Inside Indian manufacturing

The first wrong idea

The first wrong idea most software people have about manufacturing is that the factory is where the product gets copied.

You imagine the real work happening earlier. Someone has the idea. Someone designs the PCB. Someone models the enclosure. Someone makes the prototype work on a desk. Then the factory, in this lazy mental model, repeats the prototype many times.

That is almost exactly backwards.

The factory is not a photocopier. It is the place where the product is forced to tell the truth. A design that looked clean in CAD starts asking uncomfortable questions. Can this corner release from the mold? Can this screw be reached by a tired operator at the end of the shift? Does the plastic shrink the same way in May and December? Can this vendor keep tolerances stable after the first good sample? What happens when a returned unit comes back with a failure nobody saw during testing?

I did not understand this until I spent time around Homemate, a bootstrapped smart home company doing roughly 25 Cr ARR and still close enough to its own manufacturing that the business did not feel abstract. I had read about supply chains before. I knew the words: casting, molding, vendors, tooling, SKUs, returns. But knowing the words is not the same as seeing how much of a hardware company lives between them.

The surprising thing was how little of the important knowledge looked like invention.

It looked like repetition.

Manufacturing is an information system

The most useful way I found to think about a factory is not as a place with machines, but as an information system with consequences.

A software system moves bits around. A manufacturing system moves decisions around, and then those decisions become objects. If the wrong assumption enters the system, it does not merely throw an exception. It becomes inventory. It becomes a return. It becomes a customer support call. It becomes a part that cannot be used but still has to be paid for.

That is why the small details have so much weight.

In software, a vague edge case can remain vague for a long time. In hardware, it eventually becomes a box sitting somewhere. Someone has to open it, inspect it, rework it, discard it, or explain it. The physical world is less forgiving because it stores your mistakes.

This is also why the floor matters. The people closest to the line often know things that are invisible in the meeting room. They know which part feels slightly different when it comes from one batch. They know which step creates confusion when a new person joins. They know which defect appears only when the fixture has been used for too long. None of this sounds strategic, but strategy that ignores it becomes theater.

Before factory visits, I thought of manufacturing knowledge as mostly process knowledge. Afterward I started seeing it as feedback density. A good factory is a place where weak signals are noticed early. A bad one is a place where weak signals become strong signals by becoming expensive.

The hidden product

The visible product is the thing a customer buys.

The hidden product is the process that can make it again.

This distinction matters because a prototype can fool you. A prototype only has to work once, or a few times, under friendly conditions, with the builder nearby. A product has to work after it has passed through purchasing, storage, assembly, packaging, shipping, customer misuse, voltage variation, dust, heat, impatience, and returns.

This is why "it works" is a dangerously incomplete sentence in hardware.

Works how many times? Built by whom? With parts from which batch? Tested under what conditions? Does it still work when the customer plugs it in the wrong way? Does it still work when the vendor substitutes something that looks equivalent on paper? Can it be repaired? Can it be explained? Can it be made without the founder standing beside it?

The last question is the real one.

Founders love prototypes because prototypes reward intensity. You can stay up late, hack things together, make one unit beautiful, and feel the future for a moment. Manufacturing rewards a different virtue: making yourself unnecessary without letting quality collapse. The more the process depends on your personal attention, the less you have built.

That is a hard lesson for a young builder because personal intensity is often the only advantage you start with. But hardware does not scale intensity. It scales process.

Returns are a second factory

One of the most useful things I saw was that returns are not an afterthought. They are a second factory running in reverse.

A returned unit tells you something. Sometimes it tells you the customer was careless. Sometimes it tells you the documentation was unclear. Sometimes it tells you the product survived the lab but not the home. Sometimes it tells you the sales channel created expectations the product was never designed to meet.

In software, analytics can make you lazy because the dashboard arrives pre-counted. Hardware returns are messier and more honest. The object comes back with marks on it. A connector is loose. A surface is scratched. A part failed in a way that no one predicted because no one in the office uses the product like a real customer.

The temptation is to treat returns as a cost center. That is true financially, but incomplete operationally. Returns are also field research. They are expensive because they are real.

This changed how I think about Notch. A health band will not be judged by whether the first prototype can read heart rate. It will be judged by whether the hundredth person can wear it, charge it, forget about it, sweat into it, sleep with it, sync it, and still produce data clean enough to matter. Any one of those verbs can break the product.

The return path has to be designed almost as seriously as the purchase path. If a sensor fails, what do we learn? If a band is uncomfortable, how do we know whether the problem is material, size, climate, strap geometry, or user expectation? If data quality drops, how do we distinguish hardware failure from behavior?

That is the kind of question manufacturing teaches you to ask.

Vendors are part of the product

Software people talk about dependencies. Hardware people have dependencies that call them back.

A vendor is not just a supplier of parts. A vendor is a source of variance. They have incentives, constraints, cash cycles, habits, shortcuts, and other customers. A good vendor relationship is not simply cheaper pricing. It is predictability, honesty, and the ability to solve problems before they become formal disputes.

This is one reason hardware feels slower than software. You are not just waiting for parts. You are waiting for trust to become information.

The strange thing is that this makes hardware more human, not less. You can draw a supply chain as a diagram, but the real chain is a set of relationships. Who tells you when something is wrong? Who hides it until shipment? Who can do a small urgent batch without punishing you later? Who will explain why your design is annoying to make?

These questions are not separate from engineering. They are engineering under another name.

If a design requires a vendor to behave perfectly, the design is fragile. If a design works only with one supplier's best day, it is not ready. If a BOM saves a little money but creates a dependency that can stop production, the cost model is lying.

This is where I began to understand why experienced hardware people can sound conservative. They are not always less imaginative. They have just seen more ways for reality to object.

Why this matters for builders

The lesson I took from Indian manufacturing was not that hardware is hard. Everyone knows hardware is hard. The lesson was more specific: hardware is hard because every abstraction has to become a routine.

A design becomes tooling.

A tolerance becomes inspection.

A feature becomes assembly time.

A support promise becomes reverse logistics.

A cheap part becomes a risk someone must carry.

That is why factory visits are so valuable. They make business physical. You stop saying "supply chain" as if it were a cloud. You see the people, machines, dust, fixtures, boxes, ledgers, and phone calls. You see how much of a company is made of repeated small acts that do not look like genius but are harder than genius to maintain.

For me, the floor made hardware more attractive, not less. It removed the romance, but it replaced it with something better: respect.

There is a kind of ambition that wants to stay in the idea stage because ideas are clean there. Manufacturing is an antidote to that. It asks: can your idea survive being touched by the world many times?

Most cannot.

The ones that can are worth building.