The Interface Tax

There is a line item in your business that nobody has ever put in a budget, nobody reports on, and nobody owns. It does not appear in headcount. It does not appear in cost of goods. It gets paid every single week, in cash and in calendar time, and the larger and more competent your company becomes, the more of it you pay. It is the cost of the seam between the people who make the promise and the people who have to keep it.

I have spent twenty years working on both sides of that seam, and for roughly twelve of those years I have held both sides at once. Not sequentially, and not as a title stretched across a slide. At Texas Armoring Corporation and again at ReadyNet Solutions, I carried the marketing mandate and the operating mandate at the same company, at the same time, against the same P&L. The reason was never that I wanted a bigger job. It was that I could not find a version of the org chart where splitting those two mandates did not immediately start costing money.

I have come to think of that cost as an interface tax, and I think it is the most expensive thing most companies never measure.

Section 01Why the tax is invisible

Every functional leader in a healthy company can show you a scorecard where they are winning. Marketing hit its qualified pipeline number. Sales hit bookings. Operations hit on-time delivery and held gross margin. Every one of those numbers is real, and the customer still had a bad experience, and the launch still slipped a quarter, and nobody in the room is lying.

That happens because the failure did not occur inside any function. It occurred between two of them, and no function’s variance report has a column for the space in between. A promise gets made in one language and received in another. A specification gets thrown over a wall and lands slightly wrong. A launch date gets committed in a customer meeting and then arrives at the manufacturing floor as an aspiration. Each handoff loses a little fidelity and a little time, and because the loss is distributed across two owners, it belongs to neither.

This is not a new observation, and it is not mine. Chip Heath and Nancy Staudenmayer named the underlying pathology in 2000 and called it coordination neglect.¹ Their argument is that when people organize work, they instinctively spend nearly all of their attention on dividing the task and almost none on reintegrating it. They called the first bias partition focus, the reflex to keep splitting work into cleaner boxes, and the second component focus, the tendency of each specialist to optimize their own box while treating the interfaces as somebody else’s problem. Staudenmayer taught at Duke’s Fuqua School of Business, where I later did my MBA, and I did not encounter the paper until years after I had already spent a decade paying the tax it describes.

The finding that matters most is the quiet one buried in the middle of the paper: integration is not a communication problem people solve once they notice it. It is a problem people systematically fail to notice at all, even when they are actively thinking about coordination and have every incentive to get it right. Nobody wakes up choosing to under-invest in the seams. They simply cannot see them, because the seams do not have a headcount, a budget, or a name.

Melvin Conway made the structural version of this point in 1968, in what has become one of the few genuinely predictive laws in management: organizations produce systems whose structure mirrors the communication structure of the organization itself.² Conway was writing about software, but the observation is not about software. If your brand team and your production team communicate through a quarterly planning meeting and a shared spreadsheet, then you will ship products whose commercial story and physical reality are joined by roughly the strength of a quarterly meeting and a shared spreadsheet. Customers can feel that joint. They just call it something else. They call it “the website said one thing and the product did another.”

Section 02What you are actually paying

In my experience the tax gets collected in three currencies, and it helps to separate them, because each one has a different remedy.

1. The latency tax. Decisions that require information from both sides move at the speed of the slowest scheduled conversation between them. If marketing needs an operating answer to price a configuration, and the only forum where that answer exists is a weekly meeting, then the effective decision latency for every pricing question in the company is one week, regardless of how fast either team can think. Multiply that by the number of cross-boundary decisions a year and you have quietly capped your organization’s clock speed at a number nobody chose.
2. The translation tax. Every handoff is a re-encoding, and every re-encoding loses something. A customer insight arrives at product as a feature request. A capacity constraint arrives at marketing as a launch date. The nuance that made the original observation useful, the why underneath it, does not survive the trip, because the trip is optimized for transferring conclusions rather than reasoning. What arrives on the other side is a directive without a rationale, which is the least actionable object in business.
3. The truth tax. This is the expensive one and the one nobody says out loud. When two functions have been burned across a seam a few times, each begins to privately discount the other. Operations starts adding buffer to every timeline because marketing has historically over-promised. Marketing starts committing to softer language because operations has historically under-delivered. Both adjustments are individually rational. Together they mean the company’s internal numbers are no longer the company’s actual numbers, and every plan built on them inherits an error nobody can locate.

The truth tax is why I am skeptical of the standard remedy. When companies notice friction between marketing and operations, they almost always respond by adding a coordination layer: a program manager, a revenue operations function, a shared dashboard, a new recurring meeting. Sometimes those help. But you cannot fix a trust problem with a reporting structure, and a coordination layer that sits between two functions who have stopped believing each other’s numbers just adds a third set of numbers.³

Section 03The version I ran instead

At Texas Armoring, the single most profitable thing we ever did was engineer configuration upsells directly into the production process. A base armoring package regularly expanded three to four times in value through security, performance, and luxury options. Per-sale margin went up forty percent. Raw-material spend came down thirty-five percent through supplier renegotiation and category consolidation. Net income grew more than three hundred percent. Revenue went from two million to twenty-five million dollars across the decade, compounding rather than spiking, with fifty-five percent growth in year two, forty percent in year four, and thirty percent in year six.

Here is the part that matters for this argument. The decision about which options to offer, how to frame them, and what a customer would perceive as worth paying for, was a marketing decision. The decision about how to sequence the production line so those options could be added without breaking takt time or blowing up the bill of materials, was an operating decision. In most companies those two decisions live with two people who report to two different executives and have two different bonus structures.

Split that way, one of two things happens. Either the upsell never gets invented, because the marketer proposing it has no visibility into what the line could actually absorb and therefore proposes something unbuildable and gets told no. Or the upsell gets invented, gets sold, and then gets built badly, because the operating team received it as a mandate rather than as a shared bet, and had no say in the design of the thing they now have to manufacture.

Neither happened, because both decisions were made by the same person against the same P&L. That is not a claim about my talent. It is a claim about a structure. There was no interface to lose anything across.

Section 04The honest objection

The obvious rebuttal to everything above is that specialization exists for a reason, and it does. Adam Smith was not wrong about the pin factory.⁴ You cannot run a two-hundred-person manufacturer by having one person hold every mandate, and anyone who tells you the answer to organizational complexity is simply to merge more roles into fewer heroes has never had to actually staff the resulting job.

So I want to be precise about the claim. I am not arguing against division of labor. I am arguing that division of labor gets designed and integration gets improvised, and that the asymmetry is the whole problem. Companies run structured, deliberate, hotly-debated processes to decide how work should be split. They then leave the question of how it gets rejoined to goodwill, proximity, and whoever happens to be in the hallway. Integration deserves the same design rigor that partition already receives. It almost never gets it.

Which is why, for the ninety-five percent of situations where you genuinely cannot put both mandates in one seat, there is a second answer, and I built it at Texas Armoring because I had to.

Section 05Internal service standards, and why they are not bureaucracy

We wrote down what each function owed the next one. Not values. Not a RACI chart. A specific, cross-departmental service-standards system that defined what marketing owed sales, what sales owed production management, and what production management owed each of seven manufacturing disciplines: raw material, fabrication, welding, electrical, trim, upholstery, and paint and body.

People hear “internal service standards” and think bureaucracy. They are close to opposites. A standard is a promise with a deadline and a named owner. Bureaucracy is a process with neither. If a document tells you a form exists but not who is accountable for what it produces or by when, that is bureaucracy. If it tells you exactly what sales will receive from marketing, in what format, within how many days of a qualified inquiry, and who to escalate to when that does not happen, that is not bureaucracy. That is the removal of a daily argument.

A usable standard has exactly four parts, and if any one is missing you have written a preference rather than a standard:

• The deliverable. Named, specific, and defined by what the receiving function needs, not by what the sending function finds convenient to produce.
• The timeframe. An actual number of hours or days, tied to a triggering event rather than to a calendar meeting.
• The acceptance criterion. The one thing the receiver checks to know it is done. Without this, “done” is a negotiation, and negotiations cost more than the work.
• The escalation path. A named human, not a queue. If a standard has no escalation path, it will be quietly broken within a quarter and nobody will report it.

What that system bought us was not politeness. It was speed. When the interface is written down, the daily cost of crossing it collapses toward zero, because nobody has to renegotiate the terms of the handoff every time a handoff occurs.

Section 06Three diagnostics you can run this week

If you want to know your own tax rate, you do not need a consulting engagement. You need about ninety minutes.

1. Ask the same question on both sides of the wall, separately. Ask your head of marketing what your standard lead time is. Ask your head of operations the same question. Do not let them confer. The delta between the two answers is your translation tax expressed in days, and it is the number your customers are currently experiencing as inconsistency.
2. Count the artifacts that exist only to move information between two functions. Status decks, handoff templates, sync agendas, the tracker somebody maintains by hand. Every one of them is a toll booth built where a road should have been. They are not waste in the sense that they can simply be deleted; they are waste in the sense that they are the visible price of an interface you never designed.
3. Take the last five things that took longer than expected, and locate the time. Not who caused it. Where it went. In my experience the answer is almost never “a function was slow.” It is almost always “the work sat, fully complete, waiting to be received.” Queue time between functions, not work time inside them, is where the calendar disappears.

Section 07Why this matters more now, not less

Two forces are making this problem more expensive rather than less. The first is that commercial cycles have compressed while operating complexity has not. A brand promise can now be made publicly, at scale, in an afternoon. The capacity to keep it still moves at the speed of tooling, hiring, certification, and freight. The gap between how fast you can say something and how fast you can do it has never been wider, and that gap is exactly where the interface tax is collected.

The second is artificial intelligence, which is frequently sold as a coordination solution and mostly is not one. AI is extraordinarily good at reducing the cost of producing artifacts and quite bad at resolving the question of who is accountable for a decision. If two functions disagree about a lead time, giving both of them a language model produces two more confident documents, faster. The technology accelerates whatever the organization already does. If what your organization does is throw work over walls, you have just bought a faster catapult.

I have never in twenty years watched a good marketer fail. I have never watched a good operator fail either. What I have watched fail, over and over, in companies full of genuinely excellent people, is the space between them. That space is a design choice. Somebody made it, usually years ago, usually for a reason that no longer applies, and almost always without anyone writing down what it would cost.

You can keep paying that tax. It is entirely survivable; most of your competitors are paying it too, which is why the market tolerates it. But it is worth knowing that you are paying it, and roughly how much, because it is the rare structural cost that a single deliberate decision can actually remove.