Outcome analysis · 59,912 decided appeals
The assessee wins nearly three out of four appeals at the ITAT
Everyone in the profession has an instinct about this. Nobody has counted it. We read the outcome of every ITAT order on this database from 2024 onwards and sorted them by who actually brought the appeal.
BharatTax.net Research · 22 September 2026Figures computed from the BharatTax corpus, orders dated January 2024 to 21 April 2026. Text prepared with AI assistance.
The question a client asks first is the one the profession answers worst: what are my chances? The honest reply is usually a shrug and a war story. There is no published success rate for the Income Tax Appellate Tribunal in India.
There is now. Of the orders on this database from 2024 onwards, 59,912 state an outcome clearly enough to classify and name their appellant. That second part matters more than it sounds, and it is where a naive count goes wrong.
Who brings the appeal
Of the 59,912 decided appeals, 52,597 (87.8%) were filed by the assessee and 7,315 (12.2%) by the Revenue. The Tribunal is overwhelmingly a forum where the taxpayer is the one asking.
And how it goes for them
| Outcome | Assessee appeals | Revenue appeals |
|---|---|---|
| Allowed | 44.5% | 22.0% |
| Partly allowed | 28.5% | 19.8% |
| Dismissed | 13.2% | 52.9% |
| Remanded | 13.2% | 5.0% |
| Withdrawn | 0.7% | 0.2% |
Two things are worth pausing on. First, partly allowed is carrying real weight — a clean win is 44.5%, and the 72.9% figure includes the 28.5% where the assessee took something but not everything.
Second, remand is not a win. We counted “allowed for statistical purposes” — the formula used when a matter goes back to the Assessing Officer — as a remand, not an allowance. Treating it as a win, as a keyword count would, adds about a point and a half to the headline and would be the first thing a practitioner spotted.
The rate is remarkably stable
| Year of order | Assessee appeals decided | Won or part-won |
|---|---|---|
| 2024 | 9,772 | 73.5% |
| 2025 | 34,324 | 72.9% |
| 2026 (to April) | 8,477 | 72.4% |
Across three years and more than fifty thousand appeals the figure moves by barely a point. Whatever drives it is structural, not a fluctuation in the Tribunal’s mood.
Where you appeal, however, is not stable at all
Split the same assessee appeals by bench and the national average stops being useful. Benches with at least 300 decided appeals:
| Bench | Decided | Won or part-won |
|---|---|---|
| Cuttack | 695 | 90.5% |
| Bangalore | 3,018 | 85.4% |
| Chennai | 4,267 | 81.7% |
| Jaipur | 1,590 | 78.7% |
| Mumbai | 10,484 | 77.9% |
| Delhi | 10,040 | 76.1% |
| Pune | 3,018 | 73.3% |
| Surat | 981 | 69.1% |
| Ahmedabad | 2,596 | 68.2% |
| Kolkata | 2,399 | 67.8% |
| Hyderabad | 1,655 | 65.6% |
| Chandigarh | 1,661 | 64.0% |
| Indore | 716 | 57.0% |
| Visakhapatnam | 568 | 49.6% |
| Cochin | 1,586 | 48.1% |
| Raipur | 717 | 36.7% |
| Amritsar | 553 | 35.8% |
Abridged; benches between 69% and 78% are omitted for length.
How this was counted, and what it does not cover
The outcome of each order was extracted from the text of the order itself, and the appellant classified from the party named first. Orders whose outcome could not be categorised were excluded rather than guessed at — 7,697 of them.
The window is January 2024 to 21 April 2026. Earlier years are not included: the outcome extraction has been run over recent orders, covering 96% of 2026 and 92% of 2025 but well under 1% of 2015 to 2023. Reporting this as a figure for the Tribunal’s whole history would be a false generalisation, so we have not.
This is the Income Tax Appellate Tribunal only. It says nothing about what happens to these matters in the High Courts or the Supreme Court, where the selection effect is far stronger and the rates are very unlikely to look like these.
Method and sources
Computed from BharatTax.net, a free search engine covering 292,668 ITAT, High Court and Supreme Court judgments. Outcome and appellant classification by outcome_research.py. Figures are fixed at the publication date above and are not recalculated when the corpus grows.