At Aviron (YC W21), I scaled paid acquisition from five figures/month to seven figures/month across Meta, Google, CTV, and Direct Mail. As we outgrew Meta and Google in our market, I introduced CTV to expand our top-of-funnel ceiling.

To measure what was actually working, I built a triangulated attribution approach: post-purchase surveys, household-level IP-based attribution, and geographic holdout testing. I also deployed Black Crow AI for propensity modeling, scoring site visitors 1-10 based on likelihood to convert.

Low-propensity visitors received personalized offers, recovering 8-10% of conversions we would have otherwise lost. This multi-pronged approach drove our largest Black Friday and Q4 revenue on record.

At A Thinking Ape, a freemium mobile gaming studio, I built a predictive LTV model that projected customer lifetime value at day 1, 3, 7, and 30 using historical cohort conversion curves. The model updated daily as cohorts matured, adjusting projected LTV and payback periods in real time.

We segmented by acquisition channel, which allowed us to identify which channels drove the highest-value users and cut spend on channels that weren't hitting payback targets.

I introduced a product bundling strategy at Aviron that increased accessory attachment rates from 12% to over 50%, driving roughly 10% blended AOV lift across all customers. This wasn't a simple "add to cart" prompt.

It required cross-functional work with product, merchandising, and the site team to test bundle configurations, pricing tiers, and positioning to find the right combination that maximized margin without cannibalizing standalone SKU sales.