The deduction nobody made is the most expensive thing in a set of books, because nothing in the books is wrong. The expense is booked, the payment is made, the bank agrees, and the reconciliation is clean.
It surfaces at the audit, by which time the disallowance is a percentage of the expense and the interest has been running for months.
Finding it is a comparison the books support well and a judgement they do not support at all, and most of the value is in keeping those two apart.
What the books can tell you
Three things, and all three are mechanical.
Every amount booked to an expense or purchase ledger. Who it was booked against, which is the payee. And every line sitting on a tax deducted at source ledger, which is what was actually deducted.
Put those together and you have, per payee for the year, what was paid and what was deducted. That is the raw material, and it is worth collecting once for the whole year rather than per voucher, because the aggregation across the year is the thing that matters: a threshold is usually crossed by the fourth small bill, not by one large one.
What you emphatically do not have yet is whether anything was owed.
What the books cannot tell you, and must not be assumed
The nature of the payment. A ledger called "Professional Charges" may hold a works contract. A ledger called "Repairs" routinely holds both a contract and a purchase of goods. A ledger name is how somebody in the office labelled a column, and the deducting provision turns on what the payment was for, which is a question for the person who knows the client.
This is the single assumption that produces wrong exposure figures. A tool that reads "Professional" and applies the professional-fees provision will be confidently wrong on a good proportion of a real set of books, and the figure it produces looks exactly as authoritative as a right one.
The workable shape is two stages. Collect the figures once, from the books. Then let the answer about each payee's nature of payment be given by a person, and recompute the exposure the moment it changes, without going back to the accounting software.
A lower or nil deduction certificate. A payee holding one has a different rate, or none, and the certificate is not in the books either.
Whether the payee is an individual or a firm, which changes the rate for some provisions.
Every one of those is a question, and the honest output marks the rows where it is unanswered rather than defaulting them.
Important
An exposure figure built on a guess about the nature of a payment is worse than no figure. It gets quoted to the client, and it is defended in a meeting by somebody who was not told it was a guess.
The cost of getting it wrong, under the 1961 Act
For financial year 2025-26 and earlier, the provisions a practice checks against are the familiar ones:
Where tax was not deducted or not paid, section 40(a)(ia) disallows a proportion of the expenditure, and section 201(1A) charges interest from the date the tax was deductible.
The disallowance is the headline number and the interest is the one that keeps growing while the question is unresolved, which is the argument for running the check in the year rather than at the audit.
- 194C, payments to contractors
- 194J, professional and technical fees
- 194-I, rent
- 194H, commission and brokerage
- 194A, interest other than on securities
- 194Q, purchase of goods
And the same test under the 2025 Act
For tax year 2026-27 onwards the same question is asked of different provisions, because the Income-tax Act, 2025 reorganised them.
The scattered deducting provisions become serial numbers in the Table in section 393(1). The disallowance for non-deduction sits in section 35(b)(i). The interest for an assessee in default is section 398(3).
The arithmetic is unchanged. What changes is every reference in the working paper, in the letter to the client and in the note in the file. A practice running the check across two years is running it against two sets of section numbers at once, which is exactly the situation the renumbering of the TDS forms creates everywhere else as well.
The useful discipline here is to hold the rates and thresholds for each year as dated entries, per year, rather than as constants in a spreadsheet. A threshold typed into a template in 2024 is still sitting there being quietly applied to 2026, and nothing about it looks wrong.
Where the bank statement comes in
The books are the primary source here, not the statement. The deduction obligation attaches to the credit or the payment, whichever is earlier, so a payment-only view misses the ones booked and not yet paid.
What the statement contributes is the payee. A ledger grouped as "Sundry Expenses" with twenty payments in it tells you very little; the counterparty on each of those payments tells you who to ask about. That is the same recognition step that makes a bank statement worth importing properly in the first place.
It also catches the case the ledger hides entirely: a payment to a party who does not appear as a payee anywhere in the expense ledgers, because somebody booked it to a balance sheet head.
And once the exposure is settled, the entry side is ordinary: TDS on a bank statement line covers where the deduction sits in the voucher.
Pro tip
Run the check in October, not in September of the following year. The disallowance is the same figure either way; the interest is not, and the client can still fix some of it.
Conclusion
The comparison is easy and the judgement is not, and the useful tools are the ones that are honest about which is which.
What the books say, the books say: the payments, the payees, the deductions, the totals for the year. What the books do not say is the nature of each payment, and that answer belongs to a person who knows the client, not to a ledger name.
Greenote collects the first part once from Tally and recomputes the exposure every time an answer about a payee changes, with the rates and thresholds read from dated entries per financial year, under the 1961 Act or the 2025 Act as the year requires. Every row carries what was not known rather than a default. See what the TDS workspace covers.
Turn bank statements into Tally vouchers on your own PC
Greenote reads 100+ Indian banks and posts clean vouchers straight into Tally, from your own PC. The statement is not uploaded, and nothing is typed by hand.
Free trial: 2 days or 10 statements, no card. Works with Tally Prime & ERP 9.
