# A road can be 100 percent built and 17.6 percent public.

The CoST toolkit is not five separate products. It is one chain of public trust, and a chain is only as strong as its weakest link. Each link answers one public question. Pull any link out and the project stays visible but never becomes followable.

By Michael Cengkuru · 29 Jun 2026 · Interactive version: https://cengkuru.com/essays/when-a-project-becomes-public/

## The story

Public infrastructure is built in the open. Its truth is filed somewhere else. The concrete is visible, the machines are working, a surveyor sights the line. But the sign on the barrier says only "District road, 14 km". The contractor is unreadable, the contract value is not stated, the completion date is not stated, and no one is named to ask. A sign on a barrier is not a project you can follow. Uganda's portal publishes 17.6 percent of what a followable project record needs. A project can be visible in concrete and still hidden in truth.

To understand the CoST toolkit, follow five public questions. The tools are often introduced as separate products. They are better understood as one chain: each link answers one question, and a chain is only as strong as its weakest link.

**Link 1: disclose (IDS).** What facts make the project followable? IDS gives the project enough facts to be followed: 40 proactive data points published as a matter of course across identification, preparation, procurement, delivery and completion, not just the moment a contract is awarded, plus 27 reactive disclosure items released on request, the backstop when proactive publication stops short. Without these facts, citizens cannot ask precise questions. They can only suspect.

**Link 2: structure (OC4IDS).** How do those facts keep the same meaning across systems? IDS names the facts. OC4IDS gives each fact a stable place: a common structure, running project to contracts to changes to progress, so a portal, an analyst, a journalist and an oversight body all read one record the same way. OC4IDS is not a trophy checklist. Structure does not prove quality. It makes quality testable. A record can pass schema validation and still be thin. That is exactly why the next link exists.

**Link 3: review (Independent Review).** Do the published claims survive checking? Disclosure is a claim. An independent reviewer compares the published claim against the documents and the ground, then turns technical data into findings a non-specialist can use. The essay walks an illustrative trace, not a real project finding: the record claims 14 km sealed with no variations recorded; review finds the value reconciles with payments and the documents are complete, but only 11 km sealed on the ground and one variation order found, unrecorded. The check is physical: a reviewer runs the tape along the road and marks where record and ground disagree. Those two flags, a short delivery and an unrecorded change, are exactly what review exists to surface and explain in plain language. Trust begins when the record survives contact with the ground.

**Link 4: measure (ITI).** Is the system becoming more transparent over time? The Infrastructure Transparency Index does not ask whether one project is visible. It scores the whole system from 0 to 100 across four dimensions and 93 indicators: enabling environment (12), capacities and processes (25), citizen participation (12), and data publication (44). No country score is asserted in the essay; the scale is left empty on purpose. A score read once is a ranking. A score read again is a direction.

**Link 5: act (social accountability).** Who uses the evidence to demand better decisions? Data only matters when someone uses it, and the question can be real. In Uganda's GPP infrastructure dashboard, Competition view, as read in June 2026: 28.5 percent of competitive infrastructure tenders attracted only one bidder, 155 of 543 tenders in the snapshot, with 3.6 average bidders across 393 projects. A single-bid rate is a risk flag, not proof of wrongdoing. It tells oversight where to look first. The number then moves through real roles: a journalist flags it as a story, a monitoring group asks whether competition barriers explain it, the multi-stakeholder group requests a written follow-up, the procuring entity explains the figure or improves the next procurement, and back at the barrier a resident can ask a precise question, not a vague suspicion. Separately, prior assessment found the same portal meets 17.6 percent of OC4IDS requirements, 26 of 148 data elements, and 4.2 percent of project-level disclosure paths (CoST Uganda, 2022).

Now pull a link out. Remove IDS and the project has no minimum public facts: people see the road but cannot follow the decision. Remove OC4IDS and the facts exist but cannot travel cleanly across systems. Remove Independent Review and disclosure remains a claim; trust has not been tested. Remove ITI and individual projects may improve, but the system has no visible direction. Remove social accountability and evidence exists, but no one uses it to change decisions. One chain, not five products. A project becomes public when people can follow it, read it, test it, measure progress, and use the evidence.

## The data

Uganda GPP snapshot, as the essay's source drawer states it:

| Figure | Value | Date / source |
| --- | --- | --- |
| Single-bid rate, competitive infrastructure tenders | 28.5% (155 of 543 tenders) | Snapshot June 2026, Uganda GPP infrastructure dashboard, Competition view (PPDA) |
| Tenders in the snapshot | 543 | Same snapshot |
| Average bidders | 3.6 | Same snapshot |
| Projects in the snapshot | 393 | Same snapshot |
| Total value in the snapshot | UGX 4.0T | Same snapshot |
| Portal OC4IDS coverage | 17.6% of requirements (26 of 148 data elements) | CoST Uganda, 2022 |
| Project-level disclosure paths | 4.2% | CoST Uganda, 2022 |

ITI framework structure (the ITI manual's indicator split; the essay asserts no country score, the gauge shows direction, not rank):

| Dimension | Indicators | Share of 93 |
| --- | --- | --- |
| Enabling environment | 12 | 13% |
| Capacities and processes | 25 | 27% |
| Citizen participation | 12 | 13% |
| Data publication | 44 | 47% |

The 14 km claimed versus 11 km found trace in Link 3 is illustrative, not a real project finding. The only hard figures stated as fact are the CoST framing counts (40 proactive, 27 reactive, 93 indicators and the dimension split) and the Uganda GPP figures above.

## Sources

- Uganda GPP infrastructure dashboard, Competition view (PPDA): https://gpp.ppda.go.ug/open-data?tab=infrastructure&section=dashboard. Snapshot read June 2026: 543 tenders, 28.5% single-bid, 3.6 average bidders, 393 projects, UGX 4.0T.
- CoST Uganda (2022): OC4IDS coverage assessment of the GPP portal. 17.6% of OC4IDS requirements met (26 of 148 data elements); 4.2% of project-level disclosure paths.
- CoST framing: IDS (40 proactive data points, 27 reactive disclosure items), OC4IDS (common structure for infrastructure project and contracting data), Independent Review (the current term in CoST's Data Publication Manual for what was previously called Assurance), and the Infrastructure Transparency Index (0 to 100, four dimensions, 93 indicators per the ITI manual).
- Recomputation note: the GPP figures are a fixed snapshot of the live dashboard as read in June 2026. The dashboard recomputes, so a later read may differ. The snapshot date and view are stated so the essay does not silently depend on changing live data; that standing recomputation is the point, not a flaw.