I. A reform inaugurated before it was evaluated
On 15 December 2024, Pakistan Customs switched off a century-old habit. Under Customs General Order No. 6 of 2024, every import Goods Declaration filed at the Appraisement Collectorates of Karachi began flowing to a Central Appraising Unit at South Asia Pakistan Terminal, where an anonymous appraising officer — who could not see who the importer was, and whom the importer could not see either — would determine value, classification and liability. Three weeks later the Prime Minister flew to Karachi to inaugurate what was presented as a decisive blow against customs corruption, with clearance time officially claimed at nineteen hours and complaints officially claimed to have fallen by eighty per cent (both government figures announced at inauguration, neither independently verified then or since).
The sequencing deserves attention: the system was celebrated in January 2025, reviewed internally in June, and audited in August. By the time evidence arrived, the political cost of admitting failure had already been paid forward. That inversion — announcement first, measurement later — is the original defect from which most of the system’s subsequent problems descend, and it explains why the Faceless Customs Assessment (FCA) has proved so difficult to correct even by those inside the organisation who know precisely what is wrong with it.
II. The design error at the centre
The FCA rests on two propositions. First, that assessment quality suffers when the assessing officer knows the trader, because knowledge enables collusion. Second, that specialised assessment groups — chemicals, machinery, vehicles, textiles — are vehicles of capture and should be dissolved into a single undifferentiated pool.
Both propositions had been tried before. FBR’s own Review Committee on FCA, reporting in mid-2025, noted that concealing trader particulars from assessing officers and abolishing specialised groups were features of Pakistan Customs’ first computerised system roughly twenty years earlier, and that both were abandoned for reasons that remain valid. Specialisation, the committee found, builds sector-specific institutional memory, strengthens control, and — precisely because the officer has handled a hundred similar consignments — shortens clearance rather than lengthening it. Anonymity destroys that memory. It does not merely remove the corrupt appraiser’s opportunity; it removes the competent appraiser’s expertise.
The result is an assessing officer who has never seen the goods, does not know the sector, cannot see the importer’s compliance history, and works against a productivity clock under an incentive-based performance mechanism. Ask what such an officer does when confronted with an unfamiliar HS code and an invoice that looks marginally low. The rational response is not careful inquiry. It is to reach for the highest defensible value in the database, book the assessment, and move to the next file. Over-assessment is professionally costless to the officer; under-assessment is what audits look for. The system’s incentive structure quietly manufactures a bias, and the bias falls on whoever happens to be next in the queue.
III. The asymmetry: why errors do not cancel out
This is the analytical heart of the matter, and it is routinely missed in the public debate.
A random assessment system produces errors in both directions. If those errors were symmetric in consequence, they would average out and the system would be merely noisy rather than harmful. They are not symmetric, for a simple reason: the two kinds of error are corrected by different parties at radically different costs.
When the assessment falls below the correct liability, nobody complains. The importer pays, takes delivery, and the container leaves the port. Correction, if it comes at all, must come from the state — through post-clearance audit, contravention proceedings, show-cause notices and years of litigation, against a trader who has already sold the goods. In practice, the state rarely recovers.
When the assessment lands above the correct liability, the entire burden of correction falls on the importer. He files a review, waits, follows up, pays demurrage and detention while the container sits, borrows against a consignment he cannot sell, and — under an anonymous system — cannot even identify who made the decision he is contesting. Correction here is not free; it is priced in weeks and in dollars per day.
The consequence is a one-way ratchet. Every under-assessment becomes permanent revenue loss. Every over-assessment becomes a private tax on the importer’s working capital, paid whether or not the review eventually succeeds. The same pool of assessment errors thus simultaneously drains the exchequer and punishes the trader — exactly the pattern the field reports describe, and exactly the pattern that a “faceless” design without a fast, independent remedy is guaranteed to produce.
The reported data bear this out. FBR’s Review Committee found that reviews filed against assessments rose from about 6% of declarations to roughly 14% after the FCA — a more than doubling of contested assessments, which the committee read as evidence that assessment quality had deteriorated. The FBR Chairman conceded the point before the National Assembly Standing Committee on Finance, stating that the rise in reviews had not been anticipated at the design stage. The same committee reported that container clearance time rose by 57% — the opposite of the claim on which the system was launched.
IV. The revenue loss to the exchequer (reported figures)
The undercount side of the ratchet has been quantified by the state’s own auditors, and the numbers are not small.
The internal Pakistan Customs Audit covering 16 December 2024 to 15 March 2025 — a 161-page report which, by its own admission, did not cover the full population of clearances — examined 13,140 goods declarations and found discrepancies in 2,530. Its aggregate figure for the quarter, as reported by Dawn, was approximately Rs 100 billion, composed of:
Component (all reported by the audit) Amount
Detected duty and tax evasion (1,524 GDs) Rs 5.007 bn
Fines and penalties left uncollected Rs 2.43 bn
Restricted/banned goods cleared in breach of the Import Policy Order (1,000+ GDs) Rs 10.54 bn (goods value, not revenue)
Potential loss from cases never formally framed Rs 30.36 bn (potential, not realised)
Intellectual honesty requires a caveat that FBR’s critics usually skip: the Rs 100 billion headline is a composite, not Rs 100 billion of cash foregone. The restricted-goods component is the value of the goods, not the duty on them, and the largest single component is potential loss. But honesty cuts both ways. The hard components alone — detected evasion plus uncollected penalties — total a reported Rs 7.4 billion in one quarter, from a partial sample. A simple annualisation of that sampled subset (an estimate, and a floor rather than a ceiling) exceeds Rs 30 billion a year; the full-population figure is necessarily larger, because coverage was incomplete.
The Directorate General of Post Clearance Audit’s parallel examination of luxury vehicle imports is more damning still, and its figures are all reported, not estimated. Across 1,335 vehicle declarations where the duty differential exceeded Rs 1 million, importers declared a cumulative value of Rs 670 million against assessed values of Rs 7.254 billion; duty actually paid was Rs 1.293 billion against a liability the audit computed at Rs 18.78 billion. The audit found that 99.8% of Land Cruisers cleared in the period were under-invoiced, and that in no examined case did the importer produce evidence of a lawful foreign-exchange remittance — which converts a customs valuation problem into a trade-based money laundering problem, at precisely the moment Pakistan is trying to demonstrate FATF and IMF compliance. FBR has contested the most sensational illustration — the 2023 Land Cruiser recorded at a declared value of Rs 17,635 — stating that the vehicle was in fact assessed at Rs 10.05 million with Rs 47.2 million in duties and taxes recovered, and characterising the criticism as the resistance of beneficiaries of the old system. That rebuttal is fair as to the single vehicle. It does not answer the aggregate.
Underneath all of this sits a quieter structural change. The audit reported green channel clearances approaching 60% of imports and 85% of exports, with no published criteria for channel selection. (Trade press reporting from the FCA’s first fortnight recorded the green channel share doubling from 25% to 50% while a 3,500-GD backlog was being cleared — a capacity decision, not a risk decision.) The audit’s own formulation is the sharpest sentence written about the FCA by anyone: Pakistan Customs has adopted front-end facilitation without matching back-end oversight, turning the green channel itself into a risk area. The audit also recorded declarations being cancelled on importers’ request once adverse findings emerged — a loophole converting a detected evasion into no case at all — and solar panel consignments landed in 2023 but held back and cleared only after the faceless system went live, suggesting some importers understood the new system’s blind spots before it opened.
V. The other bill: what delay costs the trader (estimated, on stated assumptions)
Revenue leakage is the loss the state can see. There is a second loss that appears in no FBR statistic because it is not paid to the state at all. It is paid by importers and exporters to shipping lines and terminal operators, largely in foreign exchange. No official body has measured it. What follows is therefore the author’s estimate — an order-of-magnitude calculation, not a measurement — with every input stated so it can be challenged, and with the reported inputs distinguished from the assumed ones.
The reported inputs:
• Detention/container rent at Pakistani ports: $80–150 per container per day (Federal Tax Ombudsman proceedings), with the FTO separately recording that these sums are paid at open-market dollar rates and repatriated abroad. FPCCI office-bearers have cited demurrage of $100–200 per day.
• Karachi Port container throughput: about 2.65 million TEU in FY 2024-25 (port statistics); Port Qasim adds roughly 1.2 million TEU (Port Qasim Authority historical data).
• Clearance-time increase under FCA: 57% (FBR Review Committee).
• Rise in review filings: from 6% to 14% of GDs (FBR Review Committee).
• Post-FCA average customs dwell: about 66 hours (trade press reporting of clearance data).
The assumptions (author’s own):
• Loaded import containers passing through the FCA-covered Collectorates: taken as roughly 1 million a year (derived from the throughput figures above after netting out exports, empties and transshipment; a deliberate simplification).
• Blended demurrage-plus-detention rate: $100 per container per day — the bottom of the documented band, chosen deliberately so the estimate errs low.
• Exchange rate: Rs 278 per US dollar.
• Share of ordinary containers pushed past their free time by the added FCA delay: 30%.
• Average chargeable delay on a contested (review-filed) consignment: 10 days, reflecting trader reports of reviews running from one to several weeks.
The calculation (all four buckets are estimates):
Bucket 1 — the universal extra day. A 57% increase on a base implied by the 66-hour post-FCA dwell means roughly one added day per container. If only 30% of a million containers are pushed past free time by it: 1,000,000 × 30% × 1 day × $100 ≈ $30 million, roughly Rs 8 billion a year (estimated).
Bucket 2 — the contested tail. The reported rise in review filings from 6% to 14% implies roughly 80,000 additional contested consignments a year on a million-container base. At 10 chargeable days each: 80,000 × 10 × $100 ≈ $80 million, roughly Rs 22 billion a year (estimated). The figure is sensitive to the delay assumption: at 7 days it is about Rs 16 billion; at 21 days about Rs 47 billion.
Bucket 3 — blocked capital. Provisional release under Section 81 of the Customs Act requires security for the disputed differential. Assuming 80,000 contested consignments at an average differential of Rs 500,000, roughly Rs 40 billion is pledged across a year; held an average of three months at prevailing KIBOR-linked financing and guarantee costs, the carrying cost is on the order of Rs 1.5–2 billion a year (estimated).
Bucket 4 — the delay tax. If contesting an assessment costs on the order of Rs 280,000 in demurrage and detention alone (10 days at $100, at Rs 278/$), then any over-assessment below that threshold is rationally paid rather than fought. Assuming just one consignment in ten carries an uncontested unlawful excess averaging Rs 30,000: roughly Rs 3 billion a year (estimated) collected without lawful basis — and booked, in official reporting, as revenue performance. This bucket is the most speculative of the four and is included because its logic, not its precise magnitude, is the point: the cost of the remedy suppresses the demand for the remedy.
Cost head Annual estimate (author’s calculation)
Extra day across import containers ~Rs 8 bn
Demurrage/detention on contested consignments ~Rs 22 bn (sensitivity: 16–47)
Carrying cost of Section 81 security Rs 1.5–2 bn
Over-assessment absorbed rather than contested ~Rs 3 bn
Indicative total ~Rs 35 bn/year (plausible range ~Rs 28–60 bn)
These are estimates and should be read as such. But two conclusions survive any reasonable adjustment of the inputs.
First, the bulk of this — everything in Buckets 1 and 2 — is foreign exchange, on the order of $100 million a year (estimated), flowing to shipping lines and terminal operators as the price of a domestic administrative failure. The FTO has already recorded the repatriation mechanism as a national foreign-exchange loss; what has been missing is the FCA-specific arithmetic. A country negotiating its external account with the IMF is exporting nine-figure dollar sums annually in avoidable demurrage.
Second, even at the bottom of the estimated range, the private cost of delay is the same order of magnitude as the hard reported revenue leakage. The state loses at the bottom of the assessment distribution; the trader pays at the top; the sum of the two is the true cost of the FCA in its current form. The country is paying twice for one reform.
Nor is this the whole bill, though the remainder resists even estimation. Delayed raw material idles production, and idle plant generates neither income tax nor sales tax — a second-order revenue loss never attributed to the port. Demurrage is a deductible business expense, so every rupee of it also shaves taxable profit. Landed cost rises, and rises fastest for the compliant importer, who has nothing to hide and therefore nothing to negotiate. FPCCI’s leadership — which has supported the FCA in principle — has stated the position plainly: Pakistan’s dwell time is roughly double that of regional competitors, causing heavy losses in demurrage, detention, storage and blocked capital, while the complaints resolution cell meant to catch these cases had gone inactive.
VI. The supervision vacuum
Every functioning tax administration answers three questions: who decided, on what reasoning, and who reviews it. The FCA answers none of them cleanly.
Under the Collectorate system, an assessment was a conversation with a record. The appraiser was identifiable, the principal appraiser sat above him, assistant and deputy collectors above that, and the trader could appear with documents, comparable imports, valuation rulings and technical literature and argue the case. That proximity was abused — nobody serious denies it — but it also supplied two things the current system lacks: a supervisory chain with subject-matter competence, and a hearing in which the trader had a voice.
The FCA replaced this with a first-instance review lying with principal appraisers posted inside the same Central Appraising Unit (CGO 6 of 2024). A unit reviewing itself is not supervision; it is quality control with the wrong incentives. Virtual hearings before assistant and deputy collectors were announced from July 2025 in response to the review backlog, but a hearing right that is administratively announced rather than statutorily guaranteed, time-bound and enforceable is worth very little — as India’s parallel experience with faceless proceedings has shown, where the scheduled video hearing the officer never joins has become a documented grievance.
Contrast the Indian customs design, whatever its own faults: re-assessment requires a speaking order, the importer must be heard before an adverse re-assessment, and appeal lies to the Commissioner (Appeals) of the port of import. Pakistan launched anonymity without the corresponding procedural guarantees. The officer became invisible; the reasoning did not become visible in his place. That is the wrong half of the Indian model to have copied.
And it cuts against the officers too. Appraising staff moved into a sanitised hall on reported twelve-hour shifts, cut off from the trade whose goods they assess, judged by throughput metrics, and blamed publicly when audits surface. Forty-five clearing agent licences were suspended in the first weeks, provoking strike calls from the Karachi and All Pakistan Customs Agents Associations and an FPCCI demand for independent inquiry. A reform that alienates both sides of the counter simultaneously has a design problem, not a communications problem.
VII. Is it a control system rather than a reform?
The honest answer is that it was not designed as an instrument of control, but it functions as one, because the accountability it imposes runs in only one direction.
Consider what the FCA actually redistributes. It takes discretion away from a named officer in a Collectorate the trader can walk into, and relocates it to an unnamed officer in a hall the trader cannot enter, reachable only through a portal. The trader’s identity is concealed from the officer — sold as impartiality — while the officer’s identity is concealed from the trader, which is something else entirely: it is immunity. Impartiality requires only that the decision-maker not know whose case it is. It does not require that the affected party be unable to identify who decided, or to hold that decision to a reasoned standard. The FCA bundled the two, and only one was necessary.
The result is a state that can assess without being answerable, and a trader who can object only by absorbing roughly $100 a day for the privilege. That is the structure of control, arrived at without anyone intending it. When a system’s remedy is slower than its error rate, the remedy stops being a remedy and becomes a deterrent to complaining — which is precisely what wandering for weeks to correct an over-assessment teaches an importer to stop doing. Bucket 4 above is the estimated price of that silence.
The counter-argument deserves a fair hearing. FBR’s stated position is that opposition is loudest among beneficiaries of the old system; that clearing agents and a section of the appraising cadre have material reasons to want it restored; and that reform of this kind always produces a transitional spike in disputes as informal understandings are withdrawn. There is truth in that. It is also true that under-invoicing of vehicles and other high-value goods long predates the FCA — the audit measured a pre-existing disease under new lighting, and some of the “loss” attributed to the faceless system is loss that was previously invisible. FBR further reports (its figures, not independently verified) Rs 30.46 billion in trade fraud detected across 174 cases under the wider transformation programme as the risk engine, Cargo Tracking System and National Targeting Centre came online. A critic who ignores that is being unfair.
But the defence establishes only that the old system was bad. It does not establish that this one is better, and FBR’s own Review Committee — constituted on the Chairman’s own orders — declined to certify that it was, recommending against further rollout unless the design was revisited or its efficacy demonstrated on larger datasets.
VIII. The future
The system will not be rolled back. Too much political capital was invested at inauguration, the reform narrative facing the IMF depends on it, and the institutional direction of travel is unmistakable: FBR notified a Programme Management Unit in July 2026 to drive a National Faceless Centre across Inland Revenue as well, with a rules-based risk engine, a phased rollout and — belatedly — a pilot designed to validate the operating model before national extension, plus a compliance register tracking superior court judgments and a data protection workstream. Those are the right instincts, learned late and at cost.
What will happen instead is layering: the faceless core retained and progressively patched with the controls that should have accompanied it from the start. The question worth arguing about is which patches.
Seven would matter more than the rest:
1. A statutory speaking order. Every re-assessment departing from the declared value should carry written reasons, the comparable data relied upon and the valuation ruling invoked. Anonymity of the officer is tolerable; anonymity of the reasoning is not.
2. A time-bound review with deemed consequences. A review not decided within a fixed short period — seventy-two hours is not ambitious — should trigger provisional release against security under Section 81 as of right.
3. Shift the demurrage risk. Where a review succeeds, the delay was the department’s error and the department should bear its price: statutory delay-and-detention certificates issued automatically on a successful review, obliging waiver by terminal and carrier for the contested days. Until delay costs the assessing side something, it will remain free to the only party able to prevent it.
4. Review outside the assessing unit. First-instance review must sit outside the Central Appraising Unit, before officers with no reporting line into it.
5. Restore specialisation inside anonymity. These are not opposites. Virtual assessment groups organised by sector preserve expertise while retaining random allocation within the group. This is the single highest-value correction available, on the Review Committee’s own findings.
6. Fix valuation at source. Anonymity cannot cure under-invoicing; only current, defensible values can. Section 25A valuation rulings need systematic refresh, and vehicle imports need mandatory linkage between the declaration and the underlying banking remittance before assessment, not after audit — the precise gap the PCA report exposed.
7. Symmetric measurement. FBR publishes detections. It should also publish over-assessment reversal rates, average review disposal time, dwell time distribution by trader size, and aggregate demurrage attributable to customs-side delay — converting the estimates in Section V of this essay into official statistics. A system that measures only one direction of error will only ever correct one direction of error.
IX. Conclusion
The Faceless Assessment System diagnosed the right disease and prescribed half a cure. Corruption in customs is real, discretion was its vector, and reducing contact was a defensible instinct. But contact was never the only thing the Collectorate provided; it also supplied expertise, a supervisory chain, and a forum in which a trader could be heard. The FCA removed all three and replaced only the first.
The accounting is unforgiving even when the estimates are held to their most conservative reading. On the state’s own audit, the system leaked a reported Rs 7.4 billion in hard evasion and uncollected penalties in a single quarter of partial coverage, within a composite quarterly figure of Rs 100 billion. On the author’s estimate — built on documented rates and the government’s own delay data — importers and exporters are paying something in the order of Rs 35 billion a year in demurrage, detention and blocked capital at the other end of the same distribution, most of it in foreign exchange. Both bills are generated by the same defect: an error-prone assessment married to a remedy slower than the errors.
Until the remedy is faster than the error, the ratchet will keep turning: the exchequer will keep losing at one end and the compliant importer will keep paying at the other. Anonymity was never the reform. Accountability was — and it remains unbuilt.