What did Pune City Traffic Police deploy?

Pune deployed TraffiCure, a software-only traffic intelligence platform built on Google Maps Roads Management Insights (RMI). It went live in January 2026 and made 964 roads visible across the city, refreshed every two minutes, with no new cameras or sensors installed.

How much did average speed improve in Pune?

Pune City Traffic Police publicly reported that average vehicular speed across the city's major corridors rose from 20 km/h to 26.8 km/h over two months of use, a 34% improvement. The figure was reported by the police themselves, not estimated by a vendor.

How did the data lead to permanent road changes?

TraffiCure ranked every road segment by measured delay, diagnosed the root cause of each chronic junction, and packaged the findings into junction-level engineering briefs. Pune City Traffic Police submitted those briefs to the Pune Municipal Corporation, which builds and changes the roads, feeding a city-wide sprint across 32 key roads and 22 priority junctions.

Did TraffiCure require new hardware?

No. TraffiCure runs entirely on Google Maps RMI probe data. There were no cameras, sensors, or civil works to procure, which is how 964 roads were covered from day one and the deployment went live in weeks rather than years.

Which agencies were involved?

Pune City Traffic Police led the deployment and manages traffic flow. The Pune Municipal Corporation handles the engineering and infrastructure changes. The underlying data comes from Google Maps Roads Management Insights.

Can other cities replicate Pune's approach?

Yes. Nothing about the approach is Pune-specific. The same four steps, rank the worst links, diagnose the cause, hand engineers a brief, and prove impact before and after, work anywhere Google Maps has coverage. TraffiCure is 100% software and deploys in weeks.

Pune Traffic Case Study: 20 to 26.8 km/h with TraffiCure

The shift

Most cities can describe their traffic. Pune decided to act on it.

Camera feeds covered under 5% of Pune's network; the other 95% was invisible. TraffiCure, on Google Maps' Roads Management Insights, made the whole network visible overnight: 964 roads, every two minutes, no new hardware. But visibility was only the start. The story worth telling is what the police did next. They turned the data into a ranked problem list, a root-cause diagnosis, and a formal junction analysis submitted to the city's engineering body. That is how 20 km/h became 26.8.

The old way, react

Worst junctions chosen by complaint volume
No proof of which corridor is actually worst
Asks to PMC backed by anecdote, not data
No baseline to confirm a fix worked

The new way, diagnose & direct

Every road ranked by measured delay, daily
Worst junctions chosen by data, not noise
Data-backed junction briefs for PMC engineers
Before and after proof on every intervention

Step 1 · Find the few that matter

From 964 roads to a ranked shortlist

You can't fix a city all at once. The first job was to turn a wall of data into a short, defensible list of where intervention buys the most relief.

5% 50%

The Pareto principle of traffic. A small set of chronic bottlenecks creates most of the city's lost time. TraffiCure ranks every segment by measured delay against its own baseline, so the police could target the critical few instead of spreading thin across all 964 roads.

TraffiCure scores every segment against its own dynamic baseline, for each 15-minute window, by day of week, and ranks the whole city. The output is the list below: the corridors where peak-hour speeds collapse furthest from free-flow.

Corridor / junction approach	Peak speed	Free-flow	Drop	Pattern
Eng. College Chowk to RTO Chowk	8.6 km/h	24.6	-65%	Severe
SSPMS College to RTO Chowk	7.0 km/h	15.8	-56%	Severe
Kalyani Nagar to Good Luck Chowk	8.5 km/h	16.4	-48%	Severe
Kamgar Putla to Shahir Aman Chowk	9.5 km/h	16.4	-42%	Moderate
Gandharv Lawn to Katraj Naka	8.2 km/h	12.2	-33%	Moderate
Navle Bridge to Wadgaon Pull	29.9 km/h	37.7	-21%	High-volume

9 AM to 7 PM average · actual speed vs. free-flow · Jan 30 to Mar 11, 2026 · Source: TraffiCure Analytics on Google Maps RMI

Why this list is different

It separates structure from noise

A road slow once is a bad afternoon; one slow every weekday at the same hour is structural, a job for an engineer, not a constable. Over 20 to 30 day windows, only the structural corridors reached the brief to PMC: old-city stretches like Subanshah Darga to Govind Halwai Chowk at 4.2 km/h and Madai to Subanshah Darga at 5.4 km/h, speeds no signal tweak will fix.

Step 2 · Diagnose the cause, not the symptom

A bottleneck is a symptom. Pune went after the disease.

A ranked list tells you where. The harder question is why, the right fix depends entirely on the cause. TraffiCure pairs each hotspot with the context that explains it.

1
Profile when it fails
Temporal heatmaps show the exact days and hours a junction breaks down. Pune's pain is sharply timed: peak alert activity hits 5 to 8 PM, with 6 to 7 PM alone generating 1,669 alerts.
2
Overlay local context
Speed data is fused with what's physically there, a bus stop, a work-zone merge, a mall entrance, a low-lying underpass, pointing to the object causing it.
3
Name the root cause
It distinguishes a breakdown from passenger exchange at a bus stop from one driven by weekend retail surge or monsoon waterlogging, each a different fix.

Illustrative diagnosis, how the workflow reads a junction

Worked example

A representative root-cause readout for a chronic junction, it turns "always jammed" into two specific, fixable findings:

Finding 1, flow breakdown at the rampPrimary cause

High correlation: bus-stop proximity. Passenger boarding and alighting at the stop nearest the flyover entry ramp repeatedly stalls the merge, collapsing speed on the approach during peak hours.

Engineering implication → relocate or re-design the stop set-down; a constable cannot fix a geometry problem.

Finding 2, weekend retail surgeSecondary cause

POI-driven spillover. Traffic queuing for mall parking paralyses the westbound approach on Saturdays, a demand pattern invisible to a weekday camera review.

Engineering implication → a weekend-specific traffic management plan, not a permanent signal change.

Illustrative of TraffiCure's root-cause method. Specific causes for each Pune junction were identified from that junction's live data.

Step 3 · Hand engineering a brief it can act on

The junction analysis Pune submitted to PMC

Traffic Police manage flow; the Municipal Corporation builds the roads. To change infrastructure, the police needed to give PMC engineers more than a complaint, a data-backed analysis they could design against.

Rather than mail a long platform report, Pune distilled the analytics into a focused junction brief, the format below, for each priority location, every claim traceable to live data.

Junction Analysis Brief Submitted to PMC · Engineering review

City-wide ranking: Where this junction sits on the city's delay table, ranked by total vehicle-hours of delay per week, so capital goes to the worst first.
Temporal signature: Day-of-week by hour-of-day heatmap showing exactly when the junction fails, so interventions and enforcement are timed to the problem.
Severity metrics: Average and maximum delay in minutes, speed drop vs. baseline in km/h, and estimated upstream queue length.
Root-cause finding: The diagnosed driver, geometry, bus stop, work-zone, POI surge or waterlogging, that determines what kind of fix is needed.
Recommended action: A specific, prioritised engineering or management recommendation tied to that cause.
Exportable GIS layer: Hotspot locations and severity as Shapefile or KML, dropped straight into the engineers' ArcGIS or QGIS workflow.

The quiet shift that makes it work: the police-and-municipality conversation stops being "this area feels bad" and becomes "this junction costs the city X vehicle-hours a week, here's why, here's the fix."

22

Feeding a city-wide decongestion sprint. The ranked analysis aligned directly with Pune's multi-agency push across 32 key roads and 22 junctions, removing encroachments, retiming signals, and clearing the merges the data flagged as the highest-cost failures.

Step 4 · Prescribe the fix

It doesn't just find problems, it prescribes the fix.

For each priority junction, TraffiCure pairs the diagnosis with a specific, costed recommendation, derived from live probe data and pressure-tested in simulation before it ever reaches a work order.

Navale Bridge to Warje, NH-48 Signal timing

REC-PUN-014 · Priority 1

severe in 15 minspillback detected 16:30simulated, 12 peak scenarios

Root cause, Spillback from the service-road construction merge chokes the corridor as the evening peak begins.

Sim forecast Queue dissipates about 9 min sooner; corridor speed recovers toward free-flow before spillback compounds.

Old-city ramp junction Geometry · Infrastructure

REC-PUN-021 · Structural

-56% vs free-flowrecurs 6 of 7 weekdayscause-isolated from data

Root cause, Passenger exchange at the bus stop sitting directly on the ramp entry stalls the merge; a secondary weekend surge comes from adjacent retail parking.

Why it matters Signal timing alone cannot move a structural bottleneck; the data tells PMC exactly which lever to pull.

Khadki Underpass Resilience · Weather

REC-PUN-033 · Monsoon

waterlogging risk 17:00 to 20:00weather-fused forecast

Root cause, A low-lying chokepoint that floods on heavy monsoon rain, collapsing one of the city's critical north links exactly when rain peaks.

Sim forecast Pre-emptive diversion holds a usable alternate route through the storm window instead of a full stall.

PMPML Route 7B corridor Transit priority

REC-PUN-046 · Bus network

45% of route delay on one 2 km stretchcorridor isolated

Root cause, Probe data pins nearly half the route's total delay to a single 2 km segment, wrecking reliability.

Sim forecast Targeted timing recovers the bulk of lost minutes on the route without new lanes or hardware.

Each recommendation was generated from that junction's live data and simulation; projected impacts are simulation estimates.

Step 5 · Intervene, then prove it worked

Every fix gets a before-and-after

The data doesn't end at the recommendation. The same baselines that found a problem measure whether the fix actually moved the needle, closing the loop public spending almost never gets.

1
Simulate
A planned closure or diversion is pushed to Google Maps with a click; TraffiCure shows the real-world spillover before a single barricade goes up.
2
Analyse
Live data reveals exactly which routes and zones absorb the displaced traffic, so the diversion plan is precise, not guesswork.
3
Measure
After the change, an impact assessment quantifies the percentage change in speed, travel time and reliability for the junction and its neighbours, a cabinet-ready verdict on whether it worked.

The proof artifact, intervention impact assessment

A cabinet-ready verdict on every change

Once a recommendation is implemented, the same baselines that found the problem measure whether it worked, a before-and-after on the exact KPIs leadership cares about. A sample readout for a signal-retiming intervention on a flagged approach:

KPI · evening peak	Before	After	Change
Average speed	8.6 km/h	14.1 km/h	+64%
Peak travel time	11.4 min	7.0 min	-39%
Travel-time reliability (buffer index)	0.71	0.42	-41%
Upstream queue length	~340 m	~150 m	-56%

Illustrative impact-assessment readout. The platform generates this before and after for any intervention, on the same baselines used to detect the problem. The city-wide measured result is below.

The result

20 » 26.8 km/h. Measured by the police.

Not a vendor estimate. Pune City Traffic Police publicly reported the gain themselves, a 34% rise in average vehicular speed across the city's major corridors over two months of use.

"Over the last two months of using the application, we have observed that the average vehicular speed has increased to 26.8 kmph from 20 kmph."
, Additional Commissioner of Police Manoj Patil, Pune City Traffic Police

By the numbers

34% Speed improvement 20 to 26.8 km/h, city corridors

839K+ Data observations in the first 6 weeks

15,034 Alerts generated across 836 roads

22 Priority junctions in the decongestion sprint

11.2 min Avg incident resolution rapid-deterioration alerts

48.6 min Avg congestion resolution congestion alerts

Press

The city announced it themselves

Pune went to the press with the result. "Pune is the first city in India to implement such a project," said Police Commissioner Amitesh Kumar. The story, 34% faster, in partnership with Google, zero new hardware, was carried by the Times of India, Hindustan Times, Times Now, Pune Pulse, Free Press Journal and The Bridge Chronicle.

Why it travels

A model any city can run next quarter

Nothing about Pune's approach is Pune-specific. The same four steps work anywhere Google Maps has coverage, which is everywhere.

What it took in Pune	Why it's repeatable
No new hardware	Runs on Google Maps RMI probe data. No cameras, sensors or civil works to procure.
Live in weeks, not years	964 roads covered from day one; the baseline engine self-calibrated to local rhythms in about 2 weeks.
A ranked problem list	Pareto-based hotspot ranking surfaces the critical few links that cause most delay, in any city.
A brief engineers act on	Junction analysis with cause, recommendation and exportable GIS slots into existing planning workflows.
A way to prove impact	Before and after assessment on the same baselines justifies the spend and compounds trust.

How Pune Turned Traffic Data Into a City-Wide Decongestion Plan

Most cities can describe their traffic. Pune decided to act on it.

The old way, react

The new way, diagnose & direct

From 964 roads to a ranked shortlist

It separates structure from noise

A bottleneck is a symptom. Pune went after the disease.

Profile when it fails

Overlay local context

Name the root cause

Worked example

The junction analysis Pune submitted to PMC

It doesn't just find problems, it prescribes the fix.

Every fix gets a before-and-after

Simulate

Analyse

Measure

A cabinet-ready verdict on every change

20 » 26.8 km/h. Measured by the police.

By the numbers

The city announced it themselves

A model any city can run next quarter

Frequently asked questions.

100% software. Deploys in weeks. Every road covered from day one.