The missing-information theory of failure
When a public project goes wrong and somebody asks why nobody caught it, the answer almost always arrives in the same shape: the information was missing. The data was not published. The record was stale. The format was unusable. The portal was down.
That diagnosis is attractive because it is tractable. Missing information has an owner, a budget line, a vendor and a delivery date. You can procure your way out of it. So we do: another portal, another standard, another integration, another dashboard.
It is also often correct. Plenty of failures really are information failures, and fixing them really does help. That is the case for disclosure infrastructure, and it is a good case.
But it is not the whole account. Treating it as the whole account is how a well-built portal ends up sitting on top of a system that keeps failing in the same way it always did.
A thought experiment: build the portal perfectly
What follows is an invented proportion, not a measurement, and it is worth saying that before the numbers appear. Nothing here was measured, and there is no dataset behind it.
Imagine a product or programme team with six build levers in front of it, each one a real, fundable line item:
| Build lever | What it buys |
|---|---|
| Field completeness | Every field in the standard actually populated |
| Record freshness | Same-day updates instead of annual dumps |
| Machine readability | Clean structured data, not scanned PDFs |
| Uptime and speed | The portal is up when someone looks |
| Standard compliance | The record is comparable with other publishers |
| Search and usability | A non-specialist can find a project |
Push all six to maximum and the portal reads Flawless. Now picture the accountability outcome, meaning whether a problem that exists gets found, named and acted on by someone who can change it, moving from 2 to 20 out of 100. Those two endpoints are invented to make the size of the gap visible. They are an argument about proportion, not a coefficient.
Twenty is not nothing. It is the difference between a problem being unknowable and a problem being knowable, and every serious accountability story starts from a record someone could get to.
But knowable is not found, and found is not fixed. And if you think architecture buys 35 rather than 20, the shape of the claim survives: there is a ceiling, and the levers above it are not on the panel you control.
- FindA usable record makes a problem knowable to someone who can inspect it.
- ReportA person names the problem and brings it into an accountability process.
- RespondA decision-maker with authority acts on the evidence.
- ConsequenceThe institution and the person who surfaced the problem need a reason to keep doing so.
On the numbers. To repeat, because it is the claim most likely to be misread: the 2-to-20 range is an invented illustration of proportion, not a measured coefficient.
The six levers you cannot procure
Each of these survives a perfect portal untouched. None is an information problem.
| # | The lever | Why the portal does not touch it | What would actually move it | Who owns that |
|---|---|---|---|---|
| 1 | Officials have no incentive to expose problems | The person best placed to spot the fault is the person whose unit gets blamed for it. Publishing is a cost paid by them, a benefit collected by someone else. | Finding a problem early is what gets you promoted | The ministry's performance and appraisal rules |
| 2 | Decision-makers lack authority to act | The officer reading the red flag cannot suspend the payment or remove the contractor. The person who can is three levels up, reading something else. | A named role can halt a payment on the evidence | Whoever writes the delegation of authority |
| 3 | Agencies protect jurisdiction | Roads authority sees the works, finance ministry sees the money, audit office sees it a year later. Each is right within its mandate; none owns the whole picture. | One body answerable for the joined-up view | Cabinet, or the statute that split the mandates |
| 4 | Political priorities outrank evidence | The record says the project is failing. It is also the project that was announced and ribbon-cut before an election. Evidence does not lose the argument; it never gets to have it. | Cancelling a bad project costs less than defending it | The political principal, not the programme |
| 5 | Budgets end when the donor project ends | The portal, the pipeline, the analyst and the unit that chased findings were all on the same three-year grant. The system does not fail on the last day; it goes quietly stale. | The recurrent budget carries the running cost before the grant closes | The finance ministry, at the next budget cycle |
| 6 | Nobody is punished when the system goes stale | Publication stops and nothing happens. No hearing, no finding, no consequence. The absence of a penalty for silence is itself a rule. | Failure to publish is an audit finding with a named officer | The supreme audit institution and the regulator |
Lever 4 usually decides the others. Where political priority is settled against the evidence, incentives, authority and penalties all tend to arrange themselves behind it.
One case where the problem was found
A thesis that says "nobody is rewarded" needs a case that tests it, and the honest case is one where the system did respond, just not through the channel the thesis is about.
In Slovakia, Vahostav-SK, the country's largest construction firm, held several large state highway contracts and went into bankruptcy in 2015, its restructuring risking hundreds of subcontractors left largely unpaid, with a police investigation. Its ownership ran through shell companies in New Zealand and Cyprus to five private persons in Costa Rica. The public record had named the company and hidden the people behind it.
The problem was found. It was acted on. But the pressure did not come from a disclosure portal that someone monitored on a Tuesday. It came from a visible failure with a visible set of victims, a scandal that carried political cost, and a reform that followed: Slovakia's Register of Public Sector Partners, in November 2015, requiring beneficial owners to be disclosed before public contracts. The incentive to act was an external shock, not routine institutional demand.
That bounds the thesis rather than refuting it. Problems do get found and fixed, and often the trigger is exactly the thing this essay says is missing: someone bearing a cost. The portal can make the problem knowable earlier. It cannot supply the shock.
Why the fix is always another portal
Look at the six levers and notice what they share. Every one requires someone powerful to accept a cost: a ministry appraised on bad news, a principal losing a project they announced, a finance ministry taking on a recurrent line, an audit office naming an officer.
The build levers require none of that. They need a budget, a vendor and eighteen months.
That asymmetry is the whole explanation for the pattern. We keep fixing the tractable thing because it is the one nobody has to lose an argument over.
This is not a case against disclosure infrastructure. I build it, and the 20 is worth having. It is a case against the substitution: shipping architecture and reporting it as accountability, then being surprised when the same failures repeat on a better platform.
What to do with this
Before the next disclosure system is scoped, ask who is rewarded for finding a problem in it. Not who can access the data: who benefits, personally and institutionally, from surfacing something bad.
If the answer is nobody, you already know what the system will produce, however well it is built.
Ask the question again at the mid-term review, when the grant that pays for the analyst is halfway gone.
Evidence register
Sources and scope
This essay is an argument, not a measurement. The six institutional levers are a taxonomy drawn from recurring failure patterns in infrastructure disclosure programmes. They are not derived from a dataset, and no figure here reports a specific country, agency or project. The 2-to-20 illustration is an invented proportion, and nothing in it should be quoted as a finding.
The one real case used to bound the argument is cited: Open Ownership, case study: Vahostav-SK, in Early impacts of public beneficial ownership registers: Slovakia, 1 October 2020: https://www.openownership.org/en/publications/early-impacts-of-public-beneficial-ownership-registers-slovakia/case-study-vahostav-sk/
The narrower factual claim underneath this one, that disclosure makes a project testable without making it good, is developed with real records in the companion essays:
- The door is not the verdict: https://cengkuru.com/essays/the-door-is-not-the-verdict/
- Who decides what better means when AI ranks public projects?: https://cengkuru.com/essays/the-moat-is-the-decision-layer/
Start a conversation
If this piece names a decision you own, tell me what you are hoping to achieve and where I could help.
Start an enquiry →