Technology startups — A view from Game theory perspective
by Anand Jeyahar
Warning: Hand-wavy, claims ahead, perhaps this post is not very much better than Entreporn, but at least i haven’t come across this angle of approach(think position) before.
1. Zero-sum or Non-zero sum Game
2. Rewards-risk profile
3. cooperative vs non-cooperative
4. Symmetric vs Asymmetric
5. Simultaneous vs Sequential
6. Complete vs incomplete information
6a.Perfect vs Imperfect information
7. discrete vs continuous
8. Stochastic outcomes
1.Zero-sum or Non-zero sum games:
The most analogous concept(to sum) that springs to mind is market size/potential.
Now, this is not determinable in some cases like a totally new product(rescuetime/toggl/etc.). There are other cases where we know, the market is going to be a finite sum, with growth over time. Rarely is it the case of finite limit static over time.In most cases, it is a question of how much of market share at what point in time. But as far as the startups are considered, it is either a new product, with expected market growth in the future or it’s a disruption in an existing business/market(i.e: to wean/corner some part of the existing market by additional specialization eg: online erp services like erpnext.com)
2. Reward-risk profile:
In this case the conventional cliché is that higher the risk, higher the reward. Ofcourse that is a oversimplified,misquoted,abused cliché because it doesn’t mention time dynamics. There are a couple of things involved in this. How much risk over how much time? And what is the expected reward for the company given all the risks we have taken. During most decision-making, the common fallacy is to forget the total expected reward vs total risk company takes. That is the first mistake that causes a lot of people/startup teams to take on more risk than they can manage.
3. cooperative vs non-cooperative:
At first look, it looks like tech startup is a competitive, non-cooperative space. That may be true for the first few years, but as you grow you might need to co-operate to scale/grow bigger.(Think facebook and zynga).Some markets are inherently favouring non-cooperative(like in cases of zero-sum game) strategies. On the other hand, there are some cases where cooperation makes more sense(especially in disruptive markets, where there is an established David vs Goliath situations).In those cases you can piggy-back on other startups specializations and focus on your USP.But of course, how long does it take for different startups to realize their situation and/or USP enough to be secure about co-operation, is a tough question. Besides, in business there’s always a level of suspicion and lack of trust more than in other human social endeavours.
4.Symmetric vs asymmetric
My first reaction on this was network effects, but it can be a lot of others too. If you are building a product with network effects(say social network/content ad delivery) you are clearly in an Asymmetric market. i.e a player with huge network can offer your functionality at a lot cheaper price. (think how google server farms enable them to easily scale their apps).From the game theory stand point, it(Symmetric) means the game rewards are player agnostic. i.e the same strategy played by 2 different players will have same results. Again reality lies usually between both these extreme ends.
5. Simultaneous vs Sequential:
Now, this is perhaps the most simplest case when it comes to startup environment. Mind it simplest doesn’t mean easy/it will stay simple. It really comes down to your plans for new feature additions or release schedules. Basically an idea of who releases the next cool feature first and what is really the next coolest feature(this kind of overlaps into the next point). So i think it’s fairly safe to assume in the startup products market this leans mostly towards Sequential.
6.Complete vs Incomplete information:
This is another of those huge, tough problems that should rank alongside of P=NP problems. Anyway, in the context of startups and products, i would say complete information doesn’t exist. Everybody has some modified form of incomplete information(about market needs) and are trying to improve it continuously. This is actually why Lean startup concept by Eric Ries works. Essentially, he’s talking about find the metrics and process that let you get your information closer to completion.
6a.Perfect vs Imperfect information:
Sometimes mistaken for Complete information, there is a rather subtle difference. Perfect information is where you have all the moves made by the opposition.(Ex: chess,checkers) OTOH, in games like AOE,Rise of Nations, etc. you typically have the state of the game after the opposition’s moves rather than being able to see his moves. Now transferring it to startup context, we don’t usually have perfect information and it is most important in turn-based/sequential games. And if you are developing a new product with a Unique value proposition, then it may not matter anyway.
7.Discrete vs Continuous:
Here’s one concept that’s generic/vague enough to be applied to any problem/field. And the answer not very surprisingly, is it depends on the Sampling frequency. Ok, with that out of the way, let me get to how i think it applies to the startup industry.
This is a concept that makes more sense from the view point of a VC/investor. The simplest form of the idea is if the games’ outcome is expected to be stochastic(random/uncomputable probabilities), it is a bad investment. In theory, he can use some models(assumptions) to generate/predict a probability distribution. Of course, in practice having a heuristic of avoiding such games is less time-consuming and simpler for the investor. Now as a startup founder, this has relevance when you are pitching for investment. To begin with, if you too think the outcome is truly random and can’t be modeled you shouldn’t be doing it(unless of course you like gambling and in that case your best bet is to find investors who like the same). Otherwise, your job/pitch becomes that of conveying/educating the investor as to why/how you think your model is sound and it’s predictions reliable.(Remember reality distortion field?? this is why it became famous.) The usual tactic is to hide the model’s weak assumptions and tout the predictive abilities. But the problem with that is the investor might end up with expectations beyond the model’s capabilities and you’ll be left to keep that illusion up. Ultimately it comes to individual choice as to which level of intermediate balance to choose, but make sure you’re aware of the choice you’re making.
Relating this in startup context seems to stretch my imagination , but still possible. The classification of combinatorial, depends on the multiple possibilities and the tendency for the possibilities to with every agent in the game(Ah, that explains, why i used to take so long to make my move in chess games, too many agencies). Now there is no dearth of agencies in the startup space. Nor is there a dearth of possibilities(as a consequence of no. of agents). But the problem starts with simultaneity. In the real world there is a lot more simultaneity that sometimes thinking of possibilities and computing the combinatorics is hard. Ofcourse one famous example is the famous computer simulation of battle(with complete information) playing against real-time generals making decisions.
I think it might be a good idea for anyone, who wants to start a startup to think in these terms. Not necessarily to get a set idea of what they are going to do(some these are not mutually exclusive/orthonormal) but to have some idea of where they fit on the options.
So here’s an informal summary(with lots of help from wikipedia).
This post was originally inspired by this: HN(Hacker News) post of Entreporn:.
P.S: I must mention, i am a fan of idea. And think it’s the approach that resonates with my ideals.
P.P.S: I am new to writing these kind of posts. Neither am I an expert on Game theory or Startups, but i spent a good couple of hours, writing this post, so please do give me specific feedback/comments.