# The Portal Is Online. Diagnose Why It's Dying.

A disclosure portal stays alive only if five connections to the real workflow are made; the pipeline, not the portal, is the product.

By Michael Cengkuru · Interactive version: https://cengkuru.com/posts/why-disclosure-portals-die/

## The story

A disclosure portal is not a website. It is a public window on an internal workflow. That is why a portal can be online and still be dying: the homepage can look healthy, with clean cards, a contract value, a progress bar, while the body beneath is failing. Online is not the same as alive. Portals do not die on the homepage. They die inside the workflow.

The interactive version treats the portal as a patient in intensive care, a body with five organs. Each organ is one connection to the real workflow, and each can be in one of three states: missing, manual, or connected. Missing means the connection was never made. Manual means people prop it up by hand, which works until the effort or the money ends. Connected means the portal receives what it needs as a byproduct of the work itself.

The bloodstream is automated data extraction from source systems. Missing, no data moves; the portal shows whatever was typed in once, long ago. Manual, officers must log in and upload spreadsheets by hand, and it works until the funding ends. Connected, data flows from the e-GP system and IFMIS as a byproduct of doing the work. When this organ fails, the pipeline clogs in spreadsheets, email and forgotten uploads, and freshness dies.

The spine is the procurement-to-delivery linkage, made with OC4IDS. Missing, the record stops at contract award; there is no line to the building site. Manual, someone occasionally writes a progress note, unlinked to the contract. Connected, OC4IDS links the contract to the project, its delivery and its completion. When the spine fails, the public gets the receipt but not the roof: an award without delivery.

The immune system is independent review, the CoST Independent Review. Missing, no one checks the claims against reality, and optimistic figures face no consequence. Manual, an occasional report arrives too late and too rarely to challenge a live claim. Connected, an independent reviewer flags gaps between what was promised and what was delivered. Without it, a record can say a project is nearly complete, good quality, no delays, while the site shows no roof, no drainage, no crew.

The muscles are real user tasks: building for the practitioner, not the press release. Missing, the portal is a carousel, PDFs and a press-release archive, with no task anyone can finish. Manual, there is a search box, but it returns documents, not answers. Connected, a district engineer, a CSO monitor and a journalist each have a job the portal does. A portal without muscles is impressive at launch and useless on Tuesday; no one returns because no one had a task.

The metabolism is sustained operational funding: funding the pipeline as architecture, not a project. Missing, there is a launch event, a consultant, branding, one-off training, then nothing. Manual, a short grant keeps the lights on, with a hard end date. Connected, the data pipeline is maintained like the road network it describes. Launch money is a sugar rush; when the grant closes, the body starves.

The prognosis logic is simple and stated openly. Each organ carries a weight expressing how much it protects the portal over three years. A connected organ contributes its full weight, a manual one contributes half, a missing one contributes nothing, and no portal counts as healthy while any organ is entirely missing. The combinations produce named end-states: a press-release portal (nothing connected: you funded visibility, not survival), open-data theatre (data flows but no delivery link: you disclosed the receipt, not the infrastructure), a dashboard with no immune system (claims without accountability), technically alive but socially unused (the data flows, nobody returns), underfunded and slowly starving (connected today, abandoned by year three), or a living portal, where all five are connected.

The evidence behind this is scoped, not universal. Of the five implementations I watched, the ones that launched to a press release went dark within three years. The ones that lived shared one trait: the data was a byproduct of the work, not an extra task. The five decisions are not features to add later. They are the architecture, chosen at the start. Skip any one and the portal can still launch beautifully. It simply will not survive contact with the workflow it was meant to open.

One honesty note on the numbers you will see in the interactive vitals: the "498 days", "UGX 2.4B" and "82% complete" figures are illustrative of the failure modes described, not specific to a named project. The weights are a modeling device to drive the simulation, not measured survival probabilities, and the three-year curve illustrates causality rather than predicting a specific outcome.

The verdict the whole piece earns: the portal is not the product. The pipeline is the product. Transparency cannot be an extra task bolted onto busy people. It has to become a byproduct of the work itself.

## The data

The five organs, as the interactive model states them. Weights are out of 100 and are modeled relative importance, not survival probabilities.

| Organ (role) | Decision | Weight (of 100) | Failure if missing |
| --- | --- | --- | --- |
| Bloodstream (data extraction) | Automate data extraction from source systems | 30 | The pipeline clogs in spreadsheets, email and forgotten uploads. Freshness dies. |
| Spine (procurement to delivery) | Link procurement to delivery with OC4IDS | 22 | The public gets the receipt (UGX 2.4B awarded) but not the roof. Award without delivery. |
| Immune system (independent review) | Run the CoST Independent Review | 20 | Record says 82% complete, good quality, no delays. Site shows no roof, no drainage, no crew. |
| Muscles (real user tasks) | Build for the practitioner, not the press release | 16 | Impressive at launch, useless on Tuesday. No one returns because no one had a task. |
| Metabolism (operational funding) | Fund the pipeline as architecture, not a project | 12 | Launch money is a sugar rush. When the grant closes, the body starves. |

The ICU vitals, one per organ. The figures are illustrative of the failure modes described, not specific to a named project.

| Vital | When the organ is failing | When the organ is connected |
| --- | --- | --- |
| Data freshness | Last updated 498 days ago | Updated this week |
| Delivery evidence | Missing (contract value only) | Linked to site progress |
| Record vs reality | Unchecked | Independently reviewed |
| Public usefulness | Low (nothing to do here) | Real tasks completed |
| Operational funding | Grant closed, no owner | Maintained, named owner |

## Sources

- The five survival decisions are taken verbatim from the source analysis of five disclosure-portal implementations across Uganda and East Africa, as reported by the author. The "went dark within three years" claim is scoped to that observed sample, not a universal.
- OC4IDS is the project-level standard that links a contract to its delivery and completion; where contract-level detail is referenced it lives in the linked OCDS release.
- The CoST Independent Review (assurance) is a real mechanism.
- The interactive version offers the underlying model table as JSON, CSV and Excel downloads (export stem: portal-icu-survival-model) from its "Methodology and data" drawer.
- Interactive version: https://cengkuru.com/posts/why-disclosure-portals-die/