Field notes

Blog

Methodology posts on prospect scoring, deliverability lessons, and the occasional teardown of why a campaign worked or didn't.

2025-06-12 · methodology · 6 min read

Why LLM-personalized email beats templates by 4×

A 3.6% reply rate vs the industry's typical 0.9%, and what specifically about per-prospect generation drives the gap.

2025-06-04 · product · 7 min read

Intent scoring on Shopify stores — 40 signals that actually predict reply

From product catalog depth to checkout pixel presence, the signals we feed our scoring model and which ones turned out to be noise.

2025-05-28 · technical · 5 min read

AWS Bedrock vs self-hosted inference — the cost math at sub-1M tokens/day

Llama-3-70B on g5.12xlarge vs Bedrock pay-per-use. Why managed inference wins until you hit a very specific scale.

2025-05-14 · operations · 6 min read

Email deliverability in 2025 — what changed, what still works

Google's bulk-sender rules, Microsoft's tenant scoring shifts, and the warmup pattern we use to keep new sending domains above 99% inbox placement.