Cutting order-to-dispatch by 62% for a 400-truck 3PL

Turant Logistics moves roughly 1,100 consignments a day across nine hubs with a mixed fleet of 400-odd trucks — about a third owned, the rest attached vendors. When we met them, the dispatch desk ran on three WhatsApp groups and a shared Excel workbook that one supervisor in Gurugram owned and everybody else screenshot.

What was actually broken

• Nobody knew where anything was. Vehicle location came from calling the driver. The ops floor had a wall of phones and a whiteboard, and the customer-facing answer to "where is my load" was a callback promise.
• Paper PODs were eating the cash cycle. The signed delivery challan travelled back to the hub in the driver's dashboard, got couriered to Gurugram, and was keyed in by hand. Median time from delivery to invoice was 8.4 days. On a business running on 30-day credit, that is 8.4 days of working capital lit on fire, every single load.
• Route planning was folklore. Multi-drop runs were sequenced by whichever supervisor was on shift, from memory. Nobody could tell us what an empty-run kilometre cost them because nobody was measuring empty runs.
• Disputes were unwinnable. With no timestamp, no geotag and no photo, a "short delivery" claim from a consignee was the consignee's word against a driver's. They wrote off a meaningful slice of revenue every quarter rather than fight.

The hard constraint — the one that killed the two off-the-shelf TMS products they had already tried — was connectivity. A large share of their line-haul runs pass through corridors where a phone has no usable data for hours at a stretch. Both products they had piloted assumed a live connection at the moment of delivery. Drivers responded the way drivers always do: they stopped using the app and went back to paper, and the pilot was declared a failure of "driver adoption". It was not. It was a failure of architecture.

The app went to two hubs first, ran there for five weeks, then rolled to the remaining seven over a month. Drivers were trained in Hindi by their own hub supervisors, not by us — a detail that mattered more than any feature we shipped.

Where the numbers landed

• Order-to-dispatch fell 62%. The dispatch desk stopped being a phone bank. Loads are now assigned against live vehicle availability, and the supervisor's job changed from finding trucks to approving exceptions.
• The POD-to-invoice cycle collapsed from 8.4 days to under 19 hours — a 91% reduction. The ePOD hits the server the moment the driver reaches signal; invoicing is a queued job that fires off the POD event. The finance team did not get faster. The information got faster.
• Empty-run kilometres dropped 18% in the first full quarter, straight out of the OR-Tools sequencing and backhaul suggestions. This is the line item that paid for the project.
• 3.4M GPS pings a day are ingested without the ops database noticing, and the live map holds sub-second freshness for a full fleet on a single modest instance.

The parts we care about more

Crash-free sessions on the driver app sit at 99.6%, and — the metric we actually watch — the offline queue has never lost a POD. Not one, across a run of over 300,000 deliveries. Short-delivery disputes are now settled by opening a geotagged photo with a timestamp, which ended the write-offs almost overnight and is quietly the outcome the CFO talks about.

The system has since absorbed a fourth region and a 40% fleet increase with no architectural change and no re-platforming, which is the only real test of whether the original decisions were right.