Content authored by practitioners with experience at Amazon India, Intuit QuickBooks, and the Tata Group. Meet the team →
TDS software pilots are often run on fresh data with no answer key, so nobody can tell whether the tool's matches and exceptions are correct.
Replay a closed quarter you reconciled by hand, keep your result sealed, then compare agreements, disagreements and new finds line by line.
Success criteria agreed in advance: must-find cases, false-match tolerance, explanation quality and export completeness, with a pass or fail rule.
A signed pilot report showing exactly where the tool agreed, disagreed and found something your team missed.
A fair TDS reconciliation pilot uses one quarter you have already closed by hand, one entity group and three inputs: the credit statement download (Form 26AS or Form 168), the TDS receivable ledger and the deductor PAN and TAN master. Judge it on exceptions explained correctly, not on a headline match rate.
If you keep a manual workbook today, the TDS receivable ageing workbook is a good source for your answer key, and the TDS runbook shows the steps the tool must replace.
What do you hand over and what do you hold back?
Hand over:
- The credit statement for the quarter, as downloaded
- The TDS receivable ledger at invoice level, with section or payment code
- The deductor master with PAN and TAN
- Optionally, bank receipts for the same quarter
Hold back your manual result. Write down your total exceptions and their reasons, seal the file and share it only after the tool has run. Q4 of FY 2025-26 (January to March 2026) is a useful choice: it uses old section codes, may already have correction statements filed after April 2026, and has had time for late filers to catch up.
What should you measure?
Four counts, each compared against your answer key:
- Lines matched with a reason. Are they correct? Count any false match.
- Unmatched lines with a reason code. Does the reason agree with your team’s?
- Special cases found. Quarter drift, wrong PAN, partial credit, TAN change.
- Disagreements. Where the tool classed a line differently from your team, who was right?
The fourth count is the most useful. A tool that finds something your team missed has earned attention; a tool that disagrees and is wrong has cost you review time.
What success criteria should you agree in advance?
| Test | What to give the vendor | What a pass looks like | Fail rule |
|---|---|---|---|
| Must-find cases | Five lines you know are exceptions, unlabelled | All five flagged with the right reason | Any one missed |
| False matches | Full quarter ledger and statement | No line matched to the wrong invoice or deductor | More than the agreed tolerance, for example 0 |
| Explanation quality | Your sealed exception list | 90 percent or more of reasons agree with yours, rest explained | Reasons missing or generic |
| Export completeness | Request for one deductor’s pack | Statement line, invoice, status and sign-off in one export | Screenshots only |
| Effort | Note hours your team spent on the pilot | Lower than your manual close for the same quarter | Higher, with no new finds |
The tax head signs the criteria before the data goes out and signs the result after.
What does a pilot result look like?
Illustrative case: a closed quarter with 412 receivable lines, of which your team found 37 exceptions by hand.
- Agreements: the tool flags 34 of the 37 with the same reason.
- Disagreements: 3 of your exceptions it marks as matched. On review, 2 were true matches your team missed because the deductor quoted a different invoice reference; 1 is a false match the tool got wrong.
- New finds: 4 lines your team had matched are flagged by the tool. On review, 3 are real (credit in a later quarter that had been treated as received).
Net: one false match against a tolerance of zero is a fail on that row, which goes back to the vendor for an explanation before any decision. The 3 new finds are worth noting in the report. If the gaps relate to missing credit, the Form 168 shortfall investigation guide shows how to confirm them.
What should a pilot not include?
Do not change your chart of accounts, connect live ERP feeds or run more than one entity group in the first pilot. Keep it to one quarter so the comparison is clean. Rollout planning comes after the decision; see reconciliation software implementation.
When should you schedule the pilot?
Avoid the weeks before quarterly TDS statement due dates under the 1961 Act timetable (31 July, 31 October, 31 January and 31 May), when your tax team is busy with filings and deductors are still uploading. The middle of a quarter works best: deductors have filed for the pilot quarter, credits have had time to appear and your team has hours to review disagreements.
Plan three short sessions rather than one long one:
- Data handover (half a day). Confirm row counts with the vendor: if your ledger has 412 lines and the tool loads 409, stop and find the 3 before going further.
- Run and first review (one to two days). The vendor runs the quarter; your team reviews exceptions without the answer key.
- Comparison (one day). Open the sealed result and score each row of the criteria table.
The aim of the pilot is evidence, not a commercial negotiation. Keep the scored criteria, the vendor’s exception list and your sealed result together so procurement can see why the decision was made. If two tools pass, run the same quarter through both and compare disagreements, which is a fairer tie-break than a second demo.
How Terra Insight fits
TransactIG’s TDS reconciliation matches Form 26AS, Form 168 and AIS against your TDS receivable ledger, with an exception queue and audit-ready exports. We run evaluations as a representative-sample fit benchmark on your data; the scope and sample are agreed in discovery.