Overcome indecision and inaction
Treat life like a multi-armed bandit problem
TLDR;
Intentionally divide your time and resources into exploration and exploitation phases. Keep records to hold yourself accountable.
On to the main event…
I’ve always wanted to start a business. Even a small side hustle would be fun and fulfilling. But ever since I gave up my teenage lawn care business nearly 20 years ago, I’ve been in school or working for “the man”. I’ve thought about, embarrassingly, hundreds of business ideas. I bought domain names, started blogs, built prototype devices, tried to convince friends to join me; but alas, my ideas tend to end the same way — indecision and inaction.
Many who know me well would be surprised to hear that I have a hard time making commitments: I got married at the ripe old age of 22; I remain married, and my wife and I brought four kids into the world; I picked an undergraduate major and never changed it; I picked a graduate school and earned a PhD; I’ve had a great career so far with multiple years at each company I’ve worked for; I’ve been active in the same church my whole life. I’m the picture of stability and commitment.
My difficulty is not with staying committed, but rather, with the lead up to commitments; with the decision to commit. To be fair, I can’t commit to everything. Nobody has that kind of time. The cost of new commitments is substantial and must be weighed in the larger context.
But beyond that, I have a fear of closing doors. I am afraid of missing the “optimal opportunity” because I foolishly or prematurely committed to a suboptimal (or even bad) one. This is a classic case of the “Paradox of Choice”, in which more choices leads to lower overall consumer happiness. I definitely experience analysis paralysis.
But, surprise, surprise; life is finite! Someday, I will run out of years and breaths. I have a finite budget of time to spend while pursuing happiness.
In computer science there is a classic problem known as the “multi-armed bandit”. The scenario goes like this:
You walk into a casino with some cash in your pocket. There is a long row of slot machines, each one with a lever arm. Each machine has a different probability of winning, and a different reward when you hit the jackpot. You don’t know the probability of winning up front, or the return for doing so. Your job is to win as much as you can with the money you have.
The smart aleck may say, “just take the money and leave without playing”, but let’s face it: that’s not an option. You can already see the parallels with starting a side hustle, or any other decision in life. We have finite time or resources, and a bunch of decisions to make. We have to spend the time somehow — even not deciding is a decision.
We need more information about each slot machine to make good decisions, but we also want to put as much money as we can into the best machine. We need more information about possible businesses and side hustles, but we also need to spend time building instead of thinking. What to do?
There are various solutions to this problem (and various versions of the problem). An intuitive and effective solution is the Epsilon-Greedy algorithm, where you divide your resources between “exploration” and “exploitation”. Epsilon is the parameter that defines the fraction of resources spent on exploration.
“Exploration” is the name for the risks you must take to gather information about each slot machine. In the casino, it’s the money you spend learning about each machine. Of course, the more money you put into each machine, the better your information. If each pull on the lever costs $1, and you pull the lever only once, then your estimate of the probability of winning will be very, very crude. Your estimate could only be rounded to 0% or 100%, but nothing in between. Suppose you increase your exploration to 10 pulls on each machine — now you gain a whole order of magnitude in your estimate — your estimate can be rounded to 10%, 20%, … 90% or 100%. But, the estimate will still be unreliable. If you flip a coin 10 times, odds are heads or tails will come up in a very unbalanced pattern (e.g. 7 heads and 3 tails). The empirical probability doesn’t usually approach the true probability (50%) until the number of flips is large (think thousands). So, back to the slot machine, unless your budget is very, very large, you will never have a perfect estimate of the machine’s expected reward.
You get the idea. More resources dedicated to “exploration” will result in better information. But, the tradeoff is that fewer resources are left for “exploitation”.
“Exploitation” is an admittedly unsavory-sounding word used to describe the act of making the best decision given the information you have. If you estimate that slot machine D will give the best return on your money, then during the exploitation phase, you would put all your money in slot machine D.
The grand question becomes how to divide your resources between exploration and exploitation. Unfortunately, there’s no simple answer to this question. In the words of James McCaffrey:
The strength of the epsilon-greedy algorithm is simplicity. The key weakness of the epsilon-greedy algorithm is that it’s hard to guess a good value for epsilon.
At some level, it becomes necessary to accept the cost of exploration, and the inevitable uncertainty that will remain even after investing in exploration.
This realization is mentally freeing.
Back to my struggle with starting a side hustle, I’ve decided that I will be far more formal in my approach. I need to be intentional about exploring, accepting the cost of doing so, and the uncertainty that will remain. I also need to accept that exploration cannot last forever, and jump into exploitation — AKA commitment, decision, action. I need a way to hold myself accountable for both exploration and exploitation phases.
Here’s my approach:
- Break down what I’m looking for, define what success looks like. That will involve some combination of being fun and fulfilling (at least much of the time), some potential for being lucrative, a reasonable startup-cost, and not so time consuming as to harm higher priority commitments (family, faith, etc.). I turn these requirements into a scoring matrix.
- I brainstorm a list of viable side-hustle options (a finite set of “slot machines”).
- Unlike our casino example above, I probably have some information about each option right at the get-go. I fill in the scoring matrix with these “priors” (see my example below). Prioritize the options in the order I will explore them.
- Choose how much time I’m willing to commit to the exploration phase. I write down actual stop-dates. For example, if the average YouTuber has 1000 subscribers and $100/month by they’re 90th video (I made up these numbers) then I commit to making 90 videos. I’m specific about the actions and costs that go into the exploration phase. It may simply be learning about a topic (get actual numbers on YouTube channel startup times, for example), it may mean buying some equipment (camera, mic), or buying some software. You’re the boss, but you probably want minimize your investment, without completely jeopardizing the exploration. Crappy YouTube videos are far less likely to succeed, but you don’t need to spend thousands to make videos good enough. This is a judgement call. When in doubt, start smaller.
- Define, up front, as much as possible, some go/no-go criteria. For example, if after the YouTube trial period I haven’t gathered more than 500 subscribers, or I’m still not enjoying the process of making videos, or it’s too time consuming for my life — then quit.
- Now I act on the plan! I likely cannot try all options at once, so I work on them in priority order. If two or more can be tried simultaneously because they’re small or complimentary, that’s great. But I tend to pile too many things together at once, thus overwhelming myself and making it hard to succeed at any one thing.
- Critically, write down the experience as it happens, gather the data. Recognize that skills and enjoyment grow (editing videos will be harder at first), but enthusiasm waxes and wains. Keep updating the matrix as information is gathered. Maybe I discover that I really enjoy making YouTube videos, that I reach 1000 subscribers within a few months. Or, I discover that I hate everything about it. Or I discover that I like editing videos, but I can’t stand being on my computer after spending my day-job hours staring at a screen. The important thing is to keep notes regularly about the thought process, experience and performance over time. The goal is to avoid a perpetual exploration cycle fueled by selective memory.
- After exploring, on the pre-committed date, I make the decision to exploit, to act, to commit.
Of course, in real life, order matters. Gaining a skill while exploring one opportunity will carry over into the next opportunity and change the odds of success. Life circumstances change (the possibilities change after having children, or after an injury, or large financial setback). Opportunities open up that never existed before (remember the days before YouTube was a thing?). In reality, life is a continuous cycle of exploration and exploitation.
Mental shift
Even if you don’t set up a spreadsheet, I’ve found this mental framework very helpful for overcoming indecision and inaction. I accept exploration as the rent I pay for living, and embrace it as a fun — even joyful — part of life. More importantly, this framework helps me accept that no amount of exploration will provide certainty, or make the future knowable — it helps me accept the risk and opportunity of exploitation before my time budget runs out!
To summarize, put boundaries on your exploration, and commit to action! Attack your decision like an Epsilon-Greedy algorithm. Take time to keep a record of your costs, experiences and rewards throughout the exploration phase. Maybe even make a spreadsheet 🤓
