What separates dependence you chose from dependence you can no longer leave is not how it began

Work you took on by choice has quietly become work you cannot turn down. The clients, the kind of product you handle, the way your hours are allocated — you decided all of it yourself at the start. Yet now, removing any one of them would stop your income entirely.

What produces this state is not carelessness in the choosing. It is the order in which the information arrives: the quantity that would let you judge is settled after the decision, not before it.

When you decide for yourself to rely on something, and the option of not relying on it genuinely remains open, that is dependence you chose. The other state is the one where the option of not relying on it has disappeared. Call that dependence you can no longer leave. From the outside the two look identical. The same counterparty, the same terms, the same work. The only difference is what happens if you walk away.

What matters is that there is no moment of transition between them. No day on which the contract changed, or the other side changed their manner. The terms stay as they were, and only the set of available options quietly shrinks. You find out that it shrank after it has finished shrinking.

The fact that the outward form does not change is a practical nuisance. If something had changed, you could take that moment as a starting point and work backwards to a cause. Because nothing changed, there is no starting point to take. With no starting point, the cause gets assigned to your own character instead.

Most writing on going self-employed explains this transition as a problem on the person’s side. You were not vigilant enough, you neglected business development, you allowed the dependence to form. Each of those is partly right. Yet none of them can name the point at which the transition happened. When you place an unlocatable thing as the cause, the remedy collapses into “be more careful from now on.” The same thing has happened to people who were being careful, so that remedy returns the same result on the next pass.

If you have not gone independent yet, this may read as a discussion of something that has not started. Still, the quantity handled here sits on the side that cannot be rewritten once things are under way. What takes one sitting to write down before you begin stops responding to rewriting afterwards, however many times you try.

What follows first sets out accurately how the pair of dependence and independence is currently discussed, then shows that the line dividing the two states runs through reversibility rather than through quantity. From there it examines, in two separate mechanisms, why reversibility cannot be judged at the moment of choosing, and closes on what can actually be written down at that moment.

Dependence and independence are currently discussed as a question of quantity

Explanations of dependence and independence are now often given in a single shape. Independence is not the absence of dependence; it is having more than one thing to depend on. Placing everything with a single counterparty is what is precarious. Relying on others is not itself the problem. So increase the number of things you rely on.

This account is more accurate than the one it replaced. Before it, independence and dependence sat as simple opposites: an independent person relies on nobody, and a person who relies on others is not independent. That binary does not match what actually happens. Nobody runs a business relying on no one at all, and cases where avoiding reliance is what brings someone to a standstill are also observed. The phrase about increasing what you depend on came into use in order to break that binary.

Part of why the phrase was taken up is that it matched what readers had lived through. Among people who have gone independent, it is not rare to carry the experience of trying to stand without leaning on anyone and being worn down by it. Permission to rely on others is what loosens that particular exhaustion.

The same shape appears on the practical side. Spread your clients. Run more than one income stream. Cap the share of revenue that any single customer can represent. Each is an instruction not to concentrate anywhere, and each builds safety out of number. Increasing the count does work. Having something left when one client stops is better than having nothing left.

There is a situation the phrase does not address: the period when you cannot increase the count. In the first year of working for yourself, having exactly one counterparty is not unusual, and there is nothing to increase. When no instruction covers the period in which increasing is impossible, that period sits outside the instructions altogether. And the fixing tends to begin precisely then.

In addition, this account watches only the quantity of dependence and never the difficulty of undoing it. You can have five clients, and if all five run on the same terms and none of the five can be switched out in the time available, leaving is no easier than it was with one. The count is distributed; the reversibility is not.

There is a reason the counting version spread. Reversibility is hard to measure and count is easy. How many clients you have is something you can total today. How many of them you could actually leave is something you cannot confirm until you try. The measurable quantity becomes the indicator and the unmeasurable one drops out of the indicator — the same substitution shows up here.

And the practical failure usually happens on the second side. Being unable to move despite having spread the work is not a rare state. Counting alone cannot explain it.

The two states look identical from outside

What separates dependence you chose from dependence you can no longer leave is not how it began. It is whether you can leave now.

This distinction is needed because the beginning is useless as a test. Both start with your own decision. You decided to work with this counterparty, to handle this product, to work in this way. There is no memory of being forced. So if you ground the test in how it began, the verdict stays permanently on the side of “I chose this.”

The outward form is no use either. The wording of the contract is the same. The rate, the deadlines, the number of exchanges may all be unchanged from last year. What changed is one thing only: what would happen if you stopped this work from next month. That one thing is written nowhere in the paperwork.

Another proposed test is to watch the other side’s conduct — to gauge the degree of dependence by whether they start making unreasonable requests. That reverses the order. Unreasonable requests start landing after the options have already thinned. Conduct is the result, not the cause.

So replace the test with a single question. What would you have to give up in order to end this dependence right now? If the answer is “three months of income,” it is a dependence whose price you can measure. If the answer is “the business itself,” the options are already gone. If no answer comes, what you have learned is that you have not been measuring.

You can also attach a date to the question. Ask it again in six months, and if the answer has grown heavier, reversibility fell over those six months, and a quantity separate from whether revenue rose has moved.

What makes the question hard to ask is that the answer can read as a verdict on your own choices. If ending it would cost the business itself, the years of judgement that kept the relationship going appear to be on trial. Yet what is on trial is the state, not the judgement. Whether a decision was right and whether options remain now are settled separately.

Falling reversibility is not itself a defect. Deepening a relationship with a particular counterparty builds up something specific to that relationship, and because it cannot be carried elsewhere, it raises what you are worth there. Making non-falling reversibility the goal means never building a deep relationship at all. The problem is not the fall. It is the fall running to completion without ever being observed.

At the moment of deciding, the information the judgement needs is not yet in hand

Being unable to judge reversibility at the moment of choosing is not a shortfall of judgement. The information the judgement requires does not come into existence until after the judgement is made.

When you decide to go independent, the information in your hands runs up to that moment. How much you will earn, what kind of work will come, how you will feel about that work once you have done it for a while — all of it emerges from what happens after you begin. Prior research yields somebody else’s course of events, not your own.

What you can consult is also narrower than it is assumed to be. Other people’s accounts survive only on the results side; which things they knew at which point does not survive at all. The material the judgement needs sits in exactly the part that did not survive.

Further, the decision at that moment is made inside the frame you are standing in. What looks desirable is settled by what happens to be arranged around you, so among people who all hold the same goal, that goal looks self-evident. Something that looks self-evident cannot be compared as one option among others.

After finishing a music college, I spent a stretch of time taking composing work and living on it. From my student days, everyone around me was operating on the assumption of making a living from music, and I held that as the natural goal too. I applied to agency calls and sent tracks to what amounted to contests collecting candidates for a new release, and that continued for some time. As I kept sending, something came into focus: I did not want to write music for people I did not know, and I did not want to earn money by means of music.

I did not hold that sense before starting. There was no way to hold it. When the goal itself looks self-evident, there is no place from which to ask whether the goal matches what you actually want. The question only arose after all that sending. The material the judgement needed arrived after the judgement.

There is a view that a careful starting choice prevents this. Choose your counterparties strictly and you will not be in trouble later. Being careful in selection does work, but what selection can address is the counterparty’s characteristics, not how your own available hours will fill up. What creates the fixing is not the counterparty’s characteristics but the number of repetitions.

An exception can overturn the ordering. If you worked with the same counterparty once before going independent, that course of events is your own and can be used as evidence. Even so, the exception only covers the counterparty’s side; how your hours will fill remains unavailable in advance. Past dealings help you predict the other side, and do not help you predict yourself.

The same ordering applies to going independent. The material for judging whether to work with this counterparty comes out of the months after you start working with them, so by the time you can judge, the thing being judged is already under way.

Every choice after starting is correct at the time it is made

Reversibility falling after you begin is not the result of a wrong choice somewhere. Stack up choices that are correct at each moment and this is what you get.

Right after going independent, whether next month’s revenue materialises is the largest uncertainty you carry. In that state, a situation arrives where you choose between work that pays now for certain and work that may pay off in six months. Choosing the former is rational at that moment. The latter is uncertain and the former is not.

The same situation arrives the following month, and the same reasoning selects the former. Every single judgement can be accounted for, but the choice has an asymmetric effect. Choosing the former deepens the relationship with that counterparty by one step and adds nothing that would carry elsewhere. Choosing the latter does the reverse. Twelve repetitions later, only the dependence side has accumulated.

The asymmetry is strongest early. Early on, uncertainty is highest and certain work looks most valuable, and early choices are the ones that set the allocation for longest afterwards. The period with the least information is the period in which the longest-acting choices get repeated.

This accumulation is hard to see because it does not appear in the monthly figures. Revenue is flat, or rising. Raising something that is moving in a good direction as a problem is difficult, and in most cases it does not get raised. While the indicator moves the right way, the conditions change outside the indicator.

Economics addressed this shape early. Strotz showed that a plan you make now can fail to be the best one by the time you act on it (Strotz, 1955, The Review of Economic Studies, 23(3), 165–180). The self that made the plan and the self that executes it weight the things being compared differently. The passing of time by itself reorders the options.

What that paper deals with is the abstract problem of how an individual’s preferences shift over time. It does not take post-independence work decisions as its subject. What can be read from it goes as far as this direction: a sequence of choices that looks inconsistent in retrospect can come out of judgements that were each consistent at the time.

There is also a reason not to explain this by weak will. Explain it that way and the remedy becomes “choose the long-term side.” But the people who can choose the long-term side are the people with no uncertainty about next month’s revenue. Unless the uncertainty is reduced first, the ranking of the options is settled the same way every month.

Nobody is making a mistake, and still nobody gets out

A state can form in which every party is behaving reasonably and yet, taken as a whole, nobody can get out. It is close to what economics has called the fallacy of composition.

It never appears as a scene in which someone corners someone else. Everyone is making ordinary judgements in their own position. Even so, at the end of those accumulated judgements sits a state that nobody selected.

Look at each person’s conduct individually and all of it can be justified. The composition of justifiable actions produces an unjustifiable result. The two do not contradict each other. The result of the composition cannot be read off the properties of the individual actions.

This view has limits. That something can be explained as composition does not mean individual bad faith is absent. There really are counterparties who impose unfavourable terms. The compositional explanation is not there to make those cases harder to see. Where imposition is real, deal with that first. What is being handled here is the case where there is no imposition and the same result arrives anyway.

The same shape applies to relationships after going independent. The client is only continuing to send work. You are only accepting the work that arrives. Neither side holds any intention of binding the other. Two years on, most of your available hours are filled with that client’s projects and the record you could carry elsewhere has not grown.

Seen from the other side, the situation is the same. Because you are familiar with their process, the effort of explanation drops for them, so they send the next one to you as well. With no reason to try elsewhere, all that remains is the conclusion not to try, and both parties reduce the other’s options without binding them at all.

Look for the person who created this state and you will not find one. Since none can be found, in most cases the assignment lands on yourself: I neglected business development, I was not vigilant. Yet placing the cause on yourself does not amount to an account of the mechanism. Set something that is not an account as the cause, and the next pass returns the same result.

Objecting that this blurs responsibility is fair as far as it goes. If nobody is at fault, then nobody has to fix it. But what is being asked here is mechanism, not fault. Deciding who is to blame does not move the state, whereas knowing which condition produces the result lets you remove that condition alone.

Saying the cause lies in the structure does not mean there is nothing you can do. If the result arises out of a composition, changing the conditions of the composition changes the result. The only conditions you can change are the ones on your side, and that is enough.

The cost of undoing rises as a function of how long it has run

Simply continuing the same relationship raises the cost of leaving it. You did not do anything in order to become harder to detach. Continuing is what raised the cost.

Economics has treated this as a case where small early events produce an outcome that cannot be reversed. Arthur, analysing competition between technologies, showed that a slight early bias, under conditions of increasing returns, becomes self-reinforcing and can lock the outcome onto the inferior option (Arthur, 1989, The Economic Journal, 99(394), 116–131). The option that gets chosen becomes more advantageous the more it is used, and the more advantageous it becomes the more it is chosen. Past a certain point, a better option appearing no longer produces a switch.

What that analysis takes as its subject is technologies competing in a market. It does not treat an individual’s trading relationships. What can be borrowed from it is the shape of the mechanism: fixing arises not from anyone’s decision but from the repetition of the same choice.

Three things correspond to this on the individual side. The first is accumulated procedure specific to that counterparty: the shape of the exchanges, their internal names for things, the order of approvals. None of it holds value elsewhere, and all of it makes the work faster. The faster it gets, the higher the margin on work for that counterparty.

The second is how your hours fill. From a state where one client’s projects occupy your capacity, there is no time to look for a new counterparty. Because there is no time, the capacity stays occupied, and because it stays occupied, there is no time.

The third is the disappearance of anything to compare against. If you have not looked at terms elsewhere for a long time, you can no longer judge whether your current terms are good or bad. Nobody moves on the basis of something they cannot judge, so all that remains is the conclusion not to move.

Of the three, only the second is easy to notice. That your capacity is full registers as a daily fact. The first shows up as the pleasant sense of getting faster, so it never gets counted as a cost. As for the third, its disappearance is not observed at all. Two of the three do not look like cost.

All three grow in proportion to how long the relationship has run. So “I could leave any time” carries no meaning unless a date is attached to it. What was true as of last year can have stopped being true this year.

An answer to “just get more clients”

Lay out the mechanisms by which dependence becomes impossible to undo, and the remedy looks like spreading the work. Spreading does work, but the range over which it works has limits.

What increasing the count addresses is the quantity of dependence, not the reversibility. Split across three companies, and if all three sit in the same industry, commission work the same way and set their budgets in the same season, then when they move they move together. The count is three and the conditions are one.

Spreading also has costs. More counterparties means accumulated procedure for each of them. Increase the counterparties while total capacity stays fixed and the accumulation per company thins out, which makes you easier for any of them to replace. Spreading works as a defence only while there is slack in total capacity.

The conditions under which it does work are clear enough. Counterparties on different commissioning cycles, with budgets settled at different times, in separate industries. With those conditions in place, one stopping leaves the others running. Whether it works is settled not by the number of companies but by how independent the conditions are.

The largest limit is that spreading does not change the form in which you deliver. Increase the number of counterparties while still working in a form that requires your own hours each time, and the ceiling stays where it is. That point is handled from the side of form in Before you earn on your own, know the ceiling built into selling time.

The reason spreading is nevertheless recommended as the first move is that it can be started sooner than aligning conditions can. Adding one counterparty is something you can do this month; changing the form in which you deliver can take months. The order in which things can be started and the order in which they matter are not the same order. Mix the two and you repeat the easy-to-start move without ever entering the one that matters.

Spreading can also fail as a defence: you added counterparties and ended up accepting only work that any supplier could do. The count looks safe, the rate falls, and nothing accumulates. Ease of exit and the value of staying do not move in the same direction.

The effect of spreading also depends on how you measure it. By revenue share the work may look split across three, while by hours one company occupies eighty per cent — that pattern is not unusual. When you look at shares, pull both the money and the hours. Look at only one and the verdict that you have spread the work comes out first.

None of this rejects spreading. Spreading is one means of raising reversibility, and it is not reversibility itself. Confuse the means with the end and you arrive at a state where the work is spread and you still cannot move.

Where this does not apply: dependence that carries its exit condition from the start

There are dependences that this account does not cover. They are the ones where the condition for ending it is written on the outside at the moment of starting.

A fixed-term contract is one example. If it is settled that it ends in six months, a moment for choosing again arrives automatically. Because its arrival is fixed, reversibility does not have the chance to fall quietly.

The second is where what accumulates takes a portable form. If what you build stays on your side and can be handed to others as it is, continuing does not create fixing. Continuing increases the options instead. How to build that form is handled from the side of the run-up to independence in What you need before going self-employed is a route that reaches people, not savings.

Still, judging portability splits in practice. What you built can remain in your hands and carry meaning only inside that counterparty’s context. So put the test on transferability rather than ownership. Ask whether you could hand the same thing to a different counterparty unchanged, and the line becomes drawable.

The third is where you are the side doing the choosing. With more work than you can take, dropping one leaves the rest to fill the gap. This state carries no guarantee of continuing, though. The fact that work is plentiful loses its value as evidence at the moment work stops being plentiful. Resting on this third case alone is best avoided.

The fourth is where the counterparty holds more than one decision-maker. When the contact is not a single person and decisions are distributed on their side, a change of personnel is less likely to change your position. How fast reversibility falls depends on whether what you depend on is a person or an organisation.

There is one misreading common to all four. Having an exit condition does not mean the relationship is shallow. A fixed term still leaves you free to build whatever you can build inside it. The condition settles how the relationship ends, not how deep it goes.

There is also a form that looks as though the condition sits outside and does not. An automatically renewing contract has dates written into it and still creates no moment for choosing again. It continues unless you take the action of not renewing, so the test has returned to your own intent after all. Distinguish them not by whether a date exists but by what happens if you do nothing on that date.

In every case one thing is common. The condition for ending sits in a mechanism outside you, not in your own intent. Place it in intent and the judgement gets made at the point where that intent is weakest — that is, at the point of greatest uncertainty.

All you can write at the moment of choosing is the exit condition

What cannot be known at the start does not become knowable by trying harder to know it. Rather than making the unknown known, put in place beforehand a form that functions while it stays unknown.

Concretely, that means writing down three things before you begin. The first is the condition for ending this dependence. Write it in quantities observable from outside: “if the revenue share passes seventy per cent,” “if six months pass with no record that carries elsewhere.” Do not write it in terms of how you feel. Feeling dulls at the same rate that reversibility falls.

The second is the date of the review. Write the condition without a date and the review happens “when it starts to bother you.” It starts to bother you, generally, once it is too late. Fix the date in advance and the review occurs regardless of the state.

The third is what you do when the condition is met. Write this as allocation rather than as action: “from next month, move one tenth of capacity to the side where portable accumulation remains.” There is no basis for the figure of one tenth. Wait for a figure with a basis and the moment of starting never arrives.

There is one caution about setting the numbers. Make the condition too strict and months keep meeting it until the review itself becomes a formality. Loosening it to the level at which you would actually shift the allocation is what makes it work in the end. A condition you cannot keep is close to no condition at all.

Writing the three does not take long. What takes time is deciding which quantity to write the condition in. Choose an unobservable quantity here and the review date arrives with nothing to review. Use only what you can count at month end: revenue share, number of records, hours worked.

All three share the property that they can be written at the moment of starting. How much you will earn and whether the work will suit you are both unavailable before you begin. But “what would make me review this” is decidable while nothing has happened yet. In fact it is only decidable while nothing has happened yet.

If things are already under way, the same three can still be written. Take the present as the starting point and settle the condition, the date and the allocation. There is no need to re-evaluate past judgements. Re-evaluating them does not raise reversibility.

Writing the three does not guarantee you can act on them. The month the condition is met may happen to be your busiest. In that case, lower the allocation. Cut one tenth to one twentieth and the record that a review took place does not disappear. With the record in place, the next date can show you the change since the last one. Withdraw the condition itself and that continuity breaks.

Watch the exit side, not the quantity of dependence

This way of looking at it has a price. Writing the exit condition first makes it harder to commit deeply.

Enter a relationship holding a condition and you cannot place everything in it. To that extent, what accumulates stays shallower. The trust you would have earned by concentrating on one counterparty, and the work that could only have been made there, are not available to you. That is a real loss.

Two reasons still favour this side. The first is that what is lost is measurable. Work you could not take because you were holding a condition shows up on the spot. Options lost because you held no condition, by contrast, are not visible at the moment they are lost. Between a measurable loss and an unmeasurable one, choosing the measurable one leaves you with material to judge by.

The second is that a condition does not lower the quality of the relationship. Having decided to review every six months is not something you even need to tell the other side. Told or not, it stays inside the ordinary range of commercial dealing. Whether to commit deeply and whether to hold an exit condition are decided independently of each other.

Both of those are the case put by the side that holds a condition. The side that holds none has a case too: if you have time to write conditions, use it to bring in work. In the first few months that is sometimes the right call. The rightness has a term limit, though. After twelve months of not writing the condition, the cost of not writing it shows up as the reversibility that fell over those months.

Seen this way, the position of the account that treated it as a question of quantity changes as well. Increasing what you depend on was not wrong. What was being increased was read as the number of counterparties. What actually helps is increasing the number of ways out. One counterparty with three ways out is still dependence you can choose. Five counterparties with one way out means the options have already gone.

And this view retires the question of how things began. Whether you chose it yourself settles nothing about whether you can leave now. The memory of having chosen is not evidence that you are still choosing; it is a record that you chose once. Separating those two is where the judgement starts.

Which side your own dependence sits on becomes visible as soon as you write out the cost of ending it. If you cannot write it out, the fact that you cannot is itself the answer.

The way the ceiling is built into how income is made in the first place is mapped as a whole in Getting out of labour-intensive work.