Build an Index of Industrial Production dashboard from MoSPI data

Where to get IIP

MoSPI's eSankhyiki IIP page has Table, Chart, API and Metadata tabs and a Download button. The current series has base year 2022-23. The columns we observed were:

Check how Year encodes the month in your download before parsing it. IIP is revised, since initial monthly estimates are updated in later releases. Save each download with its date so you can show what changed.

A trap worth knowing: RBI's DBIE database still listed IIP on the 2011-12 base when we checked. If you pull history from there, you have two different series. Do not join index levels across them.

What it signals to a hiring team

Equity research desks, credit rating agencies, manufacturing companies' strategy teams and industry associations all track IIP. Many dashboards on the internet simply plot the headline. One that separates base effects from momentum, handles revisions and shows sector contributions shows you understand what the number means. That is the difference between reporting and analysis.

Cleaning steps

1. 1Filter the hierarchy. Type, Category and Sub Category are stacked in one table. Choose the level you are charting and filter to it, or totals double-count.
2. 2Parse months into a real date column. Sort by it. Never sort month names alphabetically.
3. 3Keep the published growth rate and compute your own. Where you have index levels a year apart on the same base, recompute growth and compare it with the published figure. Differences point to revisions or rounding.
4. 4Label provisional months. The latest months are usually provisional. Mark them on the chart.
5. 5Check base year on every row. Assert that it is a single value in your working table.

Code on a synthetic series

Index values below are synthetic. They follow a made-up seasonal pattern and are not MoSPI data.

import numpy as np
import pandas as pd

rng = np.random.default_rng(21)
months = pd.period_range("2023-04", "2026-07", freq="M")
season = {4: 0.98, 5: 1.02, 6: 1.00, 7: 0.99, 8: 0.97, 9: 0.99,
          10: 1.01, 11: 0.98, 12: 1.04, 1: 1.05, 2: 1.00, 3: 1.10}
rows = []
for sector, trend in [("Mining & Quarrying", 0.002), ("Manufacturing", 0.004), ("Electricity", 0.005)]:
    level = 100.0
    for m in months:
        level *= 1 + trend + rng.normal(0, 0.003)
        idx = level * season[m.month]
        if sector == "Manufacturing" and m == pd.Period("2024-05", "M"):
            idx *= 0.92   # a synthetic weak month to create a base effect a year later
        rows.append({"Base Year": "2022-23", "Category": sector, "month": m, "Index": round(idx, 1)})
df = pd.DataFrame(rows)
assert df["Base Year"].nunique() == 1

df = df.sort_values(["Category", "month"])
g = df.groupby("Category")["Index"]
df["yoy"] = (df["Index"] / g.shift(12) - 1) * 100
df["two_year"] = ((df["Index"] / g.shift(24)) ** 0.5 - 1) * 100   # annualised over two years
df["mom_3m"] = (df["Index"] / g.shift(3) - 1) * 100              # not seasonally adjusted
df["base_effect_flag"] = (df["yoy"] - df["two_year"]).abs() > 4

latest = df[df["month"] >= pd.Period("2025-04", "M")]
cols = {"yoy": 1, "two_year": 1}
print(latest[latest["base_effect_flag"]][["Category", "month", "yoy", "two_year"]].round(cols))
print(df[df["month"] == months[-1]][["Category", "Index", "yoy", "two_year"]].round(cols))

The flag compares one-year growth with the annualised two-year change. When the two diverge sharply, last year's base was unusual. In the synthetic data that happens in May 2025, the anniversary of a weak month, which is exactly what the flag should catch. The three-month change is not seasonally adjusted, so use it only to compare the same months across years, or apply a seasonal adjustment and say which method you used.

What to analyse

• Headline and sectors: year-on-year growth for General, Mining & Quarrying, Manufacturing and Electricity, month by month.
• Base-effect flags: mark months where one-year and two-year growth disagree, and explain the top three in words.
• Contributions: if the metadata gives sector weights, show each sector's contribution to headline growth. The contributions should add up to the headline, which is a check worth including.
• Revisions: compare two downloads a month apart, and show which months moved and by how much.

The finished dashboard

• Tile row: latest month's headline growth, manufacturing growth and the provisional marker.
• Line chart: year-on-year growth by sector over the full base-2022-23 span, with base-effect months highlighted.
• Bar chart: latest month's sector contributions, if weights are available.
• Revision table: month, first published value, latest value, change.
• README: source page, base year, download dates, hierarchy level, growth formulas, adjustment choices and the DBIE base warning.

Interview questions

• What is a base effect, and how did you detect it?
• Why can't you splice 2011-12 and 2022-23 index levels directly?
• How would you seasonally adjust this series, and what would you lose?
• The latest month is provisional. How would you present it to a director?
• Sector contributions do not add to the headline. What are the likely reasons?

Building a dashboard that a reader can trust is at the core of our Data Analyst programme. If you are deciding between analytics and data science, read our comparison of data analytics and data science.