467
Days
Built from my own data export: 467 physiological cycles, 182 workouts, and 411 logged behaviors, analyzed across recovery, strain, and sleep.
I wore from 2022 to 2023, stopped for about three years, and returned in 2026. The gap creates a before-and-after comparison. HRV fell 28% between the two periods, the largest change of any metric.
My HRV tracks training volume, not noise: it fell 28% when I stopped training hard, and it is already climbing back as my 2026 volume rebuilds. The drop was reversible detraining, not decline.
Rock climbing is the one activity common to both eras. Its average heart rate barely changed, while session duration and volume dropped sharply. The change was in training volume, not intensity.
My workout intensity held (avg climbing HR 116 → 114); only volume and variety changed, from near-daily climbing to a lighter spread. The fitness dip came from training less, not training worse.
The behavior with the largest effect in the data. On days I logged a drink, next-morning recovery fell 18 points and HRV dropped 15 ms. Red-recovery days were nearly 5 times more likely.
One drink roughly halves my odds of a green recovery and makes a red morning nearly 5x more likely (37 drinking vs 106 sober days). Alcohol is the biggest controllable lever on my next-day readiness.
Everyone knows alcohol hurts sleep, but what surprised me is that mine barely changed. My sleep performance and REM sleep stayed about the same whether I drank or not. What actually hurt my recovery showed up as a higher heart rate overnight, my resting bpm climbed on drinking nights. I'm no doctor, so I can't say whether that's typical, but I found it genuinely interesting that alcohol didn't seem to touch my ability to get good REM sleep.
Ranked by recovery delta, caffeine looks like an 8-point boost and stretching like a 7-point penalty. Neither is a real effect: both get logged in response to how I already feel.
Only alcohol shows a large, credible effect. Caffeine's apparent lift and stretching's apparent harm (✳) are both reverse causation, so the move is to act on alcohol and treat the rest as noise.
Why caffeine looks helpful: I drink coffee on normal, functional days and skip it when I am run down, sick, or hungover. So caffeine tags days I was already going to recover well; it marks a good day rather than causing one.
Why stretching looks harmful: I almost only logged stretching on nights I was already struggling to sleep, using it to wind down. So a poor recovery the next morning is that bad night of sleep showing through, not the stretching causing it, a textbook reverse-causation trap.
Where my past roles and the skills shown in this project line up with what these openings ask for.
This project is a working sample of what these roles ask for: take an ambiguous question, frame it as something testable, wrangle the data, and tell correlation from causation. At Cars & Bids I owned metrics end to end, turned a behavioral finding (auction-manager calls lifted submission completion 40%) into a shipped internal tool, and delivered analyses straight to the C-suite. Advanced SQL and data storytelling are daily work for me. I could possibly look into member engagement drivers, feature-launch measurement, or validating algorithm outputs against member outcomes.
My Lockheed work was quality and supply-chain analytics: owning KPIs, building automated reporting and governance rules, and cutting a shipping turnaround from 140 days to 4. That is the same job these roles describe, defining trustworthy metrics and translating them for operational stakeholders, and the behavior-to-outcome analysis here is lifecycle analytics in miniature. I could possibly look into manufacturing and field-quality KPIs, supplier-quality trends, or lifecycle communication effectiveness.
I have built data infrastructure from the ground up: defining database architecture and schemas, connecting them to Snowflake, running daily ETL across SAP, LDW and Excel in Alteryx, and writing data-governance rules to keep live datasets clean. That maps directly to the dbt and Snowflake modeling, testing and governance these roles run on. I could possibly look into expanding dbt model coverage, adding data-quality tests and source-freshness checks, or tightening warehouse naming and access standards.
I built this because marries two of my passions directly: data and fitness. More than ever, I've started taking my health seriously, and has enabled that. I'm incredibly passionate about its mission of putting real health insights in people's hands, something I've grown to love over the years.
Reading this back felt less like running queries and more like flipping through an old journal: the climbing was a gym I worked at and loved, the drinking was college, and the three-year gap is just life happening off-wrist.
A wearable is supposed to quantify the body. It turns out it also quietly kept a diary.