Three rules changed, each with a cost we accepted

Suspending speed-based pay deductions until the tracking data could carry them

When I joined Malvaneer in 2021, its 352 coaches carried tracking units from three different installers. Operations used overspeed alerts from those units to reduce drivers' monthly safety incentives, and about ₹38 lakh had been deducted in the previous year. Drivers' complaints had been treated as predictable resistance. In early 2022 my fleet data team compared the alerts with fuel slips, toll timestamps and crew records. Seventeen per cent of trip-hours had location gaps longer than fifteen minutes, and on one device type nine per cent of overspeed alerts came from position jumps implying impossible speeds.

I recommended suspending automated deductions until the data improved. Lalita Chaudhavar, the operations director, disagreed strongly. She believed drivers would read the suspension as permission to speed, and she had a serious accident in the company's history to point to. The managing director was also reluctant to replace devices fitted only three years earlier. We agreed a compromise: automated deductions would stop, but the safety desk would review every alert above the company's threshold and could still act where two independent sources supported it. Two reviewers joined the desk for that period.

We replaced 210 units of the worst type for ₹1.9 crore and corrected firmware and wiring on the rest. By March 2023 gaps had fallen to 2.6 per cent of trip-hours and false overspeed alerts to below one per cent. Deductions resumed in May 2023, and drivers could now see the trip trace behind any alert. Disputes fell from about 140 a month to 22. We gave up six months of automated enforcement and spent money the investor had expected to go on the booking engine.

The incentive scheme remains an operations decision. What changed was the evidence standard: no telematics output now affects pay without a published data-quality rate for that device type. I would involve drivers' representatives at the start next time. They had been right about the jumps for two years, and we heard them last.

Holding two berths on each conductor's handheld at a cost to online sales

Roadside boarding matters on our routes. Passengers wait for coaches at highway junctions and small towns without a counter, and in 2023 roadside sales were about seven per cent of passenger revenue. When we replaced paper waybills with a crew handheld, conductors could record those sales electronically for the first time. The difficulty was timing. A conductor in a ghat section without coverage could sell a berth that an online platform sold minutes later in Indore. In the first three months of the pilot we logged about 310 seat conflicts a month, many discovered at night when a passenger with an online ticket found the berth occupied.

There were three options. We could require a live connection for every roadside sale and lose passengers in dead zones. We could allow offline sales and resolve conflicts afterwards, accepting the night-time disputes. Or each handheld could hold a small block of berths that only the device could sell, released to online channels when the coach reached coverage at its last major stop. I recommended the third. The head of revenue opposed it: on busy routes, two fewer berths visible to online platforms meant fewer sales at the highest fares. The managing director approved the quota for a six-month trial.

We estimated ₹2.3 crore of online sales forgone in the first year, partly offset by roadside sales that were now recorded properly. Seat conflicts fell to 18 a month. Because every roadside ticket was captured on the device, the gap between conductors' cash and expected sales narrowed by 61 per cent, which the finance team valued more than I had expected. Conductors also stopped carrying paper books between depots.

The quota was a uniform two berths on every route, and that was my mistake. Some routes rarely see roadside passengers and lost online sales for no reason. I would set the quota by route and season from the beginning, using the roadside history we already had.

Buying the e-way bill layer of our parcel system and building only the counter

By 2024 parcels earned Malvaneer ₹62 crore a year, carried in the holds of passenger coaches. Counter clerks booked consignments in an old desktop tool and relied on senders to provide e-way bills where the consignment required one. Some did not. In 2023–24 our coaches were detained at roadside checks 23 times over parcel documents, sometimes for several hours with passengers aboard. One detention near Vadodara held a full sleeper coach for five hours on a wedding-season night.

My parcel systems team proposed building a complete parcel platform in-house, including e-way bill generation and GST invoicing. Their estimate was fourteen months and ₹4.8 crore, and the engineers argued, fairly, that owning the whole chain would give us control. I recommended a split. We would build counter booking, labelling and the hold manifest, which carried our operating rules, and buy e-way bill and invoice services from a licensed provider through its interface, at about ₹38 lakh a year. The counter system would refuse to label any consignment that needed an e-way bill until a valid one was recorded.

The head of parcels disagreed with the refusal rule. Clerks had long accepted parcels and sorted out the paperwork later, and some regular senders moved to other operators. In the first quarter, about four per cent of parcel value was turned away or redirected. Two of my senior engineers left within the year, partly because the most interesting work had been bought rather than built.

The system went live in six months at a build cost of ₹2.1 crore. Detentions for document reasons fell to three in 2025–26, and parcel revenue reached ₹81 crore. When the provider's service slowed on a heavy filing day, a rehearsed manual fallback kept counters open.

I signed the provider's standard contract without negotiating data export or notice of interface changes, and we spent four months renegotiating it in 2025. I would settle exit terms before the first integration, not after the dependency exists.