Built a national audience for someone nobody outside politics had heard of.

Take a first-time candidate in a local race to national recognition.

That was the KPI, and it was the whole job: a genuine national audience and real personal credibility, starting from someone who had never run for anything. I came on as Growth and Outreach Lead in November 2025.

It was also the first time the thing I was measured on was getting someone to say they would vote, rather than getting them to spend money. There is no checkout to point at, so the number is harder to define and harder to keep honest, which made it the most interesting growth problem I have had.

What I owned inside that:

• The content engine, from what got made to when it went out.
• Scripting and strategy for everything volunteer-facing.
• Weekly reads of post-level data, moving production toward whatever was turning attention into engagement.
• Co-branded campaign material with 20-plus local businesses across the district.

The local businesses meant walking in and talking to owners directly, working out what they were willing to put in a window and what they wanted back for it. Visibility in a neighbourhood is bought in conversations, not in ad spend.

The other half was the doors. I led a team of ten canvassers, trained them, and A/B tested their scripts door to door. Then I built a Claude-based layer that read the team's call logs and flagged which scripts were failing and where conversations were dropping. The team's average persuasion rate went from 35% to 50% over eight weeks.

That layer ran weekly, not once. It read every log the team filed, pulled out the objections coming up most often and the lines that were losing people, and I turned that into the next week's script revisions and training. The same loop pointed at the content side: what voters raised on the phones and at the doors, sorted by neighbourhood, decided what we made and posted next.

It is the same system in both places, pointed at two different inputs. Once it was running, neither half depended on me having a good instinct that week.

I was phone banking alongside running the channel. My own book was 100-plus voters a day, at a 15% engagement rate, meaning they picked up and had a real conversation with me, and a 65% persuasion rate on those conversations, meaning they ended in a committed yes. I held that rate for the length of the campaign. I knocked doors myself occasionally too, to see the ground reality first hand, and had some of my best conversations and "pitching practice" out there.

I took charge of the content strategy, the data analysis behind it, and the canvasser and marketing outreach, down to the scripts for the phone banking calls and the door knocking. On election day I ran the logistics for more than a hundred volunteers across polling locations and canvassing shifts.

This is the most field work I have ever done, and it was well outside what I was comfortable with. It was also the first time I had done any of it for something I actually believed in.

Saikat lost. He also built a national following and a local presence out of nothing, which is what we came to do.

We had expected the result. The last months of the race were attack ads and millions in outside spending pointed at us, and through all of it he still built the largest field movement in San Francisco and the national presence that was the goal from the beginning. It inspired a new generation of organizers, Solidarity SF among them. We backed Connie Chan in the runoff and supported progressive candidates in other races, including Abdul El-Sayed in Michigan.

My part of it was turning voter data and what we heard on the phones into a content strategy the growth came out of, and a repeatable way of getting to whatever a given neighbourhood cared about most. I also came out of it able to knock a door, if a job ever asks me to.