PROCESS

Here is How to Build

Design

Sub

What crosses

Parts that are cut clean, placed by need, and each answering for one result are still separate parts. Where one ends and the next begins there is a seam: the place where one part's result becomes the next part's starting point. A process holds or comes apart at its seams, and the seams inside a process, between its own parts, are the ones most easily left to chance. They are not crossings between processes; everything on both sides belongs to the same process and feels like ours, so nobody designs them. This discipline does.

The first question at any seam is what crosses it. The answer should be one thing: the result the part before answers for, in exactly the form the next part needs. On one side of the seam that result is a done-state; on the other it is the starting state. Nothing else needs to cross. Not the details of how the result was reached, not half-finished work, not data passed along just in case. Each of those turns a seam that carries a result into a seam that carries an inside, and the moment the next part depends on the previous part's inside, the freedom each part earned by answering for its result is gone: change one, and you change the other.

The opposite failure matters as much. If less crosses than the next part needs, it starts without something it requires and has to go looking for it, reaching back across the seam into the part before. That reaching is a thick seam by another route. So the test for what crosses is simple to state: name it, and check that it is the previous part's result and everything the next part needs to begin, and nothing more.

The highest possible standard is for every seam inside a process to carry one thing: the result the part before answers for, in the form the next part needs to begin, so that no inside crosses, no part has to reach back for what it lacks, and each part stays free on its own side.

Key Takeaway: Inside a process, the seam is where one part's result becomes the next part's starting point, and inner seams are the most easily left to chance, because both sides feel like ours. What crosses should be one thing: the previous part's result, in the form the next part needs. More than that (how it was reached, half-finished work, data just in case) makes the next part depend on the previous part's inside and undoes the freedom each part earned. Less than that sends the next part reaching back for what it lacks. Name what crosses, and check it is the result and everything needed to begin, and nothing more.

Let each inner seam carry one thing: the result the part before answers for, in the form the next part needs.

MarvinPro · PROCESS · Here is How to Build · Design · Sub · Discipline 6: The Seams Within · Section: What crosses

MarvinPro | October 2026

marvinpro.com

Who holds it in between

Work that has left one part and not yet reached the next is in the most neglected place in any process. It sits in a queue, an inbox, a file waiting for a batch, a list for someone to pick up tomorrow. It is finished as far as the part before is concerned, and not started as far as the part after is concerned, and so, very often, it belongs to no one. Inside one process this is especially easy to miss, because each part's owner looks after their own side of the seam and assumes the space between is simply part of the flow.

That space is where time goes and where things get lost. A result can wait between two parts for longer than either part takes to do its work, and nobody sees the wait, because nobody is measuring it. A result can stall there entirely and only be found when someone downstream asks where it went. So a good seam has no unowned in-between. Either the next part picks the result up the moment it is complete, so there is no wait to own, or the wait is given an owner and a limit: someone who answers for the result while it sits there, and a point at which a wait too long becomes something that person acts on.

And the seam needs one more thing on the receiving side: a check on arrival. When a result crosses, the part receiving it confirms that it got what it needs before it begins, rather than starting and discovering, halfway through its own work, that something was missing. A check on arrival turns a problem into a quick return at the seam instead of a failure deep inside the next part. It is also what makes the first question real: what crosses is only as good as the next part's confirmation that it arrived.

The highest possible standard is for no result to sit unowned between two parts: picked up the moment it is complete, or held by someone who answers for it while it waits, within a limit, and checked on arrival by the part receiving it, so nothing waits unseen and nothing missing is discovered too late.

Key Takeaway: Work that has left one part and not reached the next sits in the most neglected place in a process (a queue, an inbox, a batch), finished for one side, not started for the other, and often owned by no one. Inside one process this is easy to miss, because each owner looks after their own side. That gap is where time goes unseen and things get lost. So leave no unowned in-between: either the next part picks the result up the moment it is complete, or the wait has an owner and a limit. And the receiving part checks on arrival that it got what it needs, so a missing piece becomes a quick return at the seam, not a failure halfway through.

Leave no result unowned between two parts: pick it up when it is complete, or give the wait an owner and a limit, and check it on arrival.

MarvinPro · PROCESS · Here is How to Build · Design · Sub · Discipline 6: The Seams Within · Section: Who holds it in between

MarvinPro | October 2026

marvinpro.com

The way back

Seams are usually designed in one direction: forward, from the part before to the part after. But sometimes the part after cannot use what arrived, or begins its work and finds it cannot finish because of something that happened on the other side. Then the work has to go back across the seam. And a seam with no defined way back still gets one. It is just invisible: a message to a colleague, a phone call, a shared spreadsheet of exceptions, a workaround that one person knows. Different every time, owned by no one, and measured by nobody.

That invisible way back is how a seam that looks thin turns thick in practice. On paper only the result crosses; in reality a steady traffic of questions, corrections and side conversations flows back and forth between the two parts, carrying exactly the inside knowledge the seam was meant to keep apart. And because it is invisible, it cannot be improved. Nobody sees how often things go back, why they go back, or how long they take to come round again.

So every seam needs a way back defined as clearly as its way forward. What returns: the result, with the reason it could not be used. To whom: the part that produced it, which answers for its result and is the one able to put it right. And what happens to anything already done: if the part before has produced a result that the part after cannot accept, that result may have to be undone, not simply abandoned, and the seam has to say how. A defined way back keeps the exceptions inside the design rather than outside it, and lets them be seen, counted and reduced.

The highest possible standard is for every seam inside a process to have a way back as clearly defined as its way forward: what returns, with what reason, to which part, and what happens to the result already produced, so that exceptions travel through the design rather than around it, and can be seen and reduced.

Key Takeaway: Seams are usually designed forward only, but sometimes the next part cannot use what arrived, and the work must go back. A seam with no defined way back gets one anyway: a message, a call, a spreadsheet of exceptions, invisible and different every time. That is how a seam that looks thin turns thick in practice, with questions and corrections flowing both ways and nobody able to see or improve them. So define the way back as clearly as the way forward: what returns and why, to the part that produced it, and what happens to the result already produced, which may have to be undone rather than abandoned.

Give every seam a way back as clearly defined as its way forward: what returns, why, to whom, and what happens to what is already done.

MarvinPro · PROCESS · Here is How to Build · Design · Sub · Discipline 6: The Seams Within · Section: The way back

MarvinPro | October 2026

marvinpro.com

A real example

Return to activate and look at the seam between settling the payment and opening the account. What crosses it should be one thing: the payment settled, together with the customer and the plan the account needs. Nothing more. In particular, the card details stay inside settling. Opening the account never needs them, and carrying them across would thicken the seam and spread sensitive data into a part that has no reason to hold it. Settling answers for a settled payment; opening the account begins from one. That is the whole of what crosses.

Now look at the in-between. Suppose settling finishes at any hour, but opening the account runs in a batch each morning. A customer who pays at eleven at night has paid and has no account for nine hours, and nobody owns that wait: settling considers its work done, and opening the account has not started. If the batch fails one morning, those customers stay stranded until someone notices. The fix is to remove the unowned gap: either opening the account starts the moment a settled payment arrives, or the wait gets an owner and a limit, with someone who sees every customer waiting and acts if any wait runs past the limit. And on arrival, opening the account checks that the customer and plan are present and valid before it begins.

Then the way back. Suppose the account cannot be opened, because the customer's email is already in use on another account. The customer has paid and has nothing. With no defined way back, someone notices a complaint and sorts it out by chat, differently each time. The fix is a defined return across the seam: opening the account sends the case back to settling, with the reason; settling reverses the payment, reaching the finance process just as the original charge did, and tells the customer, through the same notice it already sends when a payment is declined. The settled payment is undone, not abandoned, and every such case is visible and counted.

The seam between settling and opening carries only the settled payment and what the account needs, has no unowned wait, is checked on arrival, and has a defined way back that undoes the payment and tells the customer.

MarvinPro · PROCESS · Here is How to Build · Design · Sub · Discipline 6: The Seams Within · A real example

MarvinPro | October 2026

marvinpro.com

The Seams Within

Parts cut clean, placed by need, and each answering for one result still meet at seams, and the seams inside a process are the most easily left to chance, because both sides feel like ours. What crosses an inner seam should be one thing: the result the part before answers for, in the form the next part needs to begin. More than that makes the next part depend on the previous part's inside and undoes the freedom each part earned; less than that sends it reaching back for what it lacks.

The space between two parts is where time goes unseen and things get lost: work finished for one side and not started for the other, often owned by no one. So leave no unowned in-between. Either the next part picks the result up the moment it is complete, or the wait has an owner and a limit. And the receiving part checks on arrival that it got what it needs, so a missing piece becomes a quick return at the seam rather than a failure halfway through its own work.

And every seam needs a way back. A seam with none still gets one, invisible and different every time, and that hidden traffic is how a seam that looks thin turns thick in practice. So define the way back as clearly as the way forward: what returns and why, to the part that produced it, and what happens to the result already produced, which may have to be undone rather than abandoned.

So the discipline is this: let every inner seam carry only the result, leave nothing unowned between parts, check on arrival, and give every seam a defined way back. Joined this way, the parts stop being pieces placed side by side and hold as one process. Cut, placed, given their functions, and joined at clean seams: that is a process built of sub-processes that works as a whole.

Join every part at a seam that carries only the result, leaves nothing unowned in between, checks on arrival, and has a defined way back, so the parts hold as one process.

MarvinPro · PROCESS · Here is How to Build · Design · Sub · Discipline 6: The Seams Within · Chapter Outcome

MarvinPro | October 2026

marvinpro.com

Think Simple.