I led DevOps at a niche eDiscovery data warehouse. We didn’t win business through marketing. Our founder had spent an entire career in New York big law, and clients came to us through those relationships. What kept them was that we would run searches nobody else would.
Leading wildcards. Terms shorter than three characters. Most platforms block searches like these because they’re expensive and messy to index. Litigators wanted them, and we were the warehouse that said yes.
Delivering that meant relying on Oracle Text features that few other Oracle customers ever touched. Rarely used features break in rarely seen ways. When ours broke, the vendor’s normal support process had little to offer, because almost nobody else had ever hit the problem.
So one of our DBAs had a job I’ve never seen anywhere else. He had a deep math background, and his explicit purpose wasn’t to fix things when they broke. It was to understand each failure well enough to talk to the mathematicians inside Oracle who built those features, get a bug opened, and then test their fixes before we trusted them in production.
That sounds like a luxury. It was actually risk management. Our critical differentiator lived in someone else’s code, in a corner the vendor rarely visited. The only way to keep it working was to have someone who could meet the vendor’s engineers at their level.
What I took from it:
- Know where your differentiator actually lives. Ours wasn’t in our own codebase. It was in a few obscure features of a database we licensed.
- If you depend on the edges of a product, managing the vendor becomes part of running the platform. Staff for it and plan for it.
- Some roles exist to translate, not to fix. They’re easy to cut and expensive to lose.
- Before you replace a platform, ask what it does for you that nothing else does. When engineering later proposed swapping Oracle for Elasticsearch, that was the first question I raised.
(And no, none of this means I like Oracle. See my last post.)
Originally posted on LinkedIn. Join the conversation there.