Risk v/s outcomes

Controlling risk for startups – Part 1

In the last article, I discussed the dynamics of skill v/s luck. I defined outcome as a combination of skill and luck – we said that for the same level of outcome, low levels of skill need a high degree of good luck and vice versa. As skills increase, we become more resilient to bad luck. A high degree of skill means that a freakishly bad event has to happen to lead to absolute failure. So skill and luck are kind of inversely related.


We also discussed the various biases that happen due to a poor understanding of where we are on our skill/luck curve and where others are on their skill luck curve. Mispricing our own skill & luck levels leads to self-serving bias and mispricing others skill/luck levels leads to envy or authority bias.  Self-serving bias is Overpricing skill in good times and under-pricing skill in bad times. Envy bias is overpricing luck and under-pricing skill in others and authority bias is overpricing skill and under-pricing luck in others

Common sources of Risk

Risk originates primarily when we either are at the low end of skill curve or when we mis-price skill/luck. This doesn’t mean that risk does not exist for entities with a high degree of skill or entities that rightly price their skill/luck levels. I restrict my discussion to these 2 sources of risk because they are very commonly found in startups. Risk profile for mid to large sized companies that have been in existence for years is a completely different discussion that is honestly beyond my scope.

Risk due to low skill level

We already discussed last time that when I am at the ‘low skill’ level, to generate the same level of outcome, I need a big tailwind of good luck. Whenever luck has a disproportionate say in outcomes, risk raises its head big-time.

This chart above is misleading because we are not in the world of ‘certainty’ but in the world of probability. Nothing in the world of ‘probability’ is a ‘fixed’ point – it is always a probability distribution. For those of you who aren’t familiar with probability distribution, please read my article explaining the concept of probability distribution in very simple terms.

Luck here means the thickness of the tail of distribution. Higher degree of luck means a greater thickness of tails, lower degree of luck means a smaller tail thickness (thinner tails). Thinner tails mean that there is less chance of outcomes moving very far away from the mean value – extreme good luck or bad luck are very rare in thin tailed probability distribution.

So a Covid19 event is an extreme event on the bottom side of the curve – for industries such as travel, airlines, hospitality, Covid19 has caused devastating losses to their business.

Risk due to low skill level

The second source of risk comes from a perception v/s reality mismatch of our position on the skill-luck curve . While I introduced biases in last article, I did not explain how biases created risks. They do so by impede our learning curve. Biases don’t let us learn the right lessons from both success and failure. Let’s see how..

Biases impede our learning curve. They don’t let us learn lessons from success & failure.

A success or failure, when looked at with an open-mind and analyzed fairly, gives us access to what I call ‘learning loop’. A success leads to a small ‘learning loop’ because mostly lessons implemented to achieve that success are the ones we learn in the previous cycles of failure. Failure leads to a relatively large learning loop – so a series of failures and successes combine to create an upward sloping learning curve that comprises of several learning loops.

When we are victims of something like a ‘self serving’ bias – this loop breaks. Since we attribute a failure to bad luck and success to our skills, we have little incentive of identifying where we fell short on skills & what we could have done differently to reduce our exposure to bad luck. Depriving ourselves of this analysis leads to an incomplete or stunted loop – and since one loop doesn’t feed into another, we lose out on the ‘compounding effect’ of lessons from both success and failures. The upward sloping learning curve ends up as a relatively flat learning curve.

Controlling & Managing Risks

Now let’s come to the key aspect of this episode – How to control and manage risk during the periods of low skill. The answer lies in all the previous topics we spoke about. Here is my prescription for reducing risks in decision making for startups in the early years.

  1. Increase skills quickly over time
  2. When skills are low, keep your exposure to negative luck low
  3. Constantly review if you are victim to some form of a bias & correct yourself promptly

I will discuss the first one in this article and leave the other two for next week.

Increase skills over time


If we plot skill v/s time… let’s assume skill increases over time.. we are picking up more know-how and specific knowledge with time that is increasing our skills consistently. Let’s say we draw 2 lines – success line (green line), a threshold line above which we can classify that our startup is a success. A failure line (red line), a threshold where we end up losing everything and incur a permanent loss.

As discussed previously, I will represent luck, a random event, by drawing a normal distribution of outcomes.  Since luck cuts both ways. I will represent luck with a bell shaped probability curve that extends both sides – the top side represents all the worlds with a positive outcome and bottom side represents all the worlds with a negative outcome. 

Lets take two points – one at an early date when skill levels are low, lets call is A (shown above). Other at a a later date, when skill levels are high, lets call it B. I draw a fat tailed curve on both sides of the skill curve at point A. This represents the occurrence of a random event that brings luck, either positive or negative . At point B, I assume, for simplicity, the same event as point A and hence assign a same distribution. 


Now at a low skill level A, all the outcomes greater than Fa lead to failure, Fa is a point where an event creates an outcome that intersects failure line. Notice that thicker the tails, greater is the chance of failure – area of the curve below point Fa represents all the worlds where our startup fails. Now come to point B, to hit the failure curve, the same curve needs a much bigger extreme event. Point of intersection is Fb – notice how the area under the curve, which is a proxy for the probability of that event, is very small – this increase in odds is a huge huge win for every entrepreneur. Even a 5% increase in odds can lead to a huge positive outcome in future.
Key observation is, so long as skills improve rapidly with time, greater the age of a startup, greater is the likelihood of success.  Most stratups improve their odds of success dramatically once they go beyond 3 year point.


Skill here means improving execution strategy for sales/marketing/product development, customer support; understanding what your target customer needs & and the gaps that exist in the market. Skill also includes vision of what you want to be, bringing people together who share that vision & constant planning of the future.

3 things that rapid skills accumulation 

1. Take big decision quickly but change that decision slowly

Major decisions on what to do, who to work with, which industry to target should happen quickly. I’ve found that the cost of waiting to make a big decision is usually much higher than the cost of making a wrong decision. Once you lock in major decisions, you define your boundary conditions within which you will build your skills. Once these decisions are made, changing your vision, industry or teams quickly can lead to stunted growth in skills.  Once a big decision is made, cost of switching the decision quickly is much higher than cost of persisting with a wrong decision.

Cost of waiting on a big decision is usually higher than cost of wrong decision

Focus on actionable metrics

Our will power is limited, don’t spread it across multiple things – focus on 2-3 metrics and relentlessly pursue improvement of those metrics. Channelling energies into improving a small number of metrics can automatically improve skills. A metric for a startup could be anything from, number of people on website, to number of client demos, to number of sales calls per month. to server response time etc. 

Decrease the number of moving parts in the business

Keep your core value proposition as simple as possible. A complicated model increase the number of skills to master. More skills to master, slower will be the growth of skills & greater will be the role of luck, specially in the initial years

To give an example, let compare 2 businesses A and B: Business A: Sells used cars online without keeping inventory – Business A is a market maker connecting buyers and sellers online.

Key challenges for business A are 

  1. Where to find customers online
  2. What offers to make at which time
  3. How to bring back customers who dropped out of website
  4. How to maintain relationship with existing customers

Business B: Tries to sell used cars online by holding inventory. Business B has higher margins and a much higher potential than business A. But Business B is way more complex than business A Business B will have to do all of what Business A does and also

  1. Find sellers online and offline
  2. Do a accurate appraisal of cars
  3. Logistics of holding and maintaining cars
  4. Arrange capital to make payments
  5. Hire people who can manage operations

Business B has a much steeper learning curve than A. And skill growth will be slower and more patchy than B. There will be days when something collapses and needs a lot of attention – these disruptions can slow down growth of skills and make business vulnerable to failure.

In the next article, I discuss how to reduce exposure to a negative luck event. While we cannot stop an extreme bad luck event from happening, we can surely cut down our downside exposure to such an event. Stay connected and stay safe.