Comment by K0balt
3 days ago
There is a fallacy in over application of the survivorship bias concept.
Literal survival for early cultures was often a matter of luck. Agriculture was an innovation that improved the odds. Early cultures that practiced agriculture outperformed those that did not, and were more likely to survive black swan events. All major cultures in existence are now based in agriculture.
Should we assume then that because we only see agrarian cultures that that is not useful information, because of survivorship bias in the resulting sample?
On the contrary, survival itself is the signal that is useful… it’s really a matter of what behavior the signal can be attributed to- was it the agriculture, or was it the human sacrifices? Was it the red ochre face paint? The storing of grain in pots instead of skins?
Failure bias is just as large of a red herring. It’s easy to imagine that it retrospect, we understand why failures happen, and sometimes the reasons are very clear. That’s why there is often more to be learned from failures than from successes. But still, it’s easy to look at the things they did right that successful example B also did, and then conclude those things weren’t critical to success because they sometimes end in failure.
The point is that we shouldn’t judge the value of information based on ideas like “survivor bias” but instead look for more methodical and logical connections between causes and outcomes, and not fall victim to cargo-culting nor casual, hand wavey dismissal of potential lessons.
Survivorship bias mitigation is a matter of determining which survivor signals are instrumental , and those which are coincidental.
Many things are fraught with risk and low probabilities of success. That does not make them primarily a matter of luck.
Aviation is a great example of an environment that is nearly 100 percent risk, where without knowledge and the correct tools the very small chance of not dying would be purely a matter of luck.
Even carefully thought out comments like the above are only hints and ideas relating to making sense of the world. They don’t talk about building quantitative predictive or causal models to disentangle the many factors driving success, failure, and everything in between.
I recommend The Book of Why by Judea Pearl as a starting point for digging into the lesser-known techniques of assessing causality. The causality work over the last couple of decades is still under-appreciated and not used often enough.
One is unlikely to find anything close to rigor when it comes to business or entrepreneurial books. They can be a starting point for analysis, but their bias to tell an interesting story and sell copies often work at odds with truth seeking.
At the risk of oversimplification, one decent model for acting comes from decision theory. (1) Look at the data probabilistically and act accordingly. (2) If you don’t have enough data, assess the cost/benefit of acquiring more.
But we don’t have the time or the discipline to make all important decisions this way, do we? Probably not. So, (3): if you act on intuition, be honest with yourself about that. Be curious about yourself and your decisions and how you can do better. Focus on areas where improvement is likely to make an outsized impact. (This leads back to #2, except it is about seeking better knowledge and self-awareness instead of just data.)
I try not to exaggerate, but these three principles might be sufficient to subsume all other business advice.
Embrace the uncertainty and move forward anyway.
Thanks for the book recommendation! I’ll look at that.
>> Embrace the uncertainty and move forward anyway.
This. This is the key factor that prevents attempts at success, or leads to failure by a thousand cuts.
There is no gain without risk. Significant gain usually comes through significant risk. Position yourself in life to be sufficiently resilient to take 10:1 bets with 100:1 odds until you can weather the 100:1 bets with 100000:1 odds. Avoid the 1:50 bets that pay at 1:50 odds like the plague… they are a comfortable quagmire of rotting aspirations.
We are not in disagreement here. Of course it makes sense to study successes and figure out differences in patterns of behavior making success more likely. Causal or not, that is another matter.
However, it's not black-or-white. Once an aspect that has likely contributed to success is identified, it's often framed as if focusing on that aspect guarantees success. That's an easy mistake to make if all observed cases of success exhibit it. But it's far from a guarantee, other aspects my play a role too. Like luck, for example. That's what I meant by mentioning survivorship bias.
What's missing from the application of the survivorship bias concept is that it really only applies if failure means death or a total wipeout. Pieter Levels for instance, a known success in the indie hacking scene, has created something like 70 projects and only a handful of them made any money. Presumably he also learned something from each project making the success of the next more likely.