A hurricane hit and devestated our service territory. Our restoration response was the best we’d ever run. Faster. More coordinated, more disciplined than anything before it.
Our public communication was a disaster. People were checking third-party apps that had nothing to do with us to see if their power was back on, because our own outage map was down. That’s how bad it got.
We fixed the outage side. Now, we had to fix public trust. We made ambitious public commitments. That exposed a new challenge, quieter but just as real: leadership had no reliable way to see restoration progress in real time. How many poles were going in today. How many miles of line were cleared. Were we on pace or falling behind. The system that was supposed to answer those questions was a legacy reporting tool feeding a spreadsheet held together by macros. It technically worked. It could maybe produce an answer once a week, if someone babysat it.
That wasn’t going to cut it for a response effort installing tens of thousands of poles, clearing thousands of miles of line, deploying hundreds of automation devices, a nine figure dollar amount in terms of commitment, all against the clock, all under public scrutiny.
I was in the room when leadership raised the question. Nobody was confident it was even feasible. I was starting to get proficient in BigQuery. Proficient enough to be dangerous. I said I’d figure it out.
I thought two weeks. It took eight.
The work wasn’t just standing up a new pipeline. It was reverse-engineering years of accumulated logic buried in someone else’s system. Every rule, every exception, every quiet assumption layered in by someone who’d never had to explain it. Migrating the data was the easy part. Migrating the logic was the actual job. Faithfully. Detail for detail. I had to understand that system better than a fresh read would ever tell you, because the new numbers had to match the old ones exactly. Not approximately. Not close enough.
Eight weeks of that got me a working replatform. It didn’t get me a trusted one.
What followed was months of going back and forth with the person who’d built the original system. Comparing outputs. Chasing down discrepancies. Getting asked, more than once, “how do I know this matches?” That back-and-forth is the part nobody puts in the demo. It’s not glamorous. It’s the actual cost of earning trust in a number.
Eventually, we sat down together and put the aggregates side by side. They matched, because I’d already done the work of understanding every piece of that logic well enough to rebuild it faithfully. The room comparison wasn’t the validation. It was confirmation.
That was the moment I took on the weight of a number other people were going to depend on. A number that would show up in public-facing reporting, that leadership would make decisions against, that had to be right because people were counting poles against it in the middle of a crisis. The responsibility didn’t transfer because I finished the migration. It transferred because I’d earned the right to be trusted with it.