Every executive I work with eventually brings me the same problem in different clothes. The board wants a decision. The data does not support one. Waiting has a cost. Being wrong has a cost. The question they ask out loud is "what would you do?" The question underneath is "how do I decide when I cannot know?"
This essay is my answer. It is the method I used in three decisions I still think about, and the one I use now as a practical, candid thought partner to CEOs and product leaders on strategic product and business challenges.
The people who wrote the field, and what the seat adds
Four people shaped how I think about this, and I recommend all of them. Jeff Bezos gave operators the plainest version of the test in Amazon's 2015 shareholder letter: some decisions are one-way doors you cannot walk back through, most are two-way doors, and the two deserve different processes. Annie Duke taught executives to separate the quality of a decision from the quality of its outcome, and to treat every choice as a bet with odds rather than a verdict. Philip Tetlock measured forecasting, showed how often expert prediction fails, and in Superforecasting described the habits of the small group who beat it. Howard Marks writes about cycles, risk and second-level thinking from inside a firm he co-founded and still co-chairs, so his memos carry the weight of capital at stake.
What I add is narrower. Bezos's test sorts decisions into two doors and stops there. The operating seat forces two more questions once you know you are standing at a one-way door: how much of this can I still take back, and by what date will I know whether I was right? Most of the strategic decisions on a CEO's desk are not cleanly one door or the other. They are mixed, and the mix is the decision. A CEO is not asked to forecast the market. She is asked to commit people and money before the forecast resolves, and to answer for it either way. The method below is how I do that.
Three decisions I still think about
Interland. We had grown from nothing to $200M in revenue and 650-plus people, and taken the company public on Nasdaq. Then the dot-com market collapsed. Every forecast we had was built on a world that no longer existed. The decision was whether to stay independent and wait for the market to come back, carrying the cash risk that came with waiting, or merge with MicronPC and survive as something different. No spreadsheet could answer that. We merged. The company survived.
Blameless, as COO. A turnaround where the honest read was that burn would kill the company before the product could prove itself. The decision was how much to cut without cutting the thing customers were paying for. We took burn down by half while revenue grew forty percent, and the company merged with FireHydrant from a position of strength.
Cox Automotive. I ran a cloud and DevOps transformation with a $250M budget: more than ten thousand servers moved to the cloud at 99.9 percent uptime, with a thousand-plus engineers changing how they worked. The uncertainty was not whether the technology worked. It was whether an organization of that size would move with it. Every phase was a bet on people, placed before the evidence was in.
Across those, and across the thirty-five M&A transactions and eighteen exits I have been part of as an operator and an investor, the same habit kept showing up. I did not have a name for it then. The name below is mine. The idea underneath it, sorting what you cannot undo from what you can, belongs to the people above, and to every operator who has learned it the expensive way.
First, is it a must
Before any ledger, I ask three questions of the decision itself, and I ask them of every leadership team I work with. Where are we going? Why? And is the thing in front of us a must or a should? If it's a should, one of two things is true: the where or the why is wrong, or we are not committed. Either way, the how will stay foggy no matter how much analysis we pile on it. If all three line up, the how gets clear and focused, and that is the moment the decision deserves the work below. Tony Robbins made "turn your shoulds into musts" a well-known line about willpower. This is not that. Here the where and the why do the deciding, not the will; a must you cannot explain is a should wearing a costume.
The Reversibility Ledger
The Reversibility Ledger is a two-column method for a one-way-door decision when the data has run out: what I cannot take back, against what I will know and when. The rule that joins the two columns is the whole method: prefer the option whose irreversible commitments are smallest relative to what you will learn by the date you named. The decision that matters is not "which option is right." It is "which option can I undo, at what cost, and how fast will I know."
Write two columns.
Column one: what this decision commits me to that I cannot take back. People I will lose. Customers I will disappoint. Money that is spent. A market position I give up. Be specific and be honest. Executives tend to underestimate this column because they picture the version of the decision that goes well.
Column two: what I will know, and when, that I do not know now. Every uncertain decision has a date on which some of the uncertainty resolves. A customer renews or does not. A quarter closes. A hire works out or does not. Name the date and the signal.
Then apply the rule. Not the option that wins on a model you do not trust. The option that keeps you in the game long enough to learn.
At Interland the merger felt like the larger commitment. On the ledger it was the smaller one. Independence in a collapsing market committed us to a cash position we could not reverse. At Blameless, cutting deeper than felt comfortable was reversible. Hiring back is possible. Running out of cash is not.
Three habits that make the ledger work
Decide the decision date first. Uncertainty does not resolve itself. Set the date on which you will decide with whatever you know then, and tell the board that date. It stops the meeting from becoming a weekly rerun.
Separate the quality of the decision from the quality of the outcome. This is Duke's rule, and it is the one operators struggle with. A good decision made on the ledger can still go badly. Review the ledger, not the result, or your team learns to hide its reasoning.
Say what would change your mind, in writing, before you decide. Tetlock's best forecasters wrote down in advance what would move them. If nothing would change your mind, you are not deciding under uncertainty. You have already decided and you are looking for cover.
When to use it, and when to bring someone in
Use the Reversibility Ledger when part of the decision cannot be undone, the data will not settle it, and waiting has a price. It is built for CEOs, founders and product leaders who have to commit before they can be sure. It is not for decisions a spreadsheet can settle. Those you should just make.
The ledger is simple. Filling it in honestly is not, because the person filling it in is the one who has to live with column one. That is where an outside operator earns the fee: not to make the decision, but to make column one complete and column two concrete when the executive is too close to it. It is much of what I do in advisory work. This is the conversation I have as an executive advisor on a first call, and it is the one I had with myself at Interland. The shapes that work takes, from standing counsel to running the change yourself, are laid out on the strategy consulting page. If there is a decision on your desk now, bring it to a discovery call.