One Bug in IRIS, and the Ledger of People the Software Cannot Name

Editor's note — the limit of the beat

The Stage-1 label on this file read cricket_asia. The content is not cricket. There is no team here, no player, no match, no league, no cricket governance body. What is here is Pakistani tax administration: a representation sent to the Federal Board of Revenue by the Lahore Tax Bar Association's Public Interest Litigation Committee, asking the FBR to correct a software error in the IRIS income tax return system.

I am a cricket beat reporter. I will not turn this into a cricket story — no invented series, no fabricated quotes, no "sources close to." What I can do is write it with the same method I have used for thirteen years: who pays, how much, and whose count that payment fails to appear in. Ticket price, train fare, a day of leave — those are the three things I ask before I look at a scorecard. Today those three questions are attached to a tax form.

---

What I did not see, and what I have

I was not standing at a bank counter in Lahore. I did not watch an elderly man carry his Behbood certificate receipt from window to window. What I have is an advocacy document — written on the committee's behalf, with Waheed Shahzad Butt's name recorded in it — and the arithmetic inside it, which is detailed and internally consistent.

The microphone was borrowed; the question is real.

That is protocol, not a stylistic habit. When I talked my way into the overflow press area at Chepauk in 2026, the mic belonged to a friend and the laptop sat on a cool box — and I decided then that I would never write from a vantage I had not occupied. Here my vantage is a document. A document written by an interested party, not produced by independent investigation, and one the FBR has not publicly answered. The source quality is medium-low. That belongs in the first paragraph of the piece, not the last.

---

Three accounts, three kinds of people

Behbood Savings Certificates, Pensioners' Benefit Account, Shuhada Family Welfare Account — three instruments in Pakistan's National Savings system built to return more than the market rate, and their intended audience is explicit: widows, pensioners, families of martyrs. That is deliberate design. This is the financial safety net for people with no second source of income.

The profit on all three is subject to withholding at source — under the Income Tax Ordinance 2026, deducted by the savings centre or the bank, before the money reaches a hand. That deduction is not an informal arrangement; it is statute. And the moment it is deducted, a liability is created on the other side: that the amount must appear in the taxpayer's account as tax already paid. IRIS is where that reconciliation is supposed to happen.

This is where it collides.

---

The technical shape of the fault

The committee's claim is that IRIS is mis-mapping the amount withheld on these three instruments. When a taxpayer files, the tax already deducted either does not appear, or appears under the wrong head, or is counted in a way that inflates the system's computed liability. The result can be any of three things: a demand notice that never existed before, a refund that stalls, or a wealth-statement mismatch the taxpayer has to disprove himself.

One thing has to be logged here, because it is the weakest joint in the whole story: the arithmetic in the representation is detailed and internally consistent, but nobody outside the committee has recomputed it. The revenue authority has not confirmed it. The number I trust least is any number no one but the committee has added up. The mechanism is specific and plausible; the magnitude is unverified. Keeping those two sentences apart is the discipline this story requires.

---

Withholding is a trust arrangement

People treat withholding as a technical process. It is a trust arrangement, and its terms are one-sided. The state says: I will take your money now, before you see it, as interest. In return I promise that at the end of the year, when you settle your accounts, this amount will appear in your account — and if your total income falls below the threshold, it will come back to you.

The place where that promise is honoured is not a counter and not an officer. It is a piece of software. IRIS. And on the day the mapping in that software goes wrong, the second half of the bargain does not return the first half. The line reading "tax paid" becomes "tax due," and the taxpayer has no cheap route to challenge the change.

This is the same thing I have watched in cricket for twenty years, in a different uniform. When the review system freezes on the wrong frame, the damage does not fall on the team with the bigger staff. It falls on the one with nobody to appeal. Same here.

---

Why those three instruments

The committee did not pick those three products at random. The selection is the fastest line in the whole argument.

A Behbood certificate usually sits in a household where there was one earner and now there is not. A Pensioners' Benefit Account belongs to someone whose income comes from a single source after the job ended. A Shuhada Family Welfare Account belongs to families the state itself handed this instrument to as compensation. All three share one thing: none of these holders keeps a chartered accountant or a tax practitioner on retainer. None of them logs into the IRIS dashboard daily.

Which means something plain: a bug that lands on a population with no capacity to comply is not a technical fault, it is distributional politics. A fault that hits a taxpayer with a lawyer gets fixed in a week — because the lawyer calls, sends reminders, and the possibility of a name in the paper arises. A fault that hits a widow who does not know she is supposed to call anyone runs for years.

And one thing needs logging that does not help the committee. The members of the Lahore Tax Bar Association are tax practitioners. A broken IRIS generates work for them. That does not discredit them — a bar association's job is to speak for the taxpayer, and the taxpayer here cannot speak. But the interest alignment belongs in the ledger, because it determines how much weight we give the document.

---

Three layers: the rule, the paper, the man

On every big story I open three layers. Here the three layers sit in different places, and that is the real difficulty.

The first layer is the rule. The Income Tax Ordinance 2026 provides for withholding at source and for that withholding to be adjusted or refunded in the return. On paper, that is clean. The second layer is the paper — the certificate or passbook on which the bank has written the amount deducted. Also clean, and stamped. The third layer is the man — the pensioner who holds both documents, both correct, and is still being told by IRIS that he owes something.

The third layer carries no stamp. That is its strength and its weakness. When the system's calculation and the citizen's paper collide, the burden of proof shifts automatically onto the citizen — because the system has a default answer and the citizen has one file. And a citizen who does not already know what he is supposed to do never reaches an appeal. He pays quietly, or stops filing, and the next year his name moves onto the list of people who did not comply.

That is where a software bug stops being an administrative problem and becomes a social event.

---

The cost ledger: the price of one error

My habit is to measure everything from the pocket of the person who carries it. So here is that arithmetic, and it is the same ledger I learned standing six hours outside the gate at Chepauk.

The cost of fixing the bug: one database query, one amended mapping rule, one backlog re-processing run. Days of work for any government IT team, not weeks. At institutional scale, close to zero.

The cost of living with it, per person: one trip to a bank branch, one token, one hour in a queue to prove a certificate is genuine. Then a tax practitioner's fee — and that fee can exceed the amount of tax in dispute. Then two or three days of leave, which is usually meaningless for the pensioner but not for the son who takes him. And if the notice goes unanswered on time, the penalty on top — which can be larger than the original error.

Here is the real economic argument of this story: what costs the institution one day to fix costs a hundred people one day each to endure. And those hundred people are precisely the ones with the least room to waste a day.

I will not run the South Africa–India comparison here, because a comparison only earns its place when I have the numbers on both sides. But one thing can be said without a number: the withholding model exists everywhere, and everywhere its weak point is the same — the gap between the deduction receipt and the return's acceptance. The wider that gap, the more money falls into it, and that money usually belongs to the person with no platform to complain from.

---

What nobody is saying: the direction of the error

Here is something that is not in the representation and is my own reading, not the source's.

Software errors run in two directions. One understates what the government is owed; the other understates what the taxpayer has paid. Technically they look identical — a mapping, an index, a condition. Administratively they are not identical at all.

An error that costs the government money usually gets fixed inside a few days, without public debate, inside the system — because someone is watching for it. An error that makes a taxpayer's credit disappear requires an association to write a representation, and then a wait.

I am not saying anything was done deliberately. I have no evidence of that, and saying so without evidence is exactly the kind of journalism I avoid. I am saying that errors in a system have a direction, and that direction is not set by coincidence — it is set by who has the power to catch the error. In this case that power arrived late, and it took an association to arrive.

---

The other reading: not a bug, a schema

The easy way to read this story is as a technical fault, and to wait for the IT wing to patch it. My reading is the reverse.

The mapping rule inside IRIS that decides which deduction goes under which head, and who gets the adjustment, is in effect a piece of tax law — written not in parliament but in a schema. When the language of a statute is converted into fields in software, every decision in that translation becomes a policy decision. And that decision is made in a room where nobody requests public comment, no accountability is fixed, and no record is published.

That is the real story. The bug is a symptom. The condition is that we have moved a large part of tax administration into a layer the citizen cannot see and for which there is no defined error-checking process. Even if this bug is fixed tomorrow, the question stays: what is the arrangement for catching the next one — an independent reconciliation, a public log, a process that tells a taxpayer what changed in his account and why.

Which is why I do not treat this as a software story. I treat it as a new chapter in the old story where the power always sits with whoever writes the ledger — and today the one writing it is not a clerk. It is a database.

---

The press box hears the game before the scoreboard admits it

There is a thing I learned in cricket that applies directly. The scoreboard is sometimes wrong. And when it is wrong, ten thousand people in the ground catch it first — because they have counted the runs, counted the balls, and their own arithmetic is colliding with the board. The press box hears the game before the scoreboard admits it.

In a tax system there is no such crowd. There are no ten thousand people keeping their own count and colliding with the system. That is why system errors take years to surface — and why, when an association finally raises its voice, that voice is alone, and carries the burden of proof by itself.

When the stands are empty, the beat becomes a question — and the question is who built the scoreboard, and who checks it.

---

The honest state of the sourcing

I want to leave this where a reader can decide for himself, because what I have is not the whole story.

What is certain: the Lahore Tax Bar Association's Public Interest Litigation Committee has asked the FBR to correct IRIS, Waheed Shahzad Butt's name is recorded on the representation, three specific savings instruments are named, and the Income Tax Ordinance 2026 is cited. The committee's arithmetic is detailed and internally consistent.

What is not certain: the precise technical form of the fault, the number of taxpayers affected, and the magnitude of the sums involved. On all three, no public response from the FBR is on record. One side's document and the other side's silence — that is the whole capital this story has, and I will not write past what it can carry.

This is the discipline I picked up after the 2026 mega auction, when a franchise media manager called my column unfair. I rewrote the lede twice that night and then set a rule: twenty-four hours before answering criticism. It applies here too. The FBR deserves time to respond, and the reader deserves to know whether a response came.

---

The question that sets next month's scoreboard

I will not close this on a conclusion, because the conclusion has not arrived. I will leave one question, and it is addressed to the people with the authority to answer it.

Will the FBR say how many returns this year show a mismatch between the amount withheld on these three instruments and the amount on the certificate the bank issued? And if that number is published, will the list of taxpayers also be published — people who will be told their accounts have been corrected without an application, without a fee, without standing in a queue?

The day that list is printed, we will know whether this was a bug or a habit. Until that day, the most important record is not the one IRIS is displaying — it is the receipt a widow keeps in her drawer, which no server is reading.