01 · Product analytics + mixed methods · 2026
Finding breakpoints across nine connected systems.
Tamarind Learning needed one view of a journey split across website, CRM, sales, support, and learning systems. I combined staff interviews, workflow tracing, record audits, and independent statistical validation to isolate failures leaders could act on.
- Role
- Contract UX/QA Researcher
- Methods
- Interviews, journey mapping, data audit
- Tools
- R, MATLAB, Zoho CRM, Excel
- Scope
- 9 systems · 69K records
Problem
One customer. Many partial truths.
Leadership could see activity inside individual tools, but not a reliable end-to-end path from lead to enrolled learner. Missing fields, overdue tasks, inconsistent entity links, and unclear metric logic made conversion and retention reporting hard to trust.
Research question: Where does journey evidence break, and which breakpoints create decision risk?
Method
Trace people, records, and rules together.
1. Stakeholder and workflow research
Interviewed sales and customer-success users. Followed how records were created, updated, handed off, and interpreted across nine system areas.
2. CRM population audit
Audited 351 leads, 8,446 contacts, 3,320 accounts, 241 opportunities, and 56 customers. Checked completeness, linkage, ownership, task state, and stage consistency.
3. Behavioral record analysis
Analyzed 69,000 learning records. Re-ran AI-assisted outputs in R and MATLAB, checked assumptions, and compared results against source rows.
4. Prioritization
Separated data defects, definition defects, workflow defects, and product friction. Ranked recommendations by decision risk and operational effort.
Evidence
Patterns with operational meaning.
Findings showed measurement failure and user friction were entangled. Missing CRM fields weakened acquisition segmentation; stale tasks obscured follow-up; learning events showed a sharp login-to-homework break; subscription logic overstated revenue.
Decision
Fix definitions before dashboards.
- Make Lead Type required and define allowed values.
- Introduce task-aging rules, ownership, and escalation.
- Define journey events and entity relationships before reporting conversion.
- Use 47-user cohort for targeted re-engagement research.
- Correct subscription metric logic and add validation checks.