If you have seen a figure for how many cases are pending in India, you have seen a number from the National Judicial Data Grid. It appears in newspaper reports, in parliamentary answers, in law review footnotes, in investor presentations about legal technology, and in the opening paragraph of roughly every article ever written about judicial delay. It is quoted constantly. It is examined rarely.
That combination, high circulation and low scrutiny, is how a good dataset acquires a bad reputation. The NJDG is a real achievement: a continuously updated, publicly accessible, national view of court pendency produced by the courts themselves. It is also a specific instrument that measures specific things in specific ways, and a great deal of the analysis built on it does not survive contact with what it actually counts.
What the NJDG Is, Structurally
The NJDG is a national aggregation layer over the case data that court establishments create as part of their ordinary administration. It sits downstream of the Case Information System deployed across the district judiciary and the High Courts under the eCourts programme. Records entered at a court establishment flow upward, and the Grid presents them as dashboards at national, State, district and establishment level.
It launched in 2015 for the district judiciary and High Courts, and the Supreme Court's own data was subsequently brought onto it, so the Grid now covers all three levels of the constitutional court hierarchy. It also carries a defined access route for institutional users under an open data policy, so that researchers and government departments can obtain data beyond what the public dashboards display.
The most important structural fact about it is the one people forget: the NJDG does not collect data. It displays data that already exists because a court had to record it in order to function. Nobody at the Grid is measuring anything. It is a window, and the view depends entirely on what was entered at the other end.
What it does not cover
The NJDG covers the constitutional court hierarchy: district and subordinate courts, High Courts, and the Supreme Court. It does not present a comparable national view of the tribunal system, of consumer commissions, or of revenue courts, each of which runs on its own systems and publishes on its own terms. Any statement about total pending litigation in India that quotes only the NJDG is a statement about courts, not about adjudication.
What It Actually Measures
The dashboards are richer than the single headline number suggests, and the disaggregations are where the useful information sits.
| Dimension | What the Grid shows | What it is genuinely good for |
|---|---|---|
| Level | Separate views for district judiciary, High Courts and the Supreme Court | Establishing the distribution of the workload, which is the single most misunderstood fact about Indian courts. |
| Civil and criminal | Split of institution, disposal and pendency between the two streams | Showing that district pendency is dominated by criminal matters, which reframes most reform arguments. |
| Age bands | Pendency grouped by how long matters have been pending, from under a year through the very old tail | Separating an ordinary queue from genuine backlog. This is the most valuable single view on the site. |
| Stage | Where pending matters sit procedurally, including matters at the evidence stage and matters awaiting service | Identifying process bottlenecks that are not judicial in nature, such as service failure. |
| Geography | State, district and establishment level drill-down | Comparing within a State, where taxonomies and practice are broadly consistent. |
| Institution against disposal | Cases filed and cases disposed over a period | Computing a clearance rate, which tells you whether the backlog is growing or shrinking regardless of the absolute number. |
| Case type | Breakdown by registered case type | Useful within a State. Risky across States, because case type taxonomies are local. |
If you take one habit from this article, take this one: stop quoting total pendency and start quoting the clearance rate and the age profile. The first is a number that impresses. The second and third are numbers that inform.
Reading the Dashboards Critically
Here are the specific things that go wrong when the Grid's numbers are used without care. None of these are criticisms of the Grid. They are properties of what it displays.
The unit of a case is not standard
What counts as one case varies with registration practice. Connected matters may be registered separately or clubbed. An execution petition arising from a decree is registered afresh and counted afresh. A criminal matter with several accused may generate one proceeding or several. Cross-State comparisons therefore compare quantities that are not quite the same quantity.
Pendency includes matters the court cannot move
A matter stayed by a higher court remains pending in the court below and appears in that court's pendency. So does a matter where the accused is absconding, or where a party cannot be served. Reading pendency as a measure of judicial productivity attributes to the court a number that partly measures things outside its control.
Disposal is a mixed category
A case decided after full trial, a case settled at a Lok Adalat, a case withdrawn, a case abated, a case transferred and a case disposed for statistical purposes all count as disposals. A disposal figure without its composition tells you throughput, not adjudication. This matters enormously around National Lok Adalat dates, when disposal counts move sharply for reasons that have nothing to do with ordinary court working.
Case types are locally defined
Case type codes derive from State practice and local statutory nomenclature. The same dispute is registered under different type names in different States. Anyone aggregating a category nationally is either doing a manual mapping or producing a number that quietly means several different things at once.
Data entry lag shows up as apparent change
Because the Grid reflects what has been entered, an establishment that clears a data entry backlog produces a visible jump in its numbers that corresponds to no change in what its judges did. Short-run movements in a small geography should be treated with suspicion for exactly this reason.
Five questions to ask of any number you take from it
- When was this taken? The dashboards update continuously, so a figure without a date is a figure from an unstated moment.
- Which level does it cover? District, High Court and Supreme Court numbers differ by orders of magnitude, and totals that mix them conceal the distribution.
- How much of this is stayed? Matters the court cannot progress sit in its pendency, and the court has no way to move them.
- What is in the disposal figure? Adjudicated outcomes, settlements, withdrawals, abatements and transfers all count, and their proportions differ by period.
- Did the period include a settlement drive? National Lok Adalat dates move disposal counts sharply for reasons unrelated to ordinary court working.
The denominator is usually missing
Pendency in a State with a large population and heavy commercial activity is not comparable with a small State's, and neither is comparable without knowing working judge strength. A pendency figure quoted without population, filings and judge strength alongside it is a number in search of a context.
The most common misuse, stated plainly
The headline pendency figure is routinely presented as a measure of judicial failure. It is nothing of the kind. It is a stock, and a stock is the accumulated difference between two flows. A system whose judges are more productive every year will still show rising pendency if filings rise faster, which they have. Use the clearance rate to talk about performance, and use the age profile to talk about harm. The total is a headline, not a finding.
The Gap Between Aggregate Statistics and Case-Level Data
The deeper limitation of the NJDG is not any of the above. It is categorical. The Grid answers questions of the form how many. It cannot answer questions of the form what was decided, or why, or whether consistently.
That distinction matters more than it sounds. Almost every genuinely interesting question about the justice system is a case-level question. Do similar facts get similar outcomes? Are two benches of the same court taking different views? What actually happens to bail applications in this category of offence, at this court, over a year? How long does an execution petition really take from decree to satisfaction, as opposed to how long the average civil matter takes?
None of those can be answered by a dashboard, because a dashboard is a count. Answering them requires the orders, the reasoning and the outcomes, which live in the judgment and order publication systems rather than in the statistics layer, and which are published unevenly and indexed thinly.
What aggregates answer well
System-level questions. Where is the workload, how is it distributed, is the backlog growing or shrinking, which stages are bottlenecks, which geographies are outliers. This is policy-grade information and it is genuinely valuable.
What aggregates cannot answer
Anything about content. What was held, on what reasoning, with what consistency, and whether comparable matters were treated comparably. A count has no opinion about the cases it counts.
Why the gap persists
The statistics layer was built from administrative fields that a court has to record anyway. The content layer would require structured extraction from the orders themselves, which nobody was ever asked to produce as part of running a court.
A dashboard can tell you that thirty thousand matters of a type are pending in a district. It cannot tell you whether they are being decided the same way, which is usually the question that matters.
How to Use the Grid Well
Take the number from the source, with a date
The dashboards update continuously. Any figure copied from an article is a figure from an unstated moment. Pull it yourself, record the date and the exact view you took it from, and cite it that way.
Prefer rates and distributions to totals
Clearance rate, age profile and stage distribution carry information. Totals carry rhetoric. If your argument depends on a total, it is probably an argument about scale rather than about performance.
Compare within a jurisdiction before comparing across
Within a State, case type taxonomies and registration practice are broadly consistent, so comparisons between districts mean something. Across States, they need caveats that most published comparisons do not carry.
Ask what the number cannot see
Before drawing a conclusion, ask which matters are stayed, which are awaiting service, which are old for reasons outside the court, and whether the period included a National Lok Adalat. Those four questions dispose of most careless readings.
Go to case-level data for any content question
If the question is about what courts decided rather than how many matters they handled, the Grid is the wrong instrument. You need the orders, and you need them across the relevant forums rather than from one.
What the NJDG Deserves to Be Said About It
It is easy to write a critical piece about a public dataset and leave the reader thinking the dataset is bad. That would be the wrong conclusion here, so let us be direct about it.
Before the Grid, nobody in India could say with any authority how many cases were pending, where, or for how long. Policy arguments were made on impressions. The judiciary itself had no continuous national view of its own workload. Building one, from administrative data produced by thousands of independent establishments, and publishing it openly rather than keeping it internal, is a transparency decision that many systems with far more money have not made.
The right criticism is not that the Grid is poor. It is that the Grid answers one class of question extremely well and has been asked, for a decade, to answer a different class of question that it was never built for. That is a problem of expectation, and the remedy is not a better dashboard. It is a research layer over case-level data.
Two different datasets, two different jobs
Statistics and judgments are not the same asset and should not be treated as one. The NJDG is the right instrument for system-level questions and the wrong one for anything about content. Judgment and order data is the right instrument for content questions, and it is far messier, far larger and far less complete. Knowing which one your question needs is most of the skill.
That is the gap CourtMesh works in. It indexes the case record itself across the Supreme Court, all twenty-five High Courts, the district judiciary and tribunals including NCLT, NCLAT, ITAT and CESTAT, drawn only from official government portals, with roughly 310 million records keyword-searchable and roughly 2 million carrying deeper semantic indexing. That makes it possible to ask what has actually been decided on a fact pattern, across forums, rather than how many matters of a type are pending. Neither replaces the other, and anyone who tells you a search index makes national statistics redundant, or the reverse, has not used both.
When the question is what was decided, not how many
The National Judicial Data Grid is the right tool for system-level questions and the wrong one for content. CourtMesh indexes the case record across the Supreme Court, all twenty-five High Courts, the district judiciary and tribunals including NCLT, NCLAT, ITAT and CESTAT, sourced only from official government portals. Search by issue, party, court, judge, act or section, and read the order at source rather than inferring outcomes from a count.
Explore CourtMesh
