Back

Building Products for the Warehouse Floor, Not the Demo

4 MINS

Building Products for the Warehouse Floor, Not the Demo

I product-managed internal supply-chain tools at Udaan before I co-founded EasyReplenish, and the two experiences taught me the same lesson from opposite ends: a supply-chain product is only as good as the messy reality it survives. A forecast that looks brilliant in a dashboard but ignores how a warehouse actually runs is just a nicer way to be wrong.

Internal tools have the most honest users

When you build an internal tool, your users sit a few desks away and they will tell you, immediately and without mercy, when it doesn't help. There's no churn dashboard to hide behind, either the planner trusts the number or they override it and go back to their spreadsheet.

That honesty is a gift. It taught me that adoption isn't won with features; it's won when the tool consistently makes a decision the user would have agonized over. The first time a planner stops second-guessing a replenishment suggestion is the moment the product is real.

Forecasts have to respect operations

A demand number is only half a decision. The other half is everything operational around it: lead times, warehouse capacity, how stock moves between six locations, what a supplier can actually deliver. I've seen elegant forecasts produce useless plans because they assumed the warehouse was a clean abstraction. It never is.

So I design backward from the floor. What constraint will break this plan? Where does the inventory physically need to be, not just how much of it? A recommendation that can't be executed is noise, no matter how accurate the math.

Trust is built one boring decision at a time

In inventory, the unglamorous decisions are the product, reorder this SKU, hold that one, shift availability toward the bestsellers. None of it is flashy. All of it compounds.

I learned to measure the second action, not the first. Anyone can get a planner to try a suggestion once. The product works when they take the suggestion the tenth time without checking it against their own gut. That's not a feature you ship. It's trust you earn by being right, quietly, again and again.

What carries over to EasyReplenish

Building for apparel brands now, the lesson holds: meet the operation where it is. Brands don't want a smarter dashboard, they want to stop losing sales to stockouts and capital to dead stock. My job is to turn the math into a decision they can act on today, on their floor, with their constraints. Everything else is decoration.

Background

Varun skipped presentations and built real AI products.

Varun Ravula was part of the April 2026 cohort at Curious PM, alongside 18 other talented participants.