Founded and Exited an Inventory Intelligence SaaS

Scaling a Supply Chain Compliance & Sustainability Platform

The product I set out to build was the wrong one, and it took fifty manufacturers to tell me why.

Founder - owned product strategy, canonical data model, integrations, pricing, go-to-market

Led a 12-person team across engineering, design, and customer success

Closphere dashboard

The Setup

Small and mid-sized manufacturers run on inventory data they do not trust.

Stock lives across Tally, Zoho, Vyapar, spreadsheets, and in-house systems built years ago by someone who has since left. Each holds a partial truth. None of them agree. A production manager placing a purchase order is working from numbers that were accurate at some point in the past, for some subset of locations, in units of measure that may or may not match what the supplier ships.

The visible symptom is stockouts. The actual problem is that nobody can say, with confidence, how much of anything they have.

My answer was the obvious one: build a better system of record. Replace the mess with a single clean source of truth. It was the right product on paper, and I spent the first stretch of the company building toward it.

Small and mid-sized manufacturers run on inventory data they do not trust.

Stock lives across Tally, Zoho, Vyapar, spreadsheets, and in-house systems built years ago by someone who has since left. Each holds a partial truth. None of them agree. A production manager placing a purchase order is working from numbers that were accurate at some point in the past, for some subset of locations, in units of measure that may or may not match what the supplier ships.

The visible symptom is stockouts. The actual problem is that nobody can say, with confidence, how much of anything they have.

My answer was the obvious one: build a better system of record. Replace the mess with a single clean source of truth. It was the right product on paper, and I spent the first stretch of the company building toward it.

The Decision

I did discovery the unromantic way, door to door across India, sitting with owners and warehouse managers who had no patience for pitch decks and infinite patience for someone who would actually listen.

Fifty-plus conversations in, the pattern was unmistakable. Nobody wanted another system of record.

Not because the existing systems were good. Because switching was unthinkable. Their accountants knew Tally. Their compliance filings ran through it. Their staff had a decade of muscle memory. Asking a 40-person manufacturer to migrate its books is asking it to stop operating for a month, and no amount of better product wins that argument easily.

What they wanted was for the systems they already had to agree with each other.

So I repositioned from ERP replacement to an ERP-agnostic intelligence layer. We stopped competing with Tally and started reading from it.

The product no longer owned the data. It owned the truth about the data.

That decision created a new problem, and it is the one I am proudest of solving. Once you read from someone else's system rather than owning it, a specific failure mode can kill you: silent sync failure. A connector stops pulling. Nothing errors. The dashboard still renders confidently, still looks correct, and a manager keeps making decisions from data that quietly stopped updating four days ago.

An inventory product that is unavailable is annoying. One that is confidently wrong destroys trust permanently, and you do not get it back. So the system's job became not only being right, but knowing when it might be wrong and saying so.

I did discovery the unromantic way, door to door across India, sitting with owners and warehouse managers who had no patience for pitch decks and infinite patience for someone who would actually listen.

Fifty-plus conversations in, the pattern was unmistakable. Nobody wanted another system of record.

Not because the existing systems were good. Because switching was unthinkable. Their accountants knew Tally. Their compliance filings ran through it. Their staff had a decade of muscle memory. Asking a 40-person manufacturer to migrate its books is asking it to stop operating for a month, and no amount of better product wins that argument easily.

What they wanted was for the systems they already had to agree with each other.

So I repositioned from ERP replacement to an ERP-agnostic intelligence layer. We stopped competing with Tally and started reading from it.

The product no longer owned the data. It owned the truth about the data.

That decision created a new problem, and it is the one I am proudest of solving. Once you read from someone else's system rather than owning it, a specific failure mode can kill you: silent sync failure. A connector stops pulling. Nothing errors. The dashboard still renders confidently, still looks correct, and a manager keeps making decisions from data that quietly stopped updating four days ago.

An inventory product that is unavailable is annoying. One that is confidently wrong destroys trust permanently, and you do not get it back. So the system's job became not only being right, but knowing when it might be wrong and saying so.

What It Cost

None of the 63 was the wrong customer. Every one of them made the product better, and I would sign all of them again.

But some wanted more than configuration could give them. They needed things the canonical model did not cover, integrations no other customer would ever use, workflows shaped around how their particular plant ran. Saying yes meant hiring. Four of my eleven people ended up customer-facing, on a business with 63 customers, and that ratio is not what a configuration-driven product company looks like. It is what one looks like when it is being pulled toward services.

The whole thesis was that onboarding should be a config change rather than an engineering project. That is what took six weeks down to seven days. Every bespoke yes moved us a little further from the thing that made the seven days possible.

I do not have a clean answer for how I should have handled it. The revenue was real, the learning was real, and a bootstrapped company does not turn down paying customers casually. But the tension between a configurable product and a customer who wants it shaped around them does not resolve on its own, and I was managing it rather than solving it.

What Shipped

Canonical model - item master, units of measure, multi-location stock, batch and lot data, and transaction types, so data from any source system could be compared on the same terms

Connectors - Tally, Zoho, Vyapar, and customers' in-house systems, with field mappings driven by configuration rather than bespoke engineering

Trust layer - local sync agent with cloud aggregation, continuous reconciliation against source ERP balances, and silent-sync-failure detection so the product could flag its own staleness

Intelligence - replenishment optimization built on reconciled data


Outcomes

$1.5M+

Paying Customers

95%+

Inventory Record Accuracy

<7 days

Avg Onboarding Time

Reconciliation effort dropped more than 70%, and replenishment optimization built on newly trustworthy data cut stockouts by 32%.

The company reached 1K+ users across 63 paying manufacturers with zero paid acquisition, ran profitably, and exited through a technology sale.

Reconciliation effort dropped more than 70%, and replenishment optimization built on newly trustworthy data cut stockouts by 32%.

The company reached 1K+ users across 63 paying manufacturers with zero paid acquisition, ran profitably, and exited through a technology sale.

What I Know Now

The hardest product decision is usually the one where the better product loses.

A single clean system of record was the better product on paper. It was also unsellable, because it asked customers to bet their operations on a switch they had no reason to make. The winning move was to be less ambitious about ownership and more ambitious about trust: read from what they already had, reconcile it continuously, and be honest about the gaps.

Capability is getting cheaper. Trust is not. I learned that on inventory data in manufacturing warehouses before I ever applied it to model outputs, and it is still the thing I build for.

The hardest product decision is usually the one where the better product loses.

A single clean system of record was the better product on paper. It was also unsellable, because it asked customers to bet their operations on a switch they had no reason to make. The winning move was to be less ambitious about ownership and more ambitious about trust: read from what they already had, reconcile it continuously, and be honest about the gaps.

Capability is getting cheaper. Trust is not. I learned that on inventory data in manufacturing warehouses before I ever applied it to model outputs, and it is still the thing I build for.

Reconciliation effort dropped more than 70%, and replenishment optimization built on newly trustworthy data cut stockouts by 32%.

The company reached 1K+ users across 63 paying manufacturers with zero paid acquisition, ran profitably, and exited through a technology sale.