ssh.autos

Recipes

Recipe · Apify Store · runs in a terminal

What 2,215 Indian government tenders on GeM show about who buys what

Railways and Defence show different buying patterns. The same export reveals where a supplier needs to open the bid document before budgeting.

COSTFree to reproduce from archived data
DATAKaggle CSV
TIMECollected October 27 to November 17, 2026 UTC
YOU NEEDPython standard library

If you sell services to Indian government buyers, sorting by bid count puts Defence well ahead of Railways in this GeM sample. Filter by service category and the gap narrows from 837 vs. 341 to 227 vs. 174. Filter for monthly cab and taxi hire and Railways accounts for 40 of 49 listings. The first alert I’d build would match the category, then use buyer names to route the results.

I used version 1 of the public GeM India Government Tenders dataset. It combines archived searches from October 27 to November 17, 2026 UTC. The export has one row per bid number and includes auctions. The searches determine which listings are visible here, so use the counts to understand this sample’s mix. Kaggle dataset · automatic/gem-tenders-scraper

Who buys what

Ministry label in GeM Listings Service category Computer category Service share
Ministry of Defence 837 227 39 27%
Ministry of Railways 341 174 17 51%
PMO 78 8 5 10%
Ministry of Home Affairs 46 5 7 11%
Ministry field blank 497 127 121 26%

I used the listing’s category ID prefixes services_ and home_info_comp_comp for the two columns. They count GeM classifications, so a computer inside a mixed bill of quantities might fall outside the computer column. Defence accounted for 54 of 58 repair and overhauling service listings in this sample. At the level of buyer routing, 672 of the 837 Defence listings carried the Department of Military Affairs label.

All 497 rows with a blank ministry still had a department. They included 121 of 228 computer-category listings. A filter that requires a ministry name would drop over half that category, even before you apply product keywords.

How much cash might a bid tie up?

An earnest money deposit (EMD) is bid security. The estimated bid value is the buyer’s stated estimate, where available. Both amounts below are in Indian rupees.

Field Rows with an amount Coverage Median among reported amounts
EMD amount 124 5.6% ₹106,938
Estimated bid value 51 2.3% ₹534,799

The export explicitly records no EMD required on 145 rows and leaves EMD status unknown on 1,946. Keep those states distinct in a shortlist. Only 18 rows have both amounts. The median of the individual EMD-to-estimate ratios in those rows is 3%. That overlap is small, so pull the bid document for the amount and any exemptions before budgeting.

When to check your queue

Month in India Standard Time Bid starts Bid deadlines
September 2026 158 31
October 2026 1,604 590
November 2026 361 1,548
Other months 92 22
Date missing 0 24

1,185 of 2,191 dated deadlines fell on a Monday or Tuesday. Among rows with both dates, the median span was 16 calendar days and 235 had a span of seven days or less. A weekly scan could miss a short listing if the scan falls near its deadline.

The October-to-November shift describes the dates captured by this collection window. It cannot establish a recurring procurement season. Use the latest deadline in the bid document when setting a reminder.

Reproduce the tables

Download version 1 from Kaggle and extract the CSV.

curl -fsSL -o gem.zip \
  'https://www.kaggle.com/api/v1/datasets/download/kirkins/gem-india-government-tenders?datasetVersionNumber=1'
unzip -p gem.zip data.csv > data.csv

Save the Python below as gem.py and run python3 gem.py data.csv. The script uses only the standard library.

import csv
import sys
from collections import Counter
from datetime import datetime, timedelta, timezone
from decimal import Decimal
from statistics import median

with open(sys.argv[1], encoding="utf-8", newline="") as file:
    rows = list(csv.DictReader(file))
assert len(rows) == len({r["bid_number"] for r in rows})
print("Unique bid numbers:", len(rows))

def category_count(group, prefix):
    return sum(r["category_id"].startswith(prefix) for r in group)

for ministry in ("Ministry of Defence", "Ministry of Railways", "PMO",
                 "Ministry of Home Affairs", ""):
    group = [r for r in rows if r["ministry"] == ministry]
    services = category_count(group, "services_")
    computers = category_count(group, "home_info_comp_comp")
    print(ministry or "(blank)", len(group), services, computers,
          f"{services / len(group):.0%}")

rail_cabs = "services_home_ve54336454_mont"
defence_repairs = "services_home_gene"
for category, ministry in ((rail_cabs, "Ministry of Railways"),
                            (defence_repairs, "Ministry of Defence")):
    buyers = Counter(r["ministry"] for r in rows if r["category_id"] == category)
    print(category, buyers[ministry], sum(buyers.values()))

print("Military Affairs in Defence:", sum(
    r["ministry"] == "Ministry of Defence"
    and r["department"] == "Department of Military Affairs" for r in rows
))
print("Blank ministry with department:", sum(
    not r["ministry"] and bool(r["department"]) for r in rows
))
print("Computer category:", category_count(rows, "home_info_comp_comp"))

emd = "emd_amount_inr"
value = "estimated_bid_value_inr"
for field in (emd, value):
    amounts = [Decimal(r[field]) for r in rows if r[field]]
    print(field, len(amounts), f"{len(amounts) / len(rows):.1%}", median(amounts))
print("EMD status:", Counter(r["emd_required"] or "unknown" for r in rows))
ratios = [Decimal(r[emd]) / Decimal(r[value]) for r in rows
          if r[emd] and r[value] and Decimal(r[value]) > 0]
print("Paired EMD/value:", len(ratios), f"{median(ratios):.0%}")

india = timezone(timedelta(hours=5, minutes=30))
def indian_date(text):
    if text:
        return datetime.fromisoformat(text.replace("Z", "+00:00")).astimezone(india).date()
    return None

starts = [indian_date(r["bid_start_at_utc"]) for r in rows]
ends = [indian_date(r["bid_end_at_utc"]) for r in rows]
months = ("2026-09", "2026-10", "2026-11")
for dates in (starts, ends):
    counts = Counter(d.strftime("%Y-%m") for d in dates if d)
    print([counts[m] for m in months],
          sum(n for m, n in counts.items() if m not in months), dates.count(None))
spans = [(end - start).days for start, end in zip(starts, ends) if start and end]
print("Monday/Tuesday deadlines:", sum(d.weekday() < 2 for d in ends if d))
print("Dated pairs, median days, seven days or less:",
      len(spans), median(spans), sum(days <= 7 for days in spans))

Data source

Every finding uses data.csv from version 1. The archive’s README documents the historical search window and how the export merges listing and detail snapshots. The CSV’s SHA-256 is 854d6555d57a78c9eb750e387d7258d31dc2a63617a3ea0d8fd4b771be9475b0. Kaggle dataset

Affiliate disclosure: The Apify Store link carries our referral code. If you sign up for a paid plan through it, ssh.autos may earn a commission at no cost to you.

Sources in this recipe

Machine-readable: openapi.json · llms.txt

Some links to Apify are affiliate links. We may earn a commission if you sign up and become a paying customer.