Case Studies
Engagements, not explanations.
Three representative projects from the last twelve months. Client names are withheld at their request. Numbers are not.
B2B SaaS
120 to 8
hours/week of manual triage
Cut support triage from 120 hrs/week to 8 hrs at a Series B SaaS platform
A Series B workflow platform was drowning in tier-1 support volume. We built a triage and response-draft system that deflected 71% of tickets and cut human review time from three full-time equivalents to one analyst working half-time.
- ticket deflection rate
- 71%
- median time-to-first-response
- 32s
D2C Consumer
+$14.20
AOV vs. holdout control
Rebuilt the merchandising engine for a D2C consumer brand, lifting AOV by $14.20
A mid-stage D2C apparel brand was running its recommendation engine on a static rules table last updated in 2022. We replaced it with a learned merchandising system trained on 26 months of clickstream and return data. Average order value rose $14.20 against a holdout control over an 8-week A/B window.
- revenue per session
- +11.3%
- return rate on recommended items
- -9.1%
Financial Services
6 days to 11 hrs
time-to-decision
Condensed underwriting decision time from 6 days to 11 hours at a mid-market fintech
A working-capital lender's underwriting queue was bottlenecked on document review. We built a document-intake and pre-decision system that compressed average time-to-decision from 6.1 business days to 11 working hours while holding default rate flat against the control cohort.
- of applications pre-decisioned without human review
- 92%
- default rate vs. control (within 0.2pts)
- flat