LibraryReportsThe Monsoon Myth
Intugine Intugine · Logistics Intelligence
India Freight Data
Report No. 05

The Monsoon Myth

Across 7,375 live trips spanning three weather regimes — dry season, quiet monsoon, and a red-alert storm — the monsoon calendar barely moved freight delays. One storm doubled them. India's freight doesn't run on seasons. It runs on events.

Every logistics plan in India carries a monsoon buffer. Contracts pad lead times from June to September; dispatchers brace for the season as a whole. The assumption underneath: monsoon months are slow months. We tested it.

Three dispatch cohorts, sampled from the much larger volume live on the platform: a dry-season cohort from April, a quiet monsoon week from late July, and the red-alert storm week from our previous study. Same platform, same delay measure, three different skies.

7,375
Trips across three cohorts
+2pp
Dry season → monsoon season (flat)
Under a live red alert
−15%
India's season rainfall vs normal — a deficit monsoon
01 / THE LADDER

Three rungs flat. One doubled.

Delay rates across the four cohorts. The first three — dry season, monsoon-calendar quiet week, and the storm week measured outside the storm — sit in one narrow band. Only the cohort under a live red alert breaks away. Hover any rung to read it.

Cohort
47.4% delayed
02 / WHAT WE RULED OUT

This comparison had three traps

A naive April-vs-July comparison lies in three ways. We controlled for each before drawing a conclusion — and the flatline survived all of them.

The censoring trap

The July cohort could only be observed for 162 hours; April for 507. Long trips literally couldn't exist in the July data yet.

Fix: both cohorts matched to a 162h window
The mix trap

April's sample was 73% one shipper; July's just 47%. Different shippers carry different baselines.

Fix: paired on the 19 shippers present in both — 22.7% vs 22.0%, statistically indistinguishable
The distance trap

July trips were significantly longer lanes (median 191 km vs 137 km). Longer lanes run longer regardless of weather.

Fix: like-for-like distance bands — 3 of 4 showed no significant speed change
03 / THE WEATHER LAYER

The monsoon that wasn't there

Why did the monsoon-season cohort stay flat? Because the monsoon itself did. In the sampled July week, IMD shows most major freight states in rainfall deficit — and the season running 15% below normal nationally. The calendar said monsoon; the sky mostly didn't.

+253%

The contrast: on the one day a real system hit — 6 July, Maharashtra, IMD red alert — daily rainfall ran 253% above normal, corridor delays peaked at 53%, and the median delayed trip stretched from 6 hours to 50. That is what actually moves freight: not the season, the event.

Why this is an Intugine problem

Budget for events, not seasons.

A blanket monsoon buffer taxes every June–September trip for a disruption most of them never meet — and still underestimates the one corridor where a red alert lands. The unit of freight risk isn't a season. It's an alert, on a lane, on a day. Cruise™ detects the alert, isolates the lane, and triggers replacement in 15 minutes — not 2-4 hours. Intugine Discover finds the truck. Vedika calls the driver.

See it on your lanes →
04 / METHOD

How this was built

Approach. Three samples drawn from the much larger volume monitored on the Intugine platform: a 6 April dispatch cohort (1,307 trips within the matched window), a 20 July dispatch cohort (1,448 trips), and the 1–7 July study from Report No. 04 (4,586 trips split into a red-alert corridor group and a rest-of-India control). A trip counts as delayed when it missed its planned platform ETA. The April–July comparison was controlled for observation-window censoring (both matched to 162h), shipper mix (paired on the 19 common shippers), and lane length (haversine-distance bands); trip origins and destinations were geocoded to states and joined to IMD rainfall departure to classify each trip's weather exposure. Shipper identities are anonymised.

Sources. Intugine platform trip data; India Meteorological Department (IMD).

Caveats. The April and July cohorts are single-day dispatch samples, not full weeks; the storm evidence comes from one event; delay is measured against planned platform ETAs. The finding — a stable baseline across seasons with an event-driven spike — is consistent across all three cohorts but should be read as evidence, not proof, pending more storm events in the series.

More Reports