A configurable private cloud platform

Every customer request turned into a new configuration, a new bill of materials, and a new thing engineering had to validate and the factory had to learn. Sellers were promising builds nobody had costed, and each one arrived as an exception rather than as a product.

What we had was a record of everything anyone had ever agreed to.

That kind of sprawl is expensive in a way that never appears as a line item. It gets paid in validation time, factory rework, slower quotes, and a handful of specialists who become the only people who can say whether a build is even valid. It also caps growth, because every new platform inherits the mess from the last one.

Product manager for the configurable platform. Across product, engineering, supply chain and the factory, I own the definition: what a governed design is, which ones exist, and what happens to a request that falls outside them.

Fix the chassis and the option rules first, then rebuild the catalog as a defined set of configurations validated together, rather than an open configurator with a price list attached.

That makes choice a product decision we make once, instead of a support problem we handle every quarter.

You give up the long tail of one-off requests, and you have to defend that in front of sellers who have already promised some of them. The exception path has to be real and staffed, or people route around the whole system and you have made things worse rather than better.

300+ fixed configurations became governed reusable designs. Buildable customer choice widened roughly three times with no matching growth in SKUs or new-product-introduction load.

Whether it worked came down to what a seller sees when the answer is no, and whether they get something they can take back to the customer.

If I did it again I would design the exception path first and the normal path second.