Photo by Roma Kaiuk🇺🇦 on Unsplash

The joke is old because it keeps working.

One person’s résumé says:
Single-handedly managed the successful upgrade and deployment of a new environmental illumination system with zero cost overruns and zero safety incidents.

Someone watching the work says:
He changed a light bulb.

We laugh because we recognize the move. A simple act gets dressed for travel. The sentence grows because it has to survive distance: from the work to the approver, from the moment to the slide, from reality to a forum where certainty is expected and nuance is expensive.

That inflation usually feels harmless. A light bulb is still a light bulb. Nothing structural breaks because someone used forty extra syllables.

The problem starts when we treat that habit (compression, simplification, “executive readability”) as neutral everywhere else.

In real systems, language is load-bearing long before it becomes decorative.

Most organizational failures don’t happen because no one knows the answer. They happen because the answer exists only somewhere, instead of anywhere it needs to be. It lives in a person, not in a form that can be inspected, repeated, or relied on when decisions have to be made under pressure.

Ask Jordan; he knows how the customer hierarchy works.

For a while, this looks like excellence. In hindsight, it’s usually a lease.

The system is “understood” as long as the right person is present to translate it. Progress is possible as long as they’re in the meeting, answering follow-up questions in real time. But when that person gets promoted, burned out, reassigned, or leaves, the organization doesn’t just lose expertise. It loses the map.

That’s what I mean by illumination.

Illumination isn’t knowing things. It’s building conditions where the same truth is available to the people who need it, in the same words, with the same boundaries, whether or not the original knower is in the room; so work can move without heroics or priesthood.

One person can change a light bulb. An organization becomes reliable only when it can keep the lights on without that person standing there, explaining how the switch works.

Category Collapse

The light-bulb joke works because the stakes are trivial.

A simple act gets reworded in business-casual. A plain sentence becomes a polished one. A small task turns into a “successful initiative.” A single decision becomes “cross-functional leadership.” Most of the time, this inflation is harmless. Irritating, perhaps, but rarely dangerous.

The failure doesn’t come from exaggeration. It comes from applying the same compression habit to things that cannot survive being vague… and organizations routinely confuse shared labels for shared meaning.

Words like customer, revenue, margin, active, churn, booked, shipped, and recognized get treated as if they name a single, stable object everyone can see. Once named, they’re assumed to be aligned. Dashboards get built. KPIs get standardized. Decisions get made.

The label does the work. Or so we pretend.

That’s category collapse: when multiple, materially different real-world concepts are flattened into a single convenient noun, and the organization proceeds as if the noun were precise enough to carry operational weight.

It rarely is.

This is how multiple teams can be “right” at the same time while the business still breaks. Finance can be correct about revenue. Sales can be correct about bookings. Marketing can be correct about pipeline. Each is describing a different reality under the same word, with different timing, different obligations, and different consequences. No one is lying. None of the numbers agree. The conflict isn’t interpersonal; it’s architectural.

The problem isn’t that people describe work differently. That’s inevitable. Perspective is normal. The problem is acting as though those descriptions are interchangeable.

Teams operate in parallel realities that only appear consistent from a distance. Up close, the systems disagree, decisions fight each other, and governance turns into arbitration.

Because if the nouns aren’t shared, the verbs can’t be trusted. And if the verbs can’t be trusted, execution is just motion; confident, well-reported, and quietly misaligned.

Define a Customer

If you want to locate the fracture line in an organization’s so-called “shared understanding,” ask a question that feels too basic to argue about:

Define a customer.

Most people answer immediately, with the same confidence they’d answer define a chair. That confidence is exactly why the question works. It feels like vocabulary. In practice, it’s architecture.

Ask Finance and you’ll usually get a definition shaped by obligation: the entity that gets invoiced, carries credit risk, and makes the ledger reconcile. In system terms, it orbits bill-to and payer relationships.

Ask Sales and you’ll get a definition shaped by ownership and persuasion: the account they sell to, the territory they carry, the relationship they manage. Sometimes that aligns with a sold-to. Sometimes it aligns with Cathy, the main contact who makes deals real enough to forecast.

Ask Operations or Customer Service and you’ll get a definition shaped by consequences: the location that receives product, rejects late deliveries, has special handling notes, or calls when something arrives damaged. In most systems, that’s the ship-to.

Ask Marketing or eCommerce and you’ll get a definition shaped by identity and reach: the email address, the user account, the cookie stitched to a profile. That “customer” may not exist in your ERP at all.

None of these answers are irrational. Each is locally correct. Each solves a real problem. The failure begins with what happens next.

The organization takes the shared label, “customer”, and treats it as a single, stable noun. Dashboards get built. KPIs get standardized. Targets get set. Governance documents get written as if the word itself guarantees common meaning.

Then, quietly, the systems start to disagree.

One dashboard reports customer growth as new billing entities. Another reports it as new ship-to locations. A third reports it as net-new eCommerce users. All three trend upward. All three can be “right.” And if no one has forced the context into the open, the disagreement becomes personal: whose numbers are wrong, whose report can’t be trusted, who is “gaming” the metric.

But this is rarely a definition problem. It’s a shared reality problem.

Because customer is almost never a single thing. In the real world it’s a composite: multiple entities bound by relationships. A parent company with subsidiaries. A buying group with distinct ship-to locations. A distributor purchasing on behalf of end users. A hospital system with separate billing structures. A franchise network that behaves like one brand and twenty legal entities.

Even in environments where structure is supposed to save you, ERP deployments included, you still have to choose which level of the hierarchy counts for this decision.

• Is the customer the bill-to with multiple sold-to accounts beneath it?
• Is it each sold-to independently?
• Do ship-to locations count separately?
• What happens when payer and bill-to diverge?
• What happens when acquisitions leave two systems disagreeing on who the entity even is?

If your answer fits in a single sentence with no qualifiers, you’re probably not defining a customer. You’re naming a convenience. Conveniences are fine… until you build compensation plans, churn narratives, pricing logic, and “strategic” forecasts on top of them.

The point here isn’t to argue for a universal definition. It’s to expose the cost of pretending the definition doesn’t matter. Without explicit context, the label becomes a shell game: the same word rotating quietly between meanings depending on who is speaking and what outcome they’re incentivized to defend.

The danger of missing context

If you’ve read The Fractal Delusion, you already know the punchline: motion isn’t proof. What changes here isn’t the lesson, but the object it’s applied to.

In that essay, the risk was structural work being mistaken for visible progress; a bridge that looks complete from the road but hasn’t been asked to carry weight. In organizations, the same failure mode shows up one layer earlier, before the work even runs. It shows up in language.

Not because people are dishonest, but because abstraction has physics. Context is what gets compressed away.

As information travels upward, it has to survive distance. It has to fit inside a meeting. It has to be comparable across initiatives. It has to become “executive readable.” And like all compression, it is not neutral. It strips texture, rounds sharp edges, and turns conditional truths into stable nouns.

This is how category collapse becomes normal.

• “Customer” becomes a single word that fits on a chart.
• “Revenue” becomes a line that can be forecast beyond the current month.
• “Active” becomes a number that can be defended.

The dangerous part is not that these simplifications are wrong. It’s that they work well enough.

They travel cleanly and produce confident decks. They create the appearance of alignment because everyone can point to the same label and believe they’re talking about the same thing.

But without invariants, truth becomes portable, and non-binding.

An invariant is not a preference or a convention. It’s a condition that must remain true for the system to hold: what counts, what doesn’t, which hierarchy level the organization is committing to, which exclusions are real, and what fails when those assumptions drift.

Without those guardrails, a word like customer becomes a suitcase term; packed differently depending on who’s carrying it and where they’re going.

That’s why two dashboards can both be “accurate” while disagreeing. They aren’t lying, just operating under different contexts that were never made explicit. This is also why governance disputes turn personal so quickly.

When the system fails to preserve definitions, enforce boundaries, or surface assumptions, the only thing left to defend is credibility. People protect “their” number because the organization never built a shared reality strong enough to make the number belong to everyone.

The same mechanism turns status lights into theater.

Without proof thresholds, green stops being a demonstrated condition and becomes a narrative judgment. Green means nothing has failed loudly yet. The work looks coherent from altitude. We believe it will hold.

Abstraction smooths sharp edges because sharp edges complicate the story. They force caveats. They trigger decisions no one has time (or appetite) to make in the current meeting. But those edges, the qualifiers, hierarchy choices, exclusions, and failure modes, are exactly what make the foundation load-bearing.

When they’re compressed away, the organization doesn’t lose information dramatically. It loses binding force. Truth stops surviving handoffs intact and starts arriving as interpretation.

Once truth becomes negotiable, execution becomes political.

The fix isn’t to eliminate compression; scale makes abstraction unavoidable. The fix is to decide deliberately which truths are not allowed to be smoothed into convenience, and to build structures that force them to survive the trip upward without changing shape.

Knowledge isn’t a person

Here’s the missing organ in most organizations: the mechanism that turns individual clarity into shared leverage.

We talk about subject matter experts as if expertise were a stockpiled asset. Hire the right person, and uncertainty recedes. Put them in the meeting, and the system becomes legible. Questions get answered. Decisions get made. Progress resumes.

For a while, this works. But illumination does not scale through talent alone.

Organizations don’t gain leverage by employing someone who can answer every question. They gain leverage when the same answer is available to everyone who needs it, in the same words, with the same boundaries, without requiring that person to be in the meeting.

That’s the difference between knowing the expert and knowing the answer.

When knowledge lives primarily inside a person, that person becomes a routing layer. Questions don’t flow through artifacts; they flow through calendar invites.

Decisions aren’t made because the system is clear. They’re made because the right individual was in the room to translate what the system “really means.” Over time, meetings become translation hubs. Progress becomes gated not by work, but by interpretation.

This is the quiet mechanics of dependency:

• Work blocks on clarifying questions that should have been answered once.
• Teams recreate the same definitions in slightly different language.
• Dashboards multiply because everyone builds the version of reality they can defend.
• Governance turns into debate because there’s no artifact strong enough to settle it.

Nothing looks broken. Output still ships, and the system appears functional.

Then the expert leaves.

When that happens, the organization doesn’t just lose a person. It discovers, usually too late, that the person took the map with them.

This is why so many “well-run” systems feel haunted after turnover. Everyone remembers that things worked. No one can explain why they worked. Rituals persist because people remember the consequences of breaking them, even if the logic has been forgotten. The organization keeps producing output while comprehension quietly atrophies underneath it.

This is the line that matters: Expertise that cannot be transmitted becomes dependency, not capability.

Real capability is not an impressive brain in a room. It’s shared artifacts: definitions that survive handoffs, standards that don’t shift based on who is speaking, and invariants that remain true even as people change.

Illumination exists when answers outlive their authors. If you want to know whether an organization has illumination or just talent, ask a single question: Can the next person find the answer without scheduling a meeting?

If the answer is no, the organization isn’t operating on shared truth. It’s operating on heroics; and heroics do not scale, no matter how many experts you hire.

Building reliable context

If illumination is the condition where truth is shared, then the next question is the only one that matters:

How do you build light sources that survive turnover, scale, and pressure?

The answer is not another policy binder. It isn’t a set of “best practices,” and it definitely isn’t a slide deck about alignment. What’s needed are primitives: small, durable structures that make shared reality cheaper than improvisation.

---

1) Defined nouns

Start with the words that steer decisions: Customer. Revenue. Active. Margin. Shipped. Recognized.

A real glossary doesn’t describe what a thing is in the abstract. It defines the boundaries that make the word operational.

• What it includes.
• What it explicitly excludes.
• Where reasonable people will disagree unless you choose a rule.

“Customer” at which level: bill-to, sold-to, ship-to, payer, parent? Does “active customer” mean purchased in the last 90 days, has an open contract, placed an order, or simply exists as a maintained account? Do distributors count as customers, or are they channels? What about end users?

Examples, deployed here, are not for decoration. They act as enforcement. If you can’t show three concrete examples and at least two exclusions, you don’t have a definition; just a story. And stories don’t survive handoffs.

---

2) Defined math

Once the nouns are stable, define the numbers that everyone pretends are self-evident.

A metric is a claim. It says: this is what we mean by success, health, growth, or risk. Claims need to be stated like engineering, not marketing.

• Which fields?
• Which filters?
• Which timing rules?
• Which currency logic?
• Which deduplication logic?
• Which hierarchy level? (again.)

Then comes lineage.

Where does the number come from? What systems feed it? What transformations occur along the way? If someone cannot trace a reported KPI back to source systems without calling three people and opening a calendar invite, the organization is not measuring. It is performing measurement to compensate for the missing ruler.

This is the smallest structural shift that ends the lifestyle where “two dashboards are both right in different ways.”

---

3) Controls that keep truth true

Definitions alone do not hold. Truth in organizations doesn’t drift because people are careless. It drifts because systems change, incentives shift, and exceptions accumulate.

So shared reality needs enforcement.

This is where governance stops being ceremonial and becomes operational:

• Clear master data ownership, with change management and validation rules
• Configuration reviews that prevent silent behavior changes
• Data contracts and validation checks so upstream breaks surface immediately
• Tripwires that block “green” status when uncertainty is present

If rollback hasn’t been executed end-to-end, readiness cannot be claimed. If a manual control fires more than n times per period, it is no longer temporary. If deduplication exceeds n% for two consecutive runs, the numbers are not “healthy.”

---

4) Staffing as a control surface

This is the part most organizations avoid saying out loud: Knowledge work requires time and continuity.

You cannot build shared reality on top of constant interruption, perpetual triage, and turnover treated as a normal operating cost. Understaffed governance does not produce “lean operations.” It produces rented competence, because the only way to survive is to buy clarity from outside or to lean on a few exhausted people who keep the rituals alive.

Governance is not overhead. It is an investment in survivability.

If an organization will fund new tools, platforms, and integrations but will not fund the people and time required to define, validate, and enforce shared truth, it is not unlucky. It is choosing dependency.

---

These four primitives aren’t glamorous. They sound like common sense, and that will never demo well. But they build the thing organizations claim to want: systems that keep working when the cast changes, when pressure spikes, and when the story has to survive distance.

This is how you stop hiring “the person who knows” and start building an environment where knowing together is the default.

In Conclusion

Actionable intelligence isn’t discovered. It’s built.

Not through insight, or alignment workshops, or the right people being in the room at the right time (though this can lead to insight). It’s built through shared language, tested assumptions, and controls that keep truth from drifting into vibes once pressure shows up.

That work is structural, and rarely shows up as rhetoric. It doesn’t announce itself as brilliance in a meeting or confidence on a slide. It shows up quietly: fewer arguments about whose numbers are right, fewer decisions that hinge on who happens to be present, and fewer systems that only make sense when a specific person is available to translate them.

Illumination is not knowing more. It’s removing ambiguity as an operating cost.

Organizations fail at this not because they lack smart people, but because they mistake access to expertise for shared understanding. They confuse fluency for alignment, and motion for readiness.

If there’s a single principle worth carrying forward, it’s this: Every statement only holds within its context.

If that context isn’t defined deliberately, it will be inferred. If it’s inferred, it will vary. And if it varies, execution stops being a technical problem and becomes a political one.

You don’t build reliable systems by hiring better translators. You build them by making translation unnecessary. By deciding which truths are load-bearing, which assumptions must survive handoffs, and which definitions are not allowed to change depending on who is speaking.

Illumination isn’t a moment of clarity. It’s an environment where clarity outlives the moment, the meeting, and the people who first understood it.

And if your systems can’t survive that, what you have isn’t intelligence. It’s memory. And memory, eventually, walks out the door.

If the people who “know how it really works” left tomorrow, would your organization still know what is true?