It was April 2020. Mumbai had been under lockdown for three weeks, and 150,000 shipments were stacked floor-to-ceiling at my DTDC branch. The last-mile had collapsed. Delivery partners had stopped showing up. Vehicles were grounded. And every day, the pile grew.
Most managers were waiting for instructions from above. I decided to build a network from scratch.
The Problem Was Structural, Not Operational
The first thing I realised was that this wasn't a manpower problem. It was a network design problem. Our standard delivery model depended on franchise partners who had their own constraints: vehicles, staff, compliance. When those constraints broke, the whole chain broke with them.
What we had instead was a city full of people on two-wheelers with no work.
Building the Network in 48 Hours
I spent the first day on the phone. Not with corporate. With auto-rickshaw unions, delivery gig workers, and small logistics brokers. I negotiated per-shipment rates on the spot: no contracts, just handshakes and WhatsApp confirmations.
By the end of day two, I had 60+ two-wheeler partners ready to ride.
The routing was done entirely in Google Sheets. I partitioned the city into 12 micro-zones, assigned partners to zones based on their home locations, and built daily manifests by hand. It wasn't elegant. It worked.
What the Numbers Showed
Over seven days, we cleared the entire 150,000-shipment backlog.
- 98% on-time delivery across the operation
- Less than 0.5% damage rate, better than our normal benchmark
- Zero customer escalations from the backlog period
The operation became a case study shared across DTDC zones. Three other branches replicated the model during subsequent lockdown waves.
What I Learned
Speed of iteration beats perfection of design. A working routing sheet in Google Sheets on day two is worth more than a perfectly optimised dispatch system on day ten.
The bottleneck is almost never where you think it is. Everyone assumed the problem was vehicle shortage. It wasn't. It was partner dependency: a network design choice that made sense in normal conditions and became a single point of failure under stress.
Operations under pressure reveal your actual system. The crisis didn't break anything that wasn't already fragile. It just made the fragility visible faster.
The lesson I took forward: design systems that degrade gracefully. When the primary network fails, what's the fallback? If you don't have an answer, you don't have a complete design.