What lunchboxes taught me about product distribution

Every morning in Mumbai, roughly 5,000 men in white cotton kurtas and Gandhi caps fan out across the city's suburbs. They knock on doors. They collect tin lunchboxes. They load these lunchboxes onto bicycles, carry them to railway stations, sort them by colour codes painted on the lids, load them onto trains, unload them at different stations, sort them again, and hand-deliver them to office workers at their desks. By 1pm, 200,000 lunchboxes have reached 200,000 people. By evening, the empty boxes travel back the same way.

No app. No GPS. No algorithm. No venture capital. The error rate is one misdelivered lunchbox in every 16 million transactions. Harvard Business School wrote a case study about it. Not because of the technology. But because of the complete absence of it.

The dabbawalas of Mumbai have been doing this since 1890. And the reason I keep thinking about them has nothing to do with logistics or supply chains or operational excellence, which is what everyone else writes about when they discover this story.

I keep thinking about them because they understood something about products that the entire tech industry is about to learn the hard way.

Building the product was never the hard part. Getting it to the right person, at the right time, reliably, repeatedly, at a cost that doesn't kill you? That's the hard part. That's the whole game. Every failed product I've worked on in twenty years across startups and enterprises had the same root cause. It wasn't bad design. It wasn't bad engineering. It was bad distribution. The product worked. It just never reached the people it was built for.

And in a world where AI just made building almost free, distribution is about to become the only thing that matters.

1. Building got cheap. Distribution didn't.

Here's a conversation I've had four times in the last six months. A founder shows me something they've built. An AI tool that does something genuinely impressive. Summarises legal contracts. Generates personalised workout plans. Automates competitor analysis. The demo is slick. The underlying model is sharp. I'm nodding along, because the product is actually good.

Then I ask: "How are people finding this?"

Silence. A pause. Then some version of: "We're going to do some content marketing. Maybe Product Hunt. We'll figure it out."

They won't figure it out. Not because they're not smart. But because the problem they're ignoring is harder than the problem they just solved. They've spent three months building something impressive and zero days thinking about how it reaches the people who need it.

AI has collapsed the cost of building to near zero. A solo founder with Claude and a weekend can ship what took a team of twelve six months to build in 2019. I've watched this happen in real time. The prototyping phase that used to take weeks now takes hours. The gap between idea and functional product has never been smaller.

But here's what hasn't changed: the cost of getting someone to notice, try, trust, and keep using your product. That's the same as it ever was. Maybe worse, because now there are a thousand other founders who also had a productive weekend.

The irony is almost painful. We've been telling founders for years that execution is everything, that ideas are cheap, that the hard part is building. But we were wrong about which part of execution was hard. Building was the part we could engineer our way out of. Distribution is the part that requires patience, relationships, and systems that don't scale the way engineers want them to.

The dabbawalas inverted the whole equation in a way that I think is quietly profound. They don't make the food. They don't care about the recipe. Some home cook in Borivali makes the meal, packs it into a tin box, and hands it over. The dabbawalas' entire business is the distribution layer. The product is outsourced. The distribution is the company.

Most product teams operate with exactly the opposite assumption. They pour 90% of their energy into the product and treat distribution as a phase that comes after. "First we'll build something great, then we'll figure out how to get it to people." This is like a dabbawala saying, "First let's perfect the lunchbox design, then we'll figure out how to deliver it across Mumbai."

The dabbawalas understood something that most product teams still haven't: the meal doesn't matter if it never arrives.

2. The best distribution systems are boring by design.

The coding system the dabbawalas use hasn't fundamentally changed in decades. Colours, numbers, and symbols painted on tin lids. Each marking tells the dabbawala where the box was picked up, which train it goes on, which station it gets off at, and which building it's headed to. A semi-literate worker can read this code and route a lunchbox across a city of 20 million people.

It is, by any startup standard, aggressively uninteresting. Nobody is trying to disrupt the handoff at Churchgate station. Nobody is building a better coding system. Nobody is pitching investors on "Uber for dabbas." They just do the same thing, six days a week, fifty-one weeks a year, and it works.

We don't talk about this kind of thing in product. We romanticise innovation. We celebrate the clever feature, the beautiful interface, the breakthrough technology. Nobody writes a breathless LinkedIn post about their activation email sequence or their customer onboarding checklist. But those boring systems are often the difference between a product people try and a product people keep.

I learned this at Freshworks, watching the difference between products that had strong distribution engines and products that didn't. The ones that grew weren't necessarily the ones with the best features. They were the ones where the entire journey from "I heard about this" to "I can't stop using this" had been thought through with the same obsessive care that a designer would give to a hero screen. The onboarding. The first-value moment. The expansion triggers. The stuff nobody puts in their portfolio.

But that work is invisible. It's the plumbing. Nobody wins a design award for a well-structured onboarding sequence. Nobody gets promoted for making the referral loop 15% more efficient. And so talented people avoid it, because it doesn't feel like real product work. I've been guilty of this myself. Early in my career, I would have considered distribution work beneath the craft of product design. I was wrong. It is the craft. Everything else is a sketch until distribution makes it real.

Sound familiar? It should. It's the same prestige problem that makes designers avoid internal tools. Distribution work is the internal tooling of growth: essential, unglamorous, and perpetually under-invested in.

Nobody has ever disrupted a reliable distribution system by making it more interesting. They've disrupted unreliable ones by making them more boring. More repeatable. More consistent. The dabbawalas know this. The best growth teams know this. But most product teams haven't figured it out yet. They're still chasing the clever hack when what they need is a boring system that works every Tuesday.

3. Distribution is a network problem, not a technology problem.

The dabba system is structured as autonomous units of about 25 people each. Flat hierarchy. Decentralised decisions. Each dabbawala knows his 30-40 customers personally. He knows which floor they work on, what time they take lunch, which colleague will accept the box if they're in a meeting. The system's reliability comes from relational trust and local knowledge, not from centralised control or sophisticated tracking.

Each dabbawala is essentially a node in a distribution graph. And the graph works because the nodes are human.

I think about this when I see startups pour money into paid acquisition as their primary distribution strategy. Performance marketing. Google Ads. Instagram campaigns. The funnel mindset: pour money in the top, hope customers drip out the bottom. It works, up to a point. But it's expensive, it's fragile, and the moment you stop spending, the tap turns off.

The distribution systems that compound are networks, not funnels. Slack grew because one team inside a company adopted it, and then the team next to them saw it in action, and then the whole floor was using it. Figma spread through design teams the same way. Not because of a brilliant ad campaign. But because of human-to-human transmission. Someone showed someone else, and that person showed someone else. No one asked them to. The product moved through the network the way gossip moves through an office: irresistibly.

I spent five years at Adobe, and one thing I noticed was how differently products grew depending on whether they had a network distribution mechanism or a funnel distribution mechanism. The products with network effects (Creative Cloud's collaboration features, for instance) grew in ways that felt almost organic. But the products that relied on traditional marketing grew in ways that felt like pushing a boulder uphill. Every month you had to push again. Stop pushing, and the boulder rolls back.

The dabbawalas don't have a growth team. They have 5,000 people who each know 30 doors to knock on. That is a growth team. And it costs almost nothing to maintain, because the network sustains itself through relationships, not through ad spend.

This is the part that's hardest for product people to accept. The best distribution strategy might not require a single clever growth hack. It might require building something that naturally flows through human networks. Something people show to other people. Not because you asked them to. But because they can't help it.

4. The real moat is the last mile.

In the dabba system, the train journey across Mumbai is the easy part. Predictable. Scheduled. The same route every day. But the hard part is the last mile: the walk from the station to the specific desk on the specific floor of the specific building in a crowded business district. That final stretch is where the system earns its reputation.

The dabbawala who handles the last mile doesn't use a bicycle. The streets near the offices are too narrow, too packed with people and handcarts and autorickshaws. He carries the crate on his head. He adapted to the constraint. He doesn't wish the streets were wider. He doesn't build a better cart. He carries the crate on his head because that's what the last mile requires.

In product terms, the last mile is activation. Getting a user from "I downloaded this" to "I can't live without this." And it's where most products quietly die.

I once worked on an IoT product at Schneider Electric, a thermostat application for home users. The product itself was solid. The interface was clean. The engineering was sound. But the last mile, getting a non-technical homeowner to connect the device to their Wi-Fi, pair it with the app, set their first schedule, and actually trust it enough to stop adjusting the thermostat manually, was brutal. We lost more users in that last-mile stretch than at any other point in the journey.

We'd built the train. We hadn't built the walk through the crowd.

I've seen this pattern repeat across twenty years and more products than I can count. The teams that obsess over the product but treat activation as a checkbox. The beautifully designed app that nobody uses past day three. The powerful tool that churns at 40% in the first month because the distance between "signed up" and "got value" is too far, too confusing, or too forgettable.

But here's what's interesting: the teams that do get the last mile right almost never do it through better technology. They do it through better understanding of the specific, messy, human context their users live in. They study the narrow streets. They learn which floor the customer sits on. They adapt to constraints instead of wishing them away.

Building the product is the train ride. Distribution is the walk through the crowd with a crate on your head. And most teams spend 90% of their energy on the train and wonder why nobody's eating lunch.

Every morning, a dabbawala in Andheri picks up a tin lunchbox from a home kitchen. He doesn't know what's inside. Doesn't need to. His job is to make sure it arrives warm, on time, at the right desk. He's been doing this for years. He'll do it again tomorrow.

There is something in that repetition that the product world hasn't learned to value. Not the meal. Not the recipe. Not the ingredients. But the daily act of carrying it from where it was made to where it is needed. The unsexy consistency of showing up at the same door at the same time.

The lunchbox arrives warm. That's the whole product.