information. On the contract, it’s because we can spot what is the most key information to focus on.

I was analyzing satellite images for 7 years during my undergrad and graduate studies. Basically, it all narrowed down to one thing — what is signal and what is noise? To be more specific, my job was to build a machine learning model to filter out the noise and capture the real signal. Later in my life, I found the methodology is surprisingly applicable to investments.

I. Signal and Noise

During my graduate studies, I published a paper on a magazine under the Chinese Academy of Sciences. My project was to identify stranded vehicles in deserts from analyzing satellite images in order to rescue the people trapped inside. Here’s a picture from my paper.

The small yellow frames you can see on each of the pictures are the vehicles identified by the machine learning model. For example, the frame in the first picture was a small car on a desert road. The real challenge was that all the photos are HD with millions of pixels, but the cars only have a few or no more than 20 pixels each. Basically, I was looking for a needle in a haystack.

It inspired me that it is the same with real life — we live in a desert full of redundant information, and we are desperate to find our own oasis. Moreover, I feel the same way when it comes to investments, repeatedly.

By 2012, I joined Trustdridge to start my career in investments. Buried myself in numerous sell-side research reports, I started to set up parameters such as supply chain, demand forces, distribution channels, etc. to evaluate companies. However, my parameters gradually increased from 5 to a dozen to more than 50!

I realized the more I understood about an industry, the more parameters would pop up. I decided to slow down to think through my methodology. Finally, seven years’ of scientific research experience led me to my Eureka moment — only noise takes many different parameters to describe, but for signal, you should be looking for the common pattern.

When you learn something new and find it harder and more subdivided, Congratulations, you have spotted the noise! As you keep going, you are able to spot common patterns from all the differences. This is when you reach your target.

My advisor used to rephrase that noise always varies in different forms, but signal extracts similarity. A typical noise such as white noise is random on every signal-point. However, when you collect all the points together, there is a Gaussian distribution.

You should never study anything from just one single perspective because you’ll get lost and drowning in a pool of parameters.

Here’s a story of my roommates.

My Roommate A had a secret crush on a girl. Luckily that girl secretly liked him too. One day after a ball game, that girl came to him sending a love vibe: “Are you hungry? I bought you some chips.”

All of a sudden my Roommate B broke into the conversation: “Those chips have been on sale. I was gonna buy some too.” Thus, unfortunately, my naïve Roommate A thought the girl only bought the chips because they were on sale, not because she had a crush on him.

At that moment, signal and noise were mixed up. Roommate A got nothing but a meaningless “wave band”, to put in engineering terms.

This was a sad lesson. We often hear complaints that how could we miss out on the changes in the world? How come we didn’t realize that a new company is a unicorn? From my story you know the signal is right there, you are just distracted by noise from Roommate B who continuously feeds you Gaussian noise.

You must constantly keep observation and awareness in order NOT to miss your true love. The key step is to kick your Roommate B (noise) out and live with your girl (signal).

II. Signal and Noise in Investments

Then again, what are the signal and the noise in investments?

Some believe 5-year is a business cycle. Then, what keeps changing and what stays the same in that period? What do you need to reconsider every 5 years? What has the least noise?

The first signal I found in investments is demographics.

In 2012–2014, I spent 2 whole years studying through the demographics of many different countries at different times. The thorough study enlightened me that all my previous investments research was nothing but noise.

Let’s begin with these 4 countries — China, Japan, Germany, and Russia. I found one thing in common — population structure gap.

There were two population booms in all four countries. Why? Because those countries were all the major belligerents during World War II. The reason is that the war caused a loss of one generation, and in order to prepare a war, the country encouraged people to give birth to more babies. Besides, suffering from a defeat and economic downturn, people have no jobs or entertainment, so they made more babies…

Take a look at Japan. This country’s population structure in 2000 is very similar to that of China in 2020. For a time span of 20 years, why the population structure could be almost the same in two different countries? Shouldn’t the two countries face the same problem at the same time as they were both the belligerents in World War II?

That was because Japan, when they initiated the war, seriously overestimated their power to dominate the world. When they occupied the 3 northeastern provinces of China at a lightning speed, they were carried away by the early success and proposed a 100-million population plan in 1939. They believed that if they wanted to colonize the whole of China, their current population was way less than enough. Therefore, the Mikado of Japan encouraged his people to multiply in an extremely aggressive way. Therefore, before the end of World War II, Japan was united in a furious “Human Being Creation Movement”.

China, on the other hand, didn’t launch the similar “Hero Mother” campaign until 10 years after World War II, which postponed the first population boom in China for almost 20 years.

World War II caused a huge impact on the world and in different ways and phase to individual countries. Most of the time we don’t see the direct connection. But a butterfly flapping somewhere may eventually get you caught into a tornado.

“Family Planning” was the direct result of the baby booming. It was written into the Chinese Constitution in 1978 and made to be the state policy in 1982. It created a world unique generation known as the “Only Child”. Almost every kid born after 1983 is an only child. They don’t have any brother or sister to fight with when growing up, so they don’t get much practice to harmonize and compromise during confrontations. This implies that an only child tends to have a big ego.

Imagine an egotistical generation first enter the professional world with their peers. They would be shocked to realize they are not “the only one” anymore. They left their comfort zone. They found it hard to adapt to the environment working with others. After 2005, there were a series booms of small individual merchants around 22, benefiting the take-off of Alibaba’s Taobao.

In the meantime, another interesting data point popped up — the divorce rate in China tripled from 2015 to 2017.

Before 2005, the divorce rate kept a steady level and even experienced a drop between 2000 to 2005. After 2005, the Only Child generation is getting married. However, a lot of them found that they could not accept living with someone who shares different personalities, and most importantly, they were not capable to compromise.

On the surface, the only child policy was no more than a birth control measure. However, it carries deep implications on the society level.

As investors, we ride the macro waves of the very era. We see single dots and we connect them into lines to form trends. We see through times, and the process and logic behind them.

The next stop: India.

If you take a close look at the population structure of India — it’s shocking. From this chart, you find India was never a part of globalization. Regardless of the two world wars, globalization or WTO, the structure is so smooth with steady growth, without any gaps.

When we look at the Indian market, we always wonder: what are we investing in?? Do you really understand your client? Which years were they born? What happened in those years? What was the context of their growth experience? What kind of personalities will grow in them?

Of course, no one is the same. However, the same generation will share a lot in common. When talking about investments, we should go back to what I mentioned at the beginning: we look for the similarities rather than the differences. We chase after the signal but filter out the noise.

You might wonder what the connection is between population structure and economy. Are these similarities and differences really have inferred value in the domain of social science? After all, can they help you generate investment returns?

To answer these questions, I began further digging into demographics by reading a pile of books. Here I recommend The Demographic Cliff written by a well-respected demographics scholar Mr. Harry S. Dent. He conducted comprehensive research on the American population structure and found out that although each family differs, they share similar expenditure patterns at some points. He put together the similarities and composed the below chart.

The chart tells you that kids from an average American family wish to buy their first apartments at age 26, a transitional house by 31, and a better house by 42. When they turn 46, they want better furniture and should have saved up the college tuition for their kids.

Dent found that in the American population structure, 46 years old is the peak of the average family expenditure. So, if we push the fertility index forward for 46 years, accounting for immigration, we can predict the expenditure peak of an average American family.

On a side note, let’s revisit the trend of the baby boomers: starting from 1934, the birth rate began to speed up through 1937 and peaked in 1961. Based on the 46-year-old figure, the expenditure peak shall be 46 years after the birth peak years of 1937–1961.

Therefore, strong economic prosperity could be expected from 1983–2007. With this prediction, Dent went to check the Dow Jones Index and found the data almost perfectly matched his prediction. Who would imagine the birth rate can be used to predict stock index?

There are even more correlations between demographic trends and economic activities than we can elaborate here. Just remember if you look into the worldwide countries’ economic data and study their population structure, birth rate, marriage rate, and divorce rate. You will definitely find that they are closely bound up with your life track.

Alan V. Oppenheim mentioned Fourier transformation in his book Signals and Systems. Fourier transformation is defined that any wave could be decomposed into an overlay of different periodic functions. In short, life is composed of multiple cycles. There are small ones and big ones, and it is important to understand the big cycles.

Today globally, food delivery platforms such as Meituan and DoorDash are expanding quickly. People are trying to figure out why food delivery has been booming. The ordinary explanations could range from the rise of micro-payments to the penetration of smartphones. Ultimately, they were all only looking for the reasons on the technical level.

On top of that, how about the reason for the population and social structure? To be more specific: without science and technology development, without a smartphone, WeChat and Alipay, would there still be online food delivery business? Was there a bigger trend? Let’s look at some more developed countries to see if they experienced the same.

I would like to take Japan as an example again, as there are better data integrity and longer time horizon of data.

The above chart shows the growth trend of Japan dining-related industries from 1963–2007. Among which, two lines showed ultra-high growth speed from 1963–1990 — at 14 and 12 times respectively. One is “Cooked food”, that is, the Bento culture today; the other is “Meals outside the home”, which is to dine out at restaurants.

So, during the years without smartphones, the two categories enjoyed the fastest growth rate are take-out and dining out? Why? Is that because of the influence of tech? Or because the delivery drivers’ business is rising up? Let’s see what happened to Japan’s social structure during that period.

According to the data above, during the 20 years from 1970–1990, the ratio of Japan’s “living alone (1 member)” and “Married Couple Only (2 members)” population increased by 9.8%. Now you know you order a take-out because you are alone (1 member) or you live with your girl/boyfriend or partner only (2 members). Larger families with kids and elders (3+ members) are less likely to order take-out.

In most of the time, we believe science and technology arouse revolution. In fact, the revolution is rather the result of social structure transition and is growing up with, for example, a boom in divorce and singles rate. Due to the quick expansion of single family and DINK family, the demand based on take-out and dining out was strong. This demand increased 14 and 12 times during only 20 years in Japan. While in the same time period, Japan’s whole consumption index only increased 7 times.

Investors tend to consider more about what revolution would science and technology bring about. However, we should trace back to the origin of the signal: what’s the change in the demographic and social structure? What does it mean to have more and more single persons in society? What is the real signal you are looking for? All the answers are well-rooted in that signal. It’s never about the noise.

III. How to deal with the world of dynamic noise

Here comes another question. In this society driven by population and social structure, how could we take advantage of signal and noise? How on earth can we use the theory?

I am definitely not the first to draw correlations between demographics and investments. There must be others using this theory as well. Many professional product managers and entrepreneurs are using it to create products and business models as well.

I tried to find supporting evidence in the economic field and found Fischer Black. Black is a legend of Wall Street, after whom the famous Black-Scholes-Merton options pricing model was named. One year after he passed away, the Nobel Economics Prize was awarded to the other two living co-authors.

In his lifetime, Black invented many quantitative pricing models, which were the basis of modern finance. He issued a paper titled Noise in 1986 which used the whole 8 pages to elaborate on what is the noise in the financial world.

He wrote in the paper that value investors are not able to prove their capability in the short run due to all the noises. Noise urges the noise-traders to make deals but their actions, in turn, became noise, which spread and further affect more noise-traders. The society then turns into the clash of the value-traders and the noise-traders.

Most of the time, there are way more noise-traders than value-traders. Therefore, stock prices fluctuate in the short run, and the fluctuations can be in the completely opposite direction of the intrinsic values. Black asked one question at the end of the paper — do these noise traders have no value at all? As he answered himself — no.

It was the noise-traders who provide liquidity to provide a functional and efficient market. If we are all value-traders, who will sell? And who will buy when we are selling?

Therefore, don’t complain about the noise. Forget about it when people tell you “your competitor startups raised a new round within 3 months” or “while you are still sorting it out, your peer VC invested in 3 other deals.”

It’s a jungle out there. Noise is everywhere in your life and career. I only hope you won’t be bothered by them because they are nothing more than noise in the end.

To follow Black’s theory, you’ll see that the stronger the liquidity in a system, the more noise there will be. Every science and technology revolution will upgrade the world’s liquidity level.

We used to believe that the liquidity level was raised high enough by the industrial revolution, but it was negligible after we’ve seen the information and the internet revolutions.

Maybe in the future, the blockchain revolution will again extend our cognition of liquidity and noise level, when you can convert anything into tradable assets.

How to invest and start a business in such a noisy world? I say, if we couldn’t avoid noise, we embrace it. I have three principles:

First of all, always keep your mind open. Be sensitive! It doesn’t matter how much the noise there is and what your position is. Upgrade your radar to catch the signal. Radar receivers are getting increasingly flatter and wider because such design allows it to catch even a tiny vibration.

Secondly, think in the long run. A signal wouldn’t appear out of nowhere. Still remember my Roommate A? This prince almost missed his Cinderella because of such noise from my Roommate B. He didn’t find that crystal shoe until after 12pm when all the magic disappeared. He simply didn’t think in the long-enough run. You must have long-run thinking to obverse and understand the signal to catch your Cinderella.

Last but not least, stay alert for changes. At any single point, noise is random and different. When expanded to a long enough run, noises balance out and revert to zero. The only way to catch the core signal is to find those changes.

I benefited tremendously from this theory in recent years, and I hope it will also help you all to become an outstanding filter in this noisy world to find your lifetime signal.