Organizational dark matter

Organizations can measure well and decide badly when their model leaves out dynamics that are already affecting coordination, trust, and capability.

The obvious interpretation is poor individual performance. From there, the discussion moves to bonuses, account allocation, promotion, or dismissal.

A closer look at the operation reveals another layer. This salesperson helps new hires get up to speed. They step in when a technical objection stalls someone else’s sale. They maintain relationships with customers who are not buying today but return months later. The CRM records some of those interactions, but it does not attribute their effect on other people’s work.

If this person leaves, their individual quota disappears from the report. So does part of the collective capability that no one had modeled.

The data did not fail; a variable was missing

In 1933, Fritz Zwicky studied the movement of galaxies in the Coma Cluster. The visible mass did not seem sufficient to explain why those galaxies remained together. They were moving too fast for the gravity attributable to the matter that could be observed.

Zwicky proposed that unseen matter was contributing the missing mass. He called it dunkle Materie: dark matter.

Decades later, Vera Rubin and Kent Ford measured the rotation of spiral galaxies. Their outer regions were also rotating at speeds that did not fit the distribution of visible matter. Their work did not identify what dark matter was, but it made a model based only on what emitted light much harder to defend.

The advance was not simply a more precise reading. It required acknowledging that the representation of the system was incomplete.

The comparison with an organization is an analogy, not a scientific equivalence. Even so, it allows us to ask a useful question:

What forces are producing consistent effects without appearing in the model we use to make decisions?

What the organizational chart and dashboard compress

Financial and operational results are essential measures. Sales, margin, productivity, turnover, and attainment allow us to recognize meaningful changes. The error arises when they are treated as a sufficient explanation of how the result was produced.

Beneath those indicators are relationships that rarely fit into a single number:

• who connects areas that do not normally coordinate;
• where the knowledge needed to unblock exceptions is concentrated;
• which leaders create openness and which produce silence;
• who absorbs invisible work to keep the team functioning;
• which dependencies make an apparently stable result fragile.

Research on organizational networks has shown that a substantial part of coordination happens through informal relationships, not only through the channels defined by the organizational chart. Other studies of knowledge teams distinguish between having expertise available and getting that expertise to the place where it is needed.

These relationships are neither mystical nor impossible to study. Nor should they automatically become another score. They are contextual variables whose effects can be observed before we attempt to measure them with false precision.

When the indicator changes, the cause has already been operating for some time

A results dashboard looks mainly at what has already happened. It can include leading signals, but only when someone has decided which mechanism they want to observe and built evidence around it.

Without that layer, management reacts late.

Early turnover appears after weeks of friction. A decline in sales comes after conversations, referrals, or technical coordination have been lost. A project delay becomes visible after several dependencies have already accumulated. The resignation of a critical person reveals a support network that never appeared as a formal responsibility.

In each case, the result is real. What is incomplete is the causal story built around it.

That is why calling the salesperson in the example an underperformer may be a valid classification of their individual quota and a poor interpretation of their total contribution. The problem is not solved by eliminating the quota. It is solved by preventing a single metric from claiming to describe all the value produced.

Do not measure everything: read consistent effects

The usual reaction to a missing variable is to create another survey, another indicator, or a new dashboard section. That impulse can increase the burden and create an appearance of control greater than the understanding gained.

Signals of energy and voice appear in meetings where no one challenges a decision, but no one builds on it either; in good news that prompts neither questions nor initiative; and in known problems discussed only outside the forum where they could be resolved.

Signals of coordination appear in deliverables that formally meet requirements but need repeated corrections; in the people who always step in when an exception arises, even though their role does not require it; and in teams that depend on particular intermediaries to speak to one another.

Signals of strain include early turnover concentrated in the same unit, workload redistributions that become permanent, and reliable people who stop participating before their results decline.

An isolated signal can have many explanations. A repeated pattern, observed from several perspectives and connected to concrete consequences, deserves a place in the management conversation.

It does not replace KPIs. It changes how they are interpreted.

Leadership stops correcting results and starts protecting capabilities

When management focuses on final indicators, intervention comes after the decline: adjust targets, replace people, reallocate accounts, or demand more monitoring.

A more complete reading moves intervention toward the dynamics that produce the result. Sometimes this means a conversation that prevents friction from becoming a rupture. In other cases, it means protecting someone who performs a critical function without formal recognition, distributing knowledge, or reducing a dependency that looks efficient until it fails.

This also changes the role of human resources. HR can design instruments, facilitate network analysis, establish ethical boundaries, and help distinguish evidence from impression. Day-to-day observation still belongs within operations. An annual process cannot replace that responsibility.

Three practices help without turning work into a collection of metrics:

• review friction and dependencies, not only progress and attainment;
• reconstruct which people and relationships made a result possible;
• record patterns with context, date, and effect, avoiding permanent labels.

The aim is not to discover an invisible truth about every person. It is to broaden the model before making a consequential decision.

A dashboard explains only what the model decided to include

Dashboards can anticipate events when they contain good leading indicators. They can also help detect anomalies and open questions. Their limit appears when a result is interpreted as a cause, a capability, or a complete contribution.

Zwicky did not resolve the anomaly by adding more galaxies to a table. Rubin did not make dark matter visible. Both bodies of work helped show that the observed behavior required the model to be reconsidered.

Something similar happens in organizations. When a result changes, the causes have often been operating for some time within relationships, decisions, and dependencies that remain outside management’s main field of view.