I Opened a Cricket File and Found a Tax Ledger: The IRIS Bug Whose Bill Is Paid by the Weakest Savers

The assignment sheet said two words: cricket, Asia Cup. I opened the file and there was not a single delivery in it. No batsman, no bowler, no scorecard, no fielding sequence. What was there instead was a letter — from the Public Interest Litigation Committee of the Lahore Tax Bar Association, addressed to the Federal Board of Revenue, alleging that the IRIS online income tax return system contains a software error that must be corrected. The letter named three savings instruments: Behbood Savings Certificates, Pensioners' Benefit Account, and Shuhada Family Welfare Account. It cited one statute, the Income Tax Ordinance 2026. And it carried one name, Waheed Shahzad Butt.

I counted. Five entities came out: one software, one revenue board, one lawyers' committee, three savings schemes, one statute. The cricket count was zero. No player, team, league, board or match appeared anywhere in the file.

I could have wrapped it in a cricket envelope. I could have built a frame about 'administrative carelessness in Pakistan during Asia Cup preparation', attached an old match reference, and let the reader believe they had read sport. But the ledger does not care what I hoped it would say. I will write what was in the file — and I will also write what was not.

What is actually in front of me

IRIS is the Federal Board of Revenue's integrated online tax system. For an ordinary taxpayer it is the route through which an income tax return is filed — registration, filing, claiming withholding credit, and checking refund status all pass through this portal. When someone says 'there is a bug in the system', they mean one link in that digital chain is computing something wrongly.

The body that sent the letter is the Public Interest Litigation Committee of the Lahore Tax Bar Association. It is part of a lawyers' body, not a government agency. The name attached to the letter is Waheed Shahzad Butt. That distinction should be registered plainly: this is an advocacy document, not the finding of an investigating agency. Its allegations are claims, not established facts.

The three instruments named — Behbood Savings Certificates, Pensioners' Benefit Account, Shuhada Family Welfare Account — are the categories in Pakistan's national savings architecture built not for the general investor but for specific groups. Behbood is for widows and senior citizens. The Pensioners' Benefit Account is for retired employees. The Shuhada Family Welfare Account is for the families of martyrs. All three share one feature: the deposited sum is the last protective layer on which a household survives a month, and all three attract deduction of tax on the profit on debt.

The relevant provision of the Income Tax Ordinance 2026 is withholding at source. The institution paying the profit deducts the tax, pays the remainder to the depositor, and issues a certificate for the amount deducted. The taxpayer then claims that deduction as a credit against their liability when filing the return. The whole structure rests on one link: the system must recognise the withholding certificate correctly and post the correct amount.

Where the arithmetic breaks

The committee's allegation sits on exactly that link. Its claim is that IRIS does not compute the deduction correctly for these three schemes, with the result that the taxpayer either pays too much tax, or has too little credit shown, or has the return rejected altogether.

Here I must state my limits plainly. The figures in the letter are not available to me in a verifiable form. I will not print those amounts, and I will not invent numbers to stand in their place. A reader is entitled to a date, a place and the name of the person who counted, attached to every number. A number I hold without those conditions is decoration to me, not evidence.

What I could check was the internal coherence of the argument, and that coherence exists. The three schemes, despite being similar in kind, carry three separate identities — a different profit rate, a different eligibility condition, a different tax treatment. The software has to hold a separate rule for each category. Where the rule for one category is mapped wrongly, the error does not stay isolated — it applies to that entire class at once. That is technically entirely plausible, and it is why the allegation deserves to be taken seriously.

Plausible is not true. I ran the back-test expecting to be right, which was error one. In this case my back-test says the shape of the claim is solid, but the evidence that would confirm it does not exist in the material I have. No official response from the FBR is recorded here. This story therefore stands at the status of 'claim', not 'confirmation'.

The size of the harm, and its invisibility

There is an aspect here that separates the case from an ordinary 'software glitch' story.

Consider a single taxpayer whose credit falls short by a few thousand rupees. For them the sum matters — in a widow's household it is a month of groceries. But for that sum they cannot hire a lawyer, cannot appeal, cannot bear a court's cost. The cost of litigation is several times the loss.

Now let the same error apply across thousands of accounts. The aggregate becomes large — and it belongs not to one person but to an entire class that has no means of complaining.

That is the real architecture of this case: the harm is real, but per person it is small enough to fall below the threshold of litigation. And harm that falls below the threshold of litigation is nobody's responsibility to fix.

The second face of the same architecture is the burden of proof. The system does not have to prove it is right. The taxpayer has to prove it is wrong. And a taxpayer already fighting the system while filing a return has neither the time nor the courage to later write to a committee. This is why such errors survive for years — not because they do not exist, but because nobody is left to expose them.

The direction of the fix, and the question of a date

The committee's demand is direct: the FBR should correct the software error. But there is a question the letter does not contain, and it is the most important one.

Will the fix apply going forward only, or retrospectively too? If the FBR repairs the portal tomorrow, what happens to the returns filed over the past three or four years under the wrong computation? Will the refund that was cut be restored? Will the rejected returns be reopened?

The date of the software fix is the line that decides whose account the loss lands in. If the correction applies only prospectively, the old loss stays in the taxpayer's account — and that taxpayer is the one with no route to complain.

Borrowing a model is easy; paying it back is the actual work. The model here was built on the assumption that a taxpayer is a generic person with a salary, a bank account and one kind of deduction. The widow, the pensioner and the martyr's family sit at the edge of that model, not in its lap. When a system rounds the people at the edge into the generic case, the resulting error is not a bug — it is the price of the design.

My own suspicion, turned against me

I have another habit to guard against. My reputation in data journalism is that of a man who finds the inverted result, and here too the temptation is to say — 'actually the FBR is right, the error is on the taxpayers' side.' I have no data behind that. Only a habit. Treating a habit as evidence is the very error I condemn in other people's copy.

So I am writing my threshold down in advance, so that I cannot move it later. This claim enters the confirmed category only if one of these happens: the FBR issues a formal response acknowledging the error; or an independent comparison of a taxpayer's return against their withholding certificate becomes public and shows the gap; or the committee publishes its full arithmetic so that a third party can re-add it. Until then this is an organised, coherent, unverified complaint.

The null result came back clean, so I wrote it down anyway. This time the result is not null — but it is not half-full either.

The first lesson

This was not a sports journalism assignment. A subject arrived that would have been easy to package as sport, and I did not package it as sport. That is the habit: when the subject arrives under the wrong label, write the subject, not the label.

The real question now is this — will the system be as quick to correct an error in a taxpayer's digital record as it is to collect? When a portal has an under-payment bug, it is found within a week and fixed within a day. When a portal has an over-collection bug, it takes a committee writing a letter.

That gap is what to watch next time. How many days until the fix, and whether it applies from a back date or a forward one.