The Invisible Bill of Materials

Ask the owner of a service business for their bill of materials and you will usually get a polite version of the same answer: we don’t have one, we don’t make anything. Ask the founder of a software platform and the answer gets even more confident, because software famously costs nothing to copy. No shelf, no boxes, no bill of materials.

Both answers are wrong, and they are wrong in a way that costs money every week. What they actually mean is: our product is invisible, so we have never taken it apart. And a product nobody has taken apart gets priced by feel, scoped by optimism, and defended by memory. If pricing your work feels like negotiation with yourself, this is usually why.

I have made the argument before that every business sits on a product, and that once you decompose that product down to its atoms, everything above it (pricing, quoting, scheduling, selling, even marketing) stops being guesswork and becomes arithmetic. I later argued that the eight phases built on that idea rest on principles that say nothing about manufacturing. This piece takes the claim to the place where it has to work hardest: businesses where there is nothing physical to point at. If the decomposition discipline only works when the product sits on a shelf, it is a manufacturing trick. If it works when the product is a week of skilled work, or a subscription that never ships anything, it is a principle. So let us take an invisible product apart and see.

What the atoms are when nothing is physical

Start with what decomposition actually produced in the factory, because the list is instructive: every material, every labor step, the measured minutes each operation takes, and the slice of overhead each unit absorbs. Notice that only the first item is physical. Three quarters of a manufacturer’s bill of materials was never on the shelf either. It was time, skill, and burden. You already know how to decompose invisible things. Manufacturers just got forced to do it first, because the visible quarter dragged the rest along.

So run the same move on a service. Imagine a caterer, because a caterer is halfway between the factory and the invisible. The menu decomposes beautifully: each dish into ingredients, prep minutes, and plate cost. But the event, the thing actually being sold, has more atoms than the food. Staff hours by role, because a chef’s hour and a server’s hour cost different amounts. Travel and load-in time. Rentals consumed. The quiet overhead of quoting, planning, and the site visit nobody bills for. An engagement is a bill of materials in which most of the materials are minutes.

Now do the platform, where the illusion is strongest. Imagine a class-booking platform used by local studios. Copying the software costs nothing, true. Serving an account does not. Every studio on the platform consumes infrastructure, support minutes, onboarding effort, and a share of the roadmap’s attention. Those are the atoms, and they are lumpy: one studio files a ticket a quarter, another files one a week and asks for a custom report on top. A subscription price is a bet about the average of those atoms, and most platforms place that bet without ever measuring what one account actually consumes. “Software has no marginal cost” is how the bet stays unexamined.

Three different columns of atoms, one identical stack underneath. Price, quote, promise, publish: the commercial layer does not care whether the atoms were boards, hours, or support tickets. It only cares whether the atoms are real. That is the entire transfer, in one picture. What follows is what happens to a few of the framework’s principles when you run them through the invisible columns, because a couple of them get sharper, and one of them gets uncomfortable.

Decompose before you systematize

In a factory, the un-decomposed product announces itself: jobs run over, material runs out, quotes come back wrong. In a service business the same disease has a politer name, and you have almost certainly said it out loud: scope creep. Scope creep is not a client behavior problem. It is what an un-decomposed product looks like from the inside. If the engagement was never taken apart into units (what is included, how many revisions, how many site visits, how many hours of which skill), then the edge of the product is wherever the client last pushed it. You cannot defend a boundary you never drew.

The test I use for manufacturers carries over verbatim, and it is still strict. Hand your notes to a stranger and ask them to price one engagement. Not your instinct, your notes. If they cannot get to a rough number, your product is not yet decomposed, and everything above it (the proposal, the schedule, the profit) is inheriting that gap. A landscaping contractor who can say “a crew day is eight labor hours, this much equipment time, this much fuel and disposal” passes the test. “It depends on the yard” does not. Of course it depends on the yard. The question is whether you have named the parts of the yard it depends on.

One source of truth

The factory version of this principle was one relational core connecting products, projects, and clients, so every question gets answered from the same place. The invisible-product version is more urgent, not less, because services and platforms fragment faster than factories.

A service business drowns in per-client folders. Imagine a repair shop: the job is in one app, the parts are in a spreadsheet, the customer history is in someone’s phone, and the actual diagnosis lives in the technician’s head. Four memories, all slightly different. Every estimate is a small archaeology project. A platform drowns the other way, in tool sprawl: usage in one dashboard, support in another, billing in a third, and nobody can answer “what does our heaviest account actually cost us” without an afternoon of exports. In both cases the fix is the same shape as the factory’s: one structure where the engagement, its atoms, and its client connect, so the question “what did this actually take” has exactly one answer. If you can only fix one thing this year, fix this. Every other principle in this piece runs on it.

Every completed job tests the system

Here is the strange part: service businesses are better positioned for this principle than factories were, and they waste the advantage almost universally.

The factory had to earn its feedback. Measuring actual minutes per operation took deliberate effort, and the payoff was real: one operation I had estimated at over twenty minutes measured out near seven, a roughly threefold error that had been quietly mispricing work for years. A service business already has the measurement layer. It is called a timesheet. The hours are being written down somewhere, for payroll if nothing else. Which means the estimate-versus-actual comparison, the single most valuable feedback loop in the whole framework, is sitting in your filing cabinet already, unread. Imagine the caterer comparing quoted kitchen hours against what the event actually consumed, event after event. Within a season they would know which menus lie. Most never look. The data is not missing. The habit is.

The platform’s version is cost-to-serve. Every account is a completed experiment in what a subscription actually costs: the support minutes, the storage, the onboarding calls. A photo-storage subscription that never measures per-account consumption is quoting fiction with a straight face, because averages hide the tail, and the tail is where the money goes. Grade every job. The grade is the product improving itself.

Price as calculation, not improvisation

In the factory, a quote is not a creative act. The invisible-product economy mostly has not gotten the message. The hourly rate, as commonly practiced, is a guess wearing a suit: it was set by looking at competitors, it absorbs every inefficiency invisibly, and it gets discounted under pressure precisely because nobody can say what the floor is. When you cannot see your costs, every negotiation is a fight with fog.

Assembled pricing changes the posture entirely. The landscaping bid built from decomposed components (area, crew hours, equipment time, disposal, margin) can be defended line by line, and more usefully, it can be corrected line by line when the actuals come back. The platform tier priced against measured cost-to-serve can survive its heaviest accounts instead of quietly subsidizing them. Notice what changed: not the price, necessarily. The confidence. A calculated price might be the same number as the guessed one. But only one of them teaches you anything when reality disagrees with it.

The loop, not the pipeline

The part of the framework I called the strongest evidence it is a framework at all is that phases five through seven form a loop, not a line: what actually happened corrects what things cost, what things cost disciplines what gets promised, and every promise is tested against the next delivery.

Strip the factory away and the loop does not just survive. It becomes easier to see. The service version of promise drift is scope creep again, viewed from the selling side: proposals promise what delivery cannot profitably do, because the promise was never connected to measured delivery truth. The platform version is the roadmap and the sales page writing checks the support queue has to cash. In every case the cure is structural, not motivational. Connect the three corners so they correct each other, and the promises calibrate themselves. Leave them disconnected and each corner drifts toward its own comfort: delivery pads, pricing wishes, promising sells. The triangle is not a metaphor. It is a wiring instruction.

What fights back

I want to be honest about where this transfer strains, because the strain is real and pretending otherwise would make this marketing. Three objections come up, and they deserve straight answers.

“Our atoms vary too much.” People are not machines and no two engagements are alike, so the minutes wobble. True, and the factory’s minutes wobbled too. The answer is not to abandon measurement but to measure the wobble: estimate-versus-actual per skill tier, per job type, over enough jobs. Variance is not the enemy of the method. Variance is data. A range you have measured beats a point you have invented, every time, and the range narrows as the loop closes.

“Software really is nearly free to copy.” The marginal-cost illusion is the platform’s most comfortable belief, and inside it hides the tail. The hundredth light account costs almost nothing; the heaviest five accounts can consume more support and infrastructure than the next fifty combined. Decomposition does not deny cheap copies. It finds the accounts where the copy is the cheapest part of the relationship. If the illusion were fully true, no platform would ever be surprised by its own support load, and most are.

“Our work is creative. It doesn’t repeat.” The caterer already answered this one. The menu is decomposed; the event is composed from it. Nobody thinks a decomposed menu makes the food less creative. It is what lets the kitchen promise a hundred plates at seven and mean it. Creativity lives in the composition. The atoms underneath it can still be named, costed, and measured, and the naming is what makes the creative promise keepable. What actually does not repeat is rarer than it feels from inside, and the parts that truly do not repeat deserve to be priced as exactly that.

The same boring first move

If you run a factory, the blueprint will turn all of this into a concrete starting plan, and the shape it produces is documented. If you run anything else, the honest version of this piece is smaller and harder than a tool: your product is invisible, which means nobody will ever force you to decompose it. No shelf will run empty. No material will run out mid-job. The forcing function has to be you.

So run the strict test once, this week. Pick one engagement, one tier, one job, and hand your notes to a stranger. If they can price it, you are further along than most, and the loop is waiting to be wired. If they cannot, you have found the real project, and it is not the proposal template or the new tool. It is the boring, load-bearing work of naming what you actually sell, one atom at a time, until the invisible product explains itself. Everything clever you want to build afterward is a function of how honestly you do that first.