Sucess of Software Startups: Inspired by Mosaic
I stumbled over some snippets out of "Mosaic" by M. Pabrai. (Here the source & a must read!!! http://outlierallocators.com/2014/02/25/excerpts-from-mosaic-perspectives-on-investing-mohnish-pabrai/)
(Mohnish Pabrai is one of my favourite investors and some kind of personal role model. One thing I share with him is a concentrated portfolio approach. I think it lies in a personal trait to have some kind of entrepreneurial spirit - being able to withstand the stress your company causes to your mind when shit hits the fan. I was an entrepreneur - before selling the startup to an investor I now work for. Mr. Pabrai was running his company before starting his investment career.)
One thing I tried to figure out since 2010 and a major cause to start this blog was to figuring out what causes a few software companies to get big and many to fail? I knew, if I could figure out some few right ideas, I could extract huge returns as an entrepreneur and investor.
Mr. Pabrai describes some really important ideas related to software startups and investing.
If I screen companies, I look for these three qualities. Some further thoughts on this inverse Bermuda Triangle is written by R. Zeckenhauser ("Investing in Unknown and Unknowable") - a must read!
We know that Lower Risk is caused by a Lower Price. A low price can be determined by historically, balance sheet based or free cash flow based.
High uncertainty is not a risk because its impact can be positive and/or negative. Humans are pretty bad in calculating probabilities of outcomes. Even apes are better in some cases. Mr. Munger recommends to add basic statistics (e.g. the works of Pascal) to the lattice of mental models. I like to use some distribution function based valuation. First I kill the company and than add back earnings of business units or values of assets in the order of confidence I have in the number. The confidence of numbers is a distribution function represented by a subjective calculation. The calculation is based on historical data + some subjective quantified numbers of positive and negative outcomes. Being conservative the negative outcomes are overweight - the number gets smaller. I rebuild the company till I have the market cap. I permute the rebuilding order and conservative assumptions. With a few permutation I can guess what uncertainties are priced in and which uncertainties are important.
High Return Possibilities are the other side of the conservative assumption. They represent positive impacts but should not represented in the intrinsic value. Its too easy to wind up yourself.
Sometimes the price seems to be a little expensive. In my experience this happens if a company has very predictable cash flows but high return possibilities. I buy them, because I wont loose much. I think of Microsoft a few month back or Corum Group http://www.corumgroup.com.au/. (At the moment I think I could build an automatic screener for the three.)
Lets see what Mr. Pabrai further has to say. (See also Amar Bhide: How Entrepreneurs Craft Strategies That Work)
I thought about every part (reinvestments, uncertainty, profit, emergence of competitive advantages) for nearly 4 years and I could not see the big picture. After reading Mosaic snippset, I see some stable relations between them: No start-up has a moat. A moat might occur over time. Its better to have a Start-up in an highly uncertain place than one in a certain. Jump to the next niche, when no moat occurs. Every jump might kill a start-up. Every moat and jump behaves probabilistic. In the software industry a moat causes in wide areas a winner-takes-it-all market.
Is a successful software company a result of skill or luck? The answer seems to be more luck than skill. Can I exploit the luck? Applying the weak law of large numbers says to jump a lot, to maximize the chance to get in a winner-takes-it-all market. To jump a lot means not burning/loosing cash for each jump. The low cost jumping is the creative part. In the IT area some on surface technically unrelated topics are closely related dune to same mathematical groundings and vice versa. This can lead to low jump costs.
One strategy to get some insights in the cost of each jump, is to apply the Kelly criterion. It says how much someone should bet for each jump to maximize the long term profit. The problem is that the bets outcome is by and large a long tail distribution (winner-takes-it-all market). Long tails having no average, therefore the bets outcome is not a tangible number. Kelly is not applicable.
In the book "Antifragile" by Nassim Taleb, he says that most long tail distributions in reality are asymmetric. The downside is much larger than the upside or vice versa. As the costs of a jump is limited and the outcome is a long tail distribution, we exploit the asymmetric behaviour. Under such a distribution Mr. Taleb suggests its better to follow processes than goals. This is because long tails outcomes cannot be expressed as numbers, therefore a goal lets say 10 Mio. revenue is totally unrelated to the real market revenue achievable therefore more a distraction than an information. A process is a step by step manual which can better adapt to the real market.
Software product management is another important part of delivering the right product to the customer. It says that a product sold, is more than the technical software. Its a brand, user experience, service, etc. which is sold too. Each jump to the next market should cause as low as capital expenditures as possible through recycling as much resources as possible - the technical software, customers, services, brands etc.
As a side note: Rocket Internet is a venture fund / enterprise which operates in the B2C market. It decentralized the idea generation and technical execution through idea acquisition and in sourcing in their own business units and centralized the administration, marketing- and brand-management in the mother enterprise. The BU profits from the experience in the consumer market of the mother and its cost structure.
In valuation of a software company, I want predictability in earnings before I will pay a fair or slightly overpriced company. This is not possible for start-ups.
Can an entrepreneur use a distribution based valuation to detect promising markets? Not really, because there is mostly no market value beforehand.
Which market should the start up penetrate? The one which has the highest ability to recycle capital expenditures - do not loose much!
In summary, software companies are at the beginning very uncertain and later due to winner-takes-it-all highly predictable (if the management is cautious about new technology - be aware of the institutional bias!). Do not listen to the hero stories, but take a close look to the processes applied in the successful company. Be very flexible early. Follow a decentralized approach in jumping new markets.
Last but not least. In real estate its "Location, location, location". In Start-ups its "Execution, execution, execution".
This blogs represents my personal opinion.
(Mohnish Pabrai is one of my favourite investors and some kind of personal role model. One thing I share with him is a concentrated portfolio approach. I think it lies in a personal trait to have some kind of entrepreneurial spirit - being able to withstand the stress your company causes to your mind when shit hits the fan. I was an entrepreneur - before selling the startup to an investor I now work for. Mr. Pabrai was running his company before starting his investment career.)
One thing I tried to figure out since 2010 and a major cause to start this blog was to figuring out what causes a few software companies to get big and many to fail? I knew, if I could figure out some few right ideas, I could extract huge returns as an entrepreneur and investor.
Mr. Pabrai describes some really important ideas related to software startups and investing.
Occasionally, you’ll see a
company like Stewart [a deadcare company] which shows three interesting characteristics —
Low Risk, High Uncertainty and High Return Possibilities. The
combination of these three attributes at the same time in the same
company makes for some very satisfying investing returns. Take
advantage of Wall Street’s handicap!
If I screen companies, I look for these three qualities. Some further thoughts on this inverse Bermuda Triangle is written by R. Zeckenhauser ("Investing in Unknown and Unknowable") - a must read!
We know that Lower Risk is caused by a Lower Price. A low price can be determined by historically, balance sheet based or free cash flow based.
High uncertainty is not a risk because its impact can be positive and/or negative. Humans are pretty bad in calculating probabilities of outcomes. Even apes are better in some cases. Mr. Munger recommends to add basic statistics (e.g. the works of Pascal) to the lattice of mental models. I like to use some distribution function based valuation. First I kill the company and than add back earnings of business units or values of assets in the order of confidence I have in the number. The confidence of numbers is a distribution function represented by a subjective calculation. The calculation is based on historical data + some subjective quantified numbers of positive and negative outcomes. Being conservative the negative outcomes are overweight - the number gets smaller. I rebuild the company till I have the market cap. I permute the rebuilding order and conservative assumptions. With a few permutation I can guess what uncertainties are priced in and which uncertainties are important.
High Return Possibilities are the other side of the conservative assumption. They represent positive impacts but should not represented in the intrinsic value. Its too easy to wind up yourself.
Sometimes the price seems to be a little expensive. In my experience this happens if a company has very predictable cash flows but high return possibilities. I buy them, because I wont loose much. I think of Microsoft a few month back or Corum Group http://www.corumgroup.com.au/. (At the moment I think I could build an automatic screener for the three.)
Lets see what Mr. Pabrai further has to say. (See also Amar Bhide: How Entrepreneurs Craft Strategies That Work)
Bhide’s research showed
that virtually all startups fall into two categories: marginal
startups (e.g., hair salons, lawn care, etc.) and promising
startups (e.g., Microsoft, HP, etc.). Marginal startups have low
uncertainty, low investment requirements and low likely profit.
Promising startups have high uncertainty, low investment
requirements and low likely profit. However, both types have two
things in common: they are low risk and arbitrage oriented. (...)
Arbitrage by definition is
low risk and typically a return slightly higher than the risk. (...)
Our hair salon has a
wonderful competitive advantage when it starts [ - no other salon in
24 miles], but that advantage is not durable [ - other salons
attracted by salons arbitrage]. Hence it would make a poor investment.
Similarly, the barriers to entry for others to create [a alternative
for Microsofts] BASIC compiler for the Altair were essentially
nonexistent. (...) The good news for entrepreneurs and investors is
that the arbitrage spread can occasionally last for years.
Given enough time, some durable barriers to entry may be created
such as brand, scale or a loyal customer base. Any sustainable
competitive advantage is nonexistent at the time of startup.
I thought about every part (reinvestments, uncertainty, profit, emergence of competitive advantages) for nearly 4 years and I could not see the big picture. After reading Mosaic snippset, I see some stable relations between them: No start-up has a moat. A moat might occur over time. Its better to have a Start-up in an highly uncertain place than one in a certain. Jump to the next niche, when no moat occurs. Every jump might kill a start-up. Every moat and jump behaves probabilistic. In the software industry a moat causes in wide areas a winner-takes-it-all market.
Is a successful software company a result of skill or luck? The answer seems to be more luck than skill. Can I exploit the luck? Applying the weak law of large numbers says to jump a lot, to maximize the chance to get in a winner-takes-it-all market. To jump a lot means not burning/loosing cash for each jump. The low cost jumping is the creative part. In the IT area some on surface technically unrelated topics are closely related dune to same mathematical groundings and vice versa. This can lead to low jump costs.
One strategy to get some insights in the cost of each jump, is to apply the Kelly criterion. It says how much someone should bet for each jump to maximize the long term profit. The problem is that the bets outcome is by and large a long tail distribution (winner-takes-it-all market). Long tails having no average, therefore the bets outcome is not a tangible number. Kelly is not applicable.
In the book "Antifragile" by Nassim Taleb, he says that most long tail distributions in reality are asymmetric. The downside is much larger than the upside or vice versa. As the costs of a jump is limited and the outcome is a long tail distribution, we exploit the asymmetric behaviour. Under such a distribution Mr. Taleb suggests its better to follow processes than goals. This is because long tails outcomes cannot be expressed as numbers, therefore a goal lets say 10 Mio. revenue is totally unrelated to the real market revenue achievable therefore more a distraction than an information. A process is a step by step manual which can better adapt to the real market.
Software product management is another important part of delivering the right product to the customer. It says that a product sold, is more than the technical software. Its a brand, user experience, service, etc. which is sold too. Each jump to the next market should cause as low as capital expenditures as possible through recycling as much resources as possible - the technical software, customers, services, brands etc.
As a side note: Rocket Internet is a venture fund / enterprise which operates in the B2C market. It decentralized the idea generation and technical execution through idea acquisition and in sourcing in their own business units and centralized the administration, marketing- and brand-management in the mother enterprise. The BU profits from the experience in the consumer market of the mother and its cost structure.
In valuation of a software company, I want predictability in earnings before I will pay a fair or slightly overpriced company. This is not possible for start-ups.
Can an entrepreneur use a distribution based valuation to detect promising markets? Not really, because there is mostly no market value beforehand.
Which market should the start up penetrate? The one which has the highest ability to recycle capital expenditures - do not loose much!
In summary, software companies are at the beginning very uncertain and later due to winner-takes-it-all highly predictable (if the management is cautious about new technology - be aware of the institutional bias!). Do not listen to the hero stories, but take a close look to the processes applied in the successful company. Be very flexible early. Follow a decentralized approach in jumping new markets.
Last but not least. In real estate its "Location, location, location". In Start-ups its "Execution, execution, execution".
This blogs represents my personal opinion.
Kommentare
Kommentar veröffentlichen