Replaced spreadsheet chaos with intelligent automation

I led the design strategy for Priceloop, focusing on three core outcomes: making machine learning accessible to non-technical users, building trust in automation, and reducing onboarding friction to minutes rather than days.

The goal was ambitious: replace brittle spreadsheet workflows with intelligent automation, without stripping away the visibility and control that pricing managers depend on.

I set strategic direction, made trade-offs, and built trust through transparency

Throughout the project, my role went beyond product design. I set the strategic direction based on research-backed hypotheses. I made tough trade-offs, prioritizing speed to value over feature breadth. I delivered measurable improvements in adoption, activation speed, and team satisfaction.

The challenge: Pricing managers trapped between manual spreadsheets and untrustworthy AI

Pricing managers faced an impossible tradeoff. On one side were spreadsheets: familiar, transparent, but painfully manual and unscalable. On the other were black-box AI tools: powerful, but opaque and untrustworthy.

Pricing managers want independence and clarity, not another black box

I began by validating assumptions about the market and user needs. Pricing managers disliked spreadsheets, but didn't trust AI either. They wanted independence, autonomy, and above all, clarity in how optimization decisions were made.

Strategy: Made ML explainable and controllable, kept spreadsheets as familiar foundation

The key strategic decision emerged early: don't build another "smart" black-box tool. Instead, make machine learning explainable and controllable. For the first version, I intentionally kept spreadsheets as the core UI while planning guided flows later.

Early access feedback: Users confused by status labels and wanted control over tests

When I tested with the early access program, users raised recurring concerns. Status labels were confusing, editable and non-editable fields looked identical, and managers wanted more control over tests and confidence scores.

Redesign: Added clear status indicators, confidence scores, and flexible test controls

I led the redesigns to address these issues. Status indicators became clearer with tooltips, editable fields gained distinct visual cues, and stronger search and filtering were added. Time-to-first meaningful action dropped to under ten minutes, feature adoption grew by 25%.

Onboarding with template library

Reducing time-to-value from days to minutes: Template-based onboarding that connected familiar use cases to specific workflows. Result: Time-to-first meaningful action dropped to under 10 minutes.

Formula editor with natural language logic

The formula editor let users define pricing rules in familiar syntax while optionally incorporating ML suggestions. This progressive complexity approach meant users could start simple and gradually adopt advanced features.

Success metrics: <10 min activation and 25% adoption growth, but then retention dropped

A few months later, retention rates dipped. The pattern was clear: once users successfully optimized pricing, many disengaged. The lesson was critical: activation drives adoption, but retention depends on continuous value.

Introduced design-dev pairing that improved delivery speed and quality

Delivering these results required deep collaboration. I co-led the recruitment of front-end engineers and introduced design–dev pairing sessions, which dramatically reduced misalignments and sped up iteration cycles.