How we make decisions

/How we make decisions

By Wilson Kageni – Founder & CEO


Data > Instinct > Chance


Of the four crucial determinants I listed in How we work, Decision-making comes first because everything that is done in the company is determined by a decision made by someone, somewhere, at some point in time, so if your framework for making decisions is wrong then it leads to you making the wrong decisions, which leads to everything else falling apart as result.

Of course, there are many different ways we could categorize the various types of decisions one has to make in life; from the trivial (e.g. which coffee mug to use today) to the mundane yet arguably important (e.g. whether or not to wash your hands when leaving the bathroom) to the crucial (e.g. whether to quit your job and start a business, whether or not to get married, whom to marry, whether or not to have a child, when to do so, et cetera). In the context of a business, the overwhelming majority of decisions are still made by individuals but often times the impact of these decisions is much larger than that of personal decisions: in the case of large companies because they affect the lives of so many people and in the case of small companies because similar to a graph line starting at zero while moving outwards, even a tiny deviation that early in the timeline can produce a massive shift in the company’s eventual performance / direction / results later on. For the purposes of this essay, we’ll adopt a simpler classification popularized by Jeff Bezos in his 2015 letter to Amazon shareholders in which he suggests that there are 2 types of business decisions. Type-1 decisions are irreversible (meaning the opportunity cost of getting them wrong is high, thus making them high risk) and Type-2 decisions are easily reversible (meaning the opportunity cost of getting them wrong is low, thus making them low risk).

It makes little sense to obsess over whether or not you are getting a Type-2 decision right because you will find out soon enough and can easily fix any damage by changing direction. A Type-1 decision on the other hand warrants deeper investigation of the assumptions behind your choice and whether or not they hold up when assessed objectively by others who do not possess the same perspectives or biases as you.

The rest of this essay is specifically about Type-1 decisions.

While business lore is filled with legendary moments where a ‘gut-decision’ saved the day just in the nick of time, I’ll posit that it is also full of many more moments when gut decisions in fact led to ruin that was otherwise avoidable. We just typically do not hear as much about the latter because there is nothing remarkable about blindly trying something only to fail and the former make for much better stories. Having this in mind, it’s good practice to be able to explain the reasoning behind each of your key decisions (beyond some natural instinct) and even better practice when that reasoning is backed by relevant data of some sort indicating that in other similar situations, a positive outcome typically follows decisions similar to yours. This is a data-driven framework of decision-making.

Choosing what to do is easy when there is precedent because it provides at least one data source which gives you an inkling of what could happen in a given situation. Multiple precedents are even better because they give you a much clearer trend of what happens in that situation under different circumstances. But what should you do when you can’t find any precedents to guide you?

Look elsewhere (outside your industry, your society, your species, etc.) and find a viable representative or proxy for which comparable data is available.


On choosing your proxies wisely

Picking a proxy is easy but picking a viable one may not always be and picking the best proxy out of a pool of potentially viable ones may be even harder. This matters because picking the wrong proxy provides the wrong reference data, making the entire exercise pointless. What you’re looking for in a viable proxy is a dynamic between components or variables that is as similar as possible to that in your dilemma, but for which the data points you need are available.

For instance, if you are trying to figure out whether mining asteroids for precious metals to sell back on earth is a viable business idea but can’t find any direct precedents since no one has attempted it yet, one would imagine that a viable proxy might be to study the beginning of any other type of mining in the history of the world (iron, gold, silver, fossil fuels, etc.) and try to understand how various factors affected which ventures succeeded or which failed and why. This sounds like it might make a good proxy because in both cases you’re assessing factors affecting outcomes in the same industry (mining) at similar points in the timeline of their existence (i.e. the very beginning) and the data for your proxy is easily accessible.

However, a deeper assessment reveals that while this proxy does indeed address the dynamics and economics associated with prospecting or mining in new territories, it does nothing to address the most unique and therefore riskiest component of your plan: Doing all this in space. The unique operating environment of outer space is actually the biggest unknown in your plan and thus the most likely potential cause of failure, meaning that for your proxy to be useful it must provide some insight into your chances of success in such a unique operating environment. In this regard, you’re actually better off studying deep-sea diving expeditions by sunken treasure hunters or ocean explorers because the data on this will at least be addressing a similar core problem to the one you would face in realizing this venture.

An even better proxy however, might be to study past ventures in the exact operating environment of space and preferably ones that have dealt with similar variables as you will be facing (terrestrial launch, long-distance multiyear travel, pre-calculated rendezvous with a relatively small & fast-moving target, synchronized landing, site mapping & navigation, terrain maneuvering, excavation, mineral analysis, data transmission, et cetera). Looking in this direction would actually yield an abundance of data on the rover missions sent by various space agencies to land on & explore the moon, Mars and recently even a comet. Of these, the data from the latter (the Rosetta mission) would be particularly relevant because successfully landing on a comet & exploring the surface is the most similar actual instance of a mission most closely resembling the first part of your plan and for which real data is available. Because the dynamic between components in the latter proxy more closely resembles that in your dilemma the data points are more likely to correlate than those in the former proxies.

Often times what we want, feel or like isn’t necessarily what’s best and when in doubt you should always believe history (provided you can be reasonably sure it is an accurate history). While in some situations it may seem that there is no ‘historical’ data available to allow us perform this analysis, some kind of data is almost always available. The world around us is filled with data points. Practice finding the right proxy and you’ll likely always be able to find the right (i.e. most useful) data.


That being said, it’s worth noting that:

i.)   Proxy data can be available but ultimately useless. Since a proxy is an attempt at estimation, it’s important to maintain a healthy level of skepticism when it comes to applying the same principles to your context. For instance, you could consider the proxy data, but assign it significantly less weight in your deliberations when compared with first hand data on other factors influencing your decision.

ii.)  Having no historical data backing your decision is fine, but only if there really is none and no appropriate proxies exist (which while rare is entirely possible). Don’t paralyze your decisions because you can’t find any applicable data. Crucially, if you can’t find the right data, don’t fall back on the wrong data because more often than not, bad data is worse than no data. Having no data at all forces you to be cautious and observant in your new direction (all good things when you are building a new data set in a pioneering field), while relying on bad data is likely to make you more confident in your choice to speed down the wrong path.

iii.) It’s okay to disagree with the historical data. The point here isn’t that you should always do what the data suggests you should do. The point is that you should always find relevant data, understand the story it tells, then make up your own mind from an informed perspective. If you have a rational justification for your position that explains why your view point may still be right although this is not reflected in the data you found, then great. Go for it.