One Field in IRIS, Three Savings Instruments: Auditing the Arithmetic in the Letter to FBR

The file that reached my desk last week carried the wrong label. The label said Asia Cricket Desk. Inside were twelve pages and not a single player's name. The pages carried the Federal Board of Revenue, IRIS, Behbood Savings Certificates, the Pensioners' Benefit Account, the Shuhada Family Welfare Account and the Income Tax Ordinance 2026.

I have spent thirty-seven years counting scorecards, seam positions and line-breaking passes. This is not my ground, and I will not pretend it should be. But the work I do is not cricket — it is counting. And everything in this file hangs on a single question: is the field in IRIS that computes the tax on these savings instruments the correct field?

The Public Interest Litigation Committee of the Lahore Tax Bar Association has said exactly this in a letter to FBR. The letter names Waheed Shahzad Butt. The committee's charge is direct: there is a software error in the IRIS income tax return system, and the cost of it is being paid by the people least able to pay — widows, pensioners, and families of martyrs.

The notebook says one thing; the replay says another. In this case the replay has not arrived. FBR has issued no public response, and until it does, this file stays open.

Income tax returns in Pakistan are no longer filed on paper. FBR's IRIS portal handles registration, the return itself, tax computation, notices and refunds. What has to be understood first is that a return is not a document but a chain of fields, and every field is wired to a section or schedule of the Income Tax Ordinance 2026. The software does not reason about tax; it executes a translation that was completed long before it reached the screen.

Behbood Savings Certificates, the Pensioners' Benefit Account and the Shuhada Family Welfare Account sit within the national savings family, and all three exist for the same purpose: to give a return above the market to people who no longer have room to take risk. Behbood is built mainly for widows and senior citizens for whom the deposit is a lifeline, the Pensioners' Benefit Account for those who have retired, and the Shuhada Family Welfare Account for households that have lost their earner.

The profit on all three is "profit on debt", which is taxable income. Ordinarily the tax is deducted at source, the savings centre issues a withholding certificate, and the filer claims that deducted amount as a credit in the return. That is where the whole game sits: the gap between what has already been deducted and what the software declares payable.

Now the structural point that is the root of this matter. Three different instruments, three different tax treatments, and one form to fill. If the form collapses all three into a single line, the computation can be right for one of them and wrong for the other two. This is not a software-engineering nicety; it is a question about the seam between statute and form, which is exactly where nobody looks.

A necessary note on source quality. This is an advocacy committee's letter, not an FBR statement, and not the output of any independent investigation. The arithmetic in it is detailed and internally consistent, but the revenue authority has not verified it. I will read it the way I read a report from the touchline — claim first, test second, verdict last.

To understand where a tax form breaks, the chain has to be laid out: statute, then the Finance Act, then the schedule, then the form field, then the instrument code, then the computation, then the credit, then the refund. There are at least eight points in that chain where a decimal can slip, and all eight are wired to one another. A software error never arrives alone; it starts at one link and travels the whole chain.

I read this letter the way I read the Bob Houghton file in 2026 — define the shape, count the actions, then state the conclusion. Here the shape is the return form, and the actions worth counting are four checks. I count them because without them we cannot even establish whether the fault is in the software or in the reading of the statute.

The instrument code check. Every savings instrument carries its own code, and the code entered in the return decides which rule governs everything downstream. If the Behbood code is swapped for the code of an ordinary term deposit, the form will compute the wrong tax even if every other entry is perfect. This is the field nobody inspects, which is precisely why it should be inspected first.

The rate check. The law does not treat these three categories identically, and the rate schedule shifts with each year's Finance Act. If the form is picking up last year's schedule while computing this year's liability, the fault is not in the software but in a translation that never reached the form after the budget. In my experience, errors in tax systems are never random; they cluster at the year boundary, when the law moves and the form stays behind.

The credit check. Tax deducted at source is the filer's own money and must land in the field the computation actually reads. If it lands in a field the computation ignores, the filer sees a liability for tax already paid. That is not a technical glitch; it is double collection, and double collection always falls first on the person who cannot afford a lawyer to fight it.

The refund check. Suppose the computation is fixed. The next question is the repair path. If correction requires the filer to file a revised return, answer a notice and then wait months for a refund, then fixing the software means nothing at all. The only real fix is one where the system recomputes the affected accounts itself instead of making the citizen run the race again.

Reading these four checks, I found myself back in the notebook I filled at the Ambedkar Stadium during the 2026 Nehru Cup final. That night I counted twenty-three throw-in sequences and eleven second-ball recoveries, because I had become convinced the game was decided in the spaces between lines rather than in possession. The same holds here. The argument is about the refund figure, while the game is being decided in the one field that will never appear in a headline.

And this is the human part, the part a spreadsheet has no room for. A widow holding Behbood certificates does not have a tax consultant. The retired teacher with a Pensioners' Benefit Account has kept the arithmetic of her life on paper, not on a portal. For a Shuhada family the account is the last piece of security left. For all three, a software error does not mean money alone — it means a notice, a reply, a queue, and the sense of having been counted wrongly by the system. When the computation is wrong, what falls first is the trust the citizen placed in the system, and no refund restores that.

Now the weakest link in the letter, and I will say it against my own trade. The arithmetic is internally consistent. Consistency is not truth. A spreadsheet can be wrong in its first cell and remain flawlessly consistent in every row beneath it. When a committee builds its own calculation and then verifies its own calculation, we get what the scorebook gives us rather than what the field gives us — a closed document. That is why FBR's reply is not a formality in this file; it is the other half of the evidence.

The most interesting structural pattern here is that all three instruments emerge from one place and enter the tax net at one place. That is economy in design, and every economy in design creates a single point of risk. If one form serves all three, one wrong field hits all three at once. I have argued for years that data piped onward without a reconciliation layer carries its own error with it — and when that data is calculating someone's tax liability, the error does not stop there. It enters the audit trail and is read as fact years later.

An old habit is useful here. Every draft leaves a fingerprint on the squad sheet — meaning the form we fill carries within it the mark of that year's budget, that year's politics and that year's hurry. The IRIS form computing the wrong tax today is not one engineer's mistake but the sum of many small decisions, each of which pushed a new provision into an old box. A form's history is written in its code, and code never lies — it simply does what it was taught.

Now the angle that troubles me most, and it appears nowhere in the letter. The entire debate is about the amount — how much was over-collected, who gets it back. Nobody is asking who approved that form, who tested it, and which official signed off on the change to that one field. When a system's error surfaces, we demand an accounting of the outcome and not of the process — and the process is where the next error is already sitting.

I also accept that the committee's reading of the statute may be one of two defensible readings. Tax law has many places where two honest readers arrive at two outcomes, and there the court's finding is final. This letter is an argument, not a ruling. Those who read it as a ruling will make the mistake I never make in a match report — treating the scorebook as the field.

And here I am afraid of my own habit. I was never among those who mistook the notebook for the match, but when a letter arrives with twelve pages of clean arithmetic, reading it is a comfort. That comfort is the danger. The calculation the committee has given us is the committee's own; what is running inside FBR's system is still recorded in its notebook, not in ours.

Three concrete things are worth watching. Whether FBR concedes the error — which will show not in a reply but in a change to the IRIS form next quarter. What the repair path is: automatic recomputation of affected accounts, or a revised return demanded from every filer. And most importantly, whether those over-collected get their money back with interest or without it, because that single decision will reveal who pays for the system's mistakes.

Anyone who closes this file as a routine tax dispute will miss the real lesson. This is a case about who owns the translation when a statute is poured into software. In my notebook this file has not closed; it has only changed formations. And when FBR's answer comes, I will ask for the name of that one field — because the pattern is not in the noise; it is in the code that repeats.