The State That Cannot Contract

Executive Summary. Bangladesh's e-Government Procurement portal is large enough to prove that digitisation is no longer the main procurement question. At noon on 23 August 2026, its ministry view reported 1,122,259 tenders or proposals invited, 122,068 being processed, and 872,015 contracts awarded. The division view reported a smaller denominator but a similar conversion pattern. These are cumulative administrative counts. They do not reveal how many qualified bidders competed, whether the final price was reasonable, whether delivery was late, whether specifications changed, or whether citizens received the asset or service. Bangladesh has digitised the doorway into contracting without yet publishing a joined record of contract performance. The next reform is a public contract ledger that connects planning, tender, award, amendments, payments, milestones, completion, and service results under one identifier. Procurement should be judged by delivered value, not by how many files passed through a portal.

A tender can be perfectly digital and still fail in the physical world. A road contract may be awarded online and delivered late. A medicine order may be processed through a compliant workflow and arrive after the stockout. A school building can have a complete electronic award record while its roof leaks. The portal records the transaction. The citizen experiences the outcome.

That gap matters because procurement is where budgets become real things. A budget allocation is an intention. A contract translates it into a promised price, specification, and date. Delivery determines whether the intention becomes a bridge, textbook, diagnostic service, or functioning drain. If the data chain stops at award, the state can count procedures but cannot manage performance.

Bangladesh digitised procurement at genuine scale

The live e-GP tender report shows the size of the administrative achievement. Its ministry view listed 1,122,259 invitations, 122,068 records being processed, and 872,015 awards at the recorded time. Moving a procurement system of this scale onto a common digital surface reduces search cost, standardises stages, creates timestamps, and gives oversight bodies a base from which to investigate.

Source: Bangladesh e-GP portal, ministry view, record updated 23 August 2026 at noon.

The correct critique starts by acknowledging that achievement. Paper procurement fragments information across offices and filing rooms. A central portal makes notices discoverable and gives suppliers a shared entry point. It can reduce discretionary withholding of tender information. It can also make routine reporting cheaper because fields are already captured during the process.

But scale changes the question. Once more than a million invitations have passed through the system, counting transactions is no longer evidence that the system is learning. The portal should be able to answer which agencies attract competition, which categories produce repeated single bidding, which contracts accumulate amendments, which suppliers deliver late, and which assets fail inspection. If those questions remain outside the public data model, digitisation automates administration without completing accountability.

The archived procurement-performance page reinforces the point. It lists quarterly files from FY 2015-16 through FY 2023-24 Q1. Reports exist, but the public surface is organised as downloadable files rather than a current, queryable history tied to each contract. A regulator may read reports. A spending ministry needs live operational signals.

Two portal views reveal why denominators matter

The portal also displays a division view: 742,277 invitations, 85,918 being processed, and 583,772 awards. These figures are not errors simply because they differ from the ministry totals. They appear to be different administrative slices. The problem is that a citizen cannot infer the relationship between them from the headline table alone.

Source: Bangladesh e-GP portal, division view, record updated 23 August 2026 at noon.

Every performance ratio inherits its denominator. Using the ministry view, awards equal 77.7% of invitations and current processing equals 10.9%. Using the division view, the corresponding shares are 78.6% and 11.6%. The similarity may be reassuring, but it does not make the two populations interchangeable.

Source: Bangladesh e-GP portal; BDPolicyLab calculations from the displayed cumulative counts.

This is not a complaint about labels alone. Procurement dashboards are easily used to imply speed or success. A high award share could mean an efficient pipeline, an old cumulative denominator, cancelled notices excluded from view, or weak competition that makes award easy. Without a data dictionary and stage transitions, the ratio cannot bear a policy claim.

The remedy is simple in concept: publish record-level status with stable definitions. Each tender should show whether it was cancelled, re-tendered, awarded, challenged, amended, completed, or closed. Aggregates should be generated from those statuses, and the portal should state exactly which records enter each denominator. Transparency begins where ambiguity about the population ends.

Award is the midpoint, not the result

An award identifies a selected supplier and a promised contract. It does not prove performance. The contract may change through time extensions, price adjustments, quantity variations, subcontracting, or specification revisions. Payments may move faster or slower than work. A completed physical asset may still fail the service test.

Procurement data therefore need a spine that continues after award. The award identifier should connect to the signed contract, beneficial ownership, implementation schedule, payment ledger, amendments, inspection findings, completion certificate, and any warranty claim. For a service contract, the ledger needs measurable outputs. For goods, it needs receipt and acceptance. For works, it needs milestones and independent verification.

This does not mean publishing confidential commercial details indiscriminately. The public needs enough to evaluate value, competition, and delivery. Sensitive personal or security information can be protected under explicit rules. Blanket opacity after award is not a privacy policy. It is a break in the accountability chain.

The state also needs this joined record for its own purchasing power. If agencies cannot see supplier performance across contracts, a contractor that fails repeatedly can continue bidding with a clean-looking tender file. If price and delivery histories are not comparable, procuring entities negotiate from memory. A performance ledger turns each contract into information for the next one.

Competition must be measured where discretion enters

Electronic submission does not automatically produce competition. Specifications can be tailored, qualification criteria can exclude capable firms, lot sizes can favour incumbents, and short preparation periods can deter bidders. These choices occur before the bid arrives. A portal that counts only invitations and awards sees the endpoints but not the competitive design.

Useful public measures would include valid bids per tender, the share of awards with one qualified bid, bidder concentration by procuring entity and category, time allowed for response, challenge outcomes, and the gap between estimated and awarded price. These indicators should be interpreted with context. A specialised purchase may legitimately attract few bidders. A repeated pattern across ordinary categories needs scrutiny.

Data quality also needs enforcement. Supplier identities must be resolved across spelling variants and related companies. Procuring entities need consistent codes. Amendments need machine-readable reasons rather than free text. Contract values need currencies and tax treatment. Without clean identifiers and definitions, a large portal becomes a large collection of incomparable records.

The aim is not a league table that punishes difficult procurement. It is early warning. An unusual pattern should trigger review, not an automatic accusation. Good analytics narrow the search for human oversight. They do not replace due process.

The counterargument: more disclosure will slow procurement and invite litigation

The strongest objection is practical. Procurement officials already manage complex rules and deadlines. Adding fields, inspections, and public reporting could delay urgent purchases. Suppliers may challenge every anomaly, while officials become reluctant to exercise legitimate judgement. A perfect data model could make the state slower.

That risk is real if reporting is designed as a parallel paperwork system. It is much smaller if performance data are produced by the workflow itself. A payment already has an authorisation. An extension already has an approval and reason. An inspection already produces a decision. The task is to connect those events to the contract identifier and publish appropriate fields, not to write a second narrative report.

Emergency procurement also needs speed, but speed increases the need for later disclosure. An emergency pathway can simplify competition while requiring rapid publication of supplier, price, justification, delivery, and review. The choice is not speed or accountability. It is whether accountability happens before, during, or immediately after the exceptional purchase.

Litigation is not avoided by weak records. Ambiguous specifications, missing reasons, and disconnected files create disputes. A clear event trail can protect honest officials by showing what they knew and why they acted.

Three moves from e-procurement to contract intelligence

1. Create one public identifier from plan to completion. Owner: Bangladesh Public Procurement Authority. Success signal: every tender page links to award, signed terms, amendments, payments, milestones, completion, and complaint outcomes through a documented data schema.
2. Publish competition and delivery measures, not only volumes. Owner: BPPA with the Implementation Monitoring and Evaluation Division. Success signal: procuring entities can compare valid bids, single-bid patterns, time extensions, cost changes, completion delays, and inspection outcomes by category and supplier.
3. Use risk signals for review while preserving due process. Owner: procuring entities and oversight bodies. Success signal: unusual patterns generate documented review queues, decisions are recorded, and no supplier or official is sanctioned by an unexplained algorithmic score.

Bangladesh has already paid the institutional cost of moving procurement online. The possibility now is to make the data follow the contract into the world. A state can count awards and still be unable to contract. It becomes capable when it can specify clearly, buy competitively, monitor delivery, learn from failure, and show citizens what arrived.

---

Data note: portal counts are cumulative administrative totals captured at the stated update time. They do not measure competition, timeliness, savings, quality, or completion. Ministry and division views are separate displayed denominators. Sources retrieved 2026-08-23.

Sources

1. Bangladesh e-GP portal, Tender report: https://www.eprocure.gov.bd/Tenders.jsp
2. Bangladesh e-GP portal, Procurement Performance Indicators reports: https://www.eprocure.gov.bd/ReportDownload.jsp
3. Bangladesh Public Procurement Authority, Project progress: https://bppa.gov.bd/reports/details/reports-project-progress/38.html