The dashboard was green.
The steering committee moved to the next agenda item because the update sounded controlled. Schedule, budget, scope, and risk all carried the same signal. The deck was clean. The dependency map looked orderly. The action log had owners. Nobody asked whether the evidence still supported the color.
That is not a failure of attention. It’s how governance is supposed to work when a report says the project is fine: you believe the report, and you move on to the next item.
The problem is that a status report and a fact are not the same category of thing.
A green status is a claim. It says: current evidence, interpreted by the people closest to the work, supports continuing on the current path. That’s a legitimate thing to claim. It is also, structurally, just a claim — waiting to be tested against a different kind of evidence than the one that produced it.
Most enterprises don’t have a mechanism for that test. They have a mechanism for producing the claim.
Truth Latency
I call the gap between those two things Truth Latency: the delay between when operational reality changes and when leadership sees enough evidence to act on it.
It is not the same as lying. Nobody has to falsify anything for Truth Latency to exist. A project can be staffed by competent, honest people at every layer and still generate a status report that no longer matches what’s actually happening underneath it — because the report was never built to catch the moment reality shifted. It was built to summarize confidence.
Projects rarely fail overnight. They decay. A dependency that was supposed to close this cycle quietly reopens under a different name. A vendor’s language shifts from “will deliver” to “targeting” to “evaluating options,” a little softer each cycle, without anyone deciding to soften it. A technical demo that was promised for last month gets rescheduled once, then again, absorbed into the normal noise of a busy program.
None of that shows up as a red status. Each individual softening looks like ordinary variance. It’s only the pattern, held up against what was true a few cycles ago, that shows the drift.
By the time the official record admits what changed, the operational reality has usually already moved on without it.
How the Report Loses the Signal
The mechanism is not mysterious. It’s structural.
Technical reality gets filtered as it moves upward through a reporting hierarchy, and the filtering is rarely a single decision — it’s the compounding effect of many reasonable ones. An engineer flags a real integration risk to a team lead. The team lead, who has five other risks to report and thirty minutes to do it, folds it into a broader “on track, monitoring closely” line. The program manager, compressing six team updates into one steering slide, keeps the language but drops the specificity. By the time it reaches the executive who’s deciding whether to release the next tranche of funding, the risk is a single bullet that reads as routine.
Nobody lied. Everyone made a locally reasonable compression decision. The signal still didn’t survive the trip.
This is what I call Human Middleware — skilled people spending their judgment translating fragmented reality into a form the next layer up can consume quickly, instead of exercising judgment on the substance. It’s not a character flaw in the people doing it. It’s what happens when an enterprise has no other mechanism for reconciling operational evidence with executive attention, so it makes people do the reconciling by hand, one compression at a time.
Activity Is Not Evidence
The other half of the mechanism is subtler: activity keeps generating the appearance of progress long after actual movement has stopped.
Meetings continue. Invoices continue. The steering deck still moves through its template on schedule. The vendor still shows up with an update. None of that is fake — the meetings really happen, the invoices are for real work. But activity and evidence are different claims. Activity says people are doing things. Evidence says the thing being funded is actually getting closer to done.
The test isn’t whether the team is busy. It’s whether the last funding period produced evidence — a working demo, an accepted deliverable, a closed dependency, a resolved technical question — that the underlying investment thesis is still valid. When the answer is no, and the funding continues anyway, that’s not yet a crisis. It’s the early, quiet shape of one.
Where Delay Becomes Capital Risk
Truth Latency describes delayed awareness. There’s a related, sharper problem: Decision Latency — delayed action, even after awareness catches up.
A signal can surface, get logged, get discussed, and still not force a decision, because nothing in the governance structure requires one. The dependency stays open across three reporting cycles under three slightly different labels. Everyone in the room has technically seen it. No one has been required to decide anything about it.
That gap consumes optionality even before anyone declares a formal problem. Every cycle the decision gets deferred, alternatives narrow, commercial leverage shifts, and the cost of course-correcting grows — quietly, without a single red flag ever appearing on a dashboard.
This is where delayed truth stops being a reporting inconvenience and becomes a capital question: is the enterprise still funding a belief the current evidence supports, or is it funding the belief it approved eight months ago, on inertia?
Two Signals Worth Learning to Read
Two patterns are worth watching for, specifically because they show up before a status ever turns red.
Silence can be a signal. Not the absence of noise — the absence of a topic that should be there. No mention of user-acceptance testing as go-live approaches. No update on a dependency that used to get discussed every week and has simply stopped coming up. Silence isn’t proof of a problem. It’s a prompt to ask why a topic that should be active has gone quiet.
Linguistic decay is the other one. Commitments move from high-agency language to low-agency language before the formal status changes: “will deliver” becomes “targeting,” “confirmed” becomes “should be,” “fixed” becomes “we’re optimizing.” That drift is not mind-reading and it isn’t proof of deception — people hedge for all kinds of reasonable reasons. But it’s an early-warning pattern worth testing against the evidence underneath it, rather than dismissing as just how people talk.
Neither signal is a verdict. Both are worth a look before the quarter’s funding decision gets made on the strength of a deck that hasn’t changed its language in months for reasons no one has actually checked.
The Governance Move
None of this argues for distrust of every report, or for burying leadership in more granular status data. More reporting created this problem; it won’t solve it.
The actual move is smaller and more specific: before approving the next funding period, ask what evidence — not what activity — supports continuing. Compare this cycle’s steering deck to last cycle’s, looking for language that’s softened rather than sharpened. Treat a dependency that survives three reporting cycles under three different names as a decision point, not a status line. Ask, plainly, what changed in the work, not what changed in the wording.
That’s a governance discipline, not a technology purchase, and it doesn’t require distrusting the people producing the report. It requires distinguishing the report from the fact.
The project didn’t become dangerous the day the dashboard turned red. It became dangerous the day the evidence turned red and capital kept moving anyway — while the dashboard, for entirely ordinary reasons, hadn’t caught up yet.
The question was never whether the project is green.
The question is what evidence still supports green.
This is the first essay in Truth Arrives Late, a series working through the evidence discipline behind Forensic Capital Governance*, my book on evidence-based capital stewardship, arriving early September. More at forensicgovernance.com.*