We build the systems that don't exist yet.

• The bank was already running two BI vendor solutions (IBM and Beta Systems), both limited to analyzing internal company data only.
• Proposed and delivered a solution to analyze public sentiment across the entire population toward the brand and an upcoming campaign — something the incumbent tools could not do.
• Publicly won the tender for a customer satisfaction measurement platform, against established vendors.

12-month contract. €50,000 paid upfront. First reference in the banking sector.

• Designed a system correlating social media sentiment with Bitcoin price, achieving predictive signal on a ~72-hour horizon.
• Processed 113 TB of blockchain data, 30 million conversations, 1.5 million unique authors.
• Classified BTC wallets into 7 size groups and validated findings through cross-referencing and reverse validation.

Key finding: small holders — not large players — correlate most strongly with price movements.

• Word Cloud analysis identified an unexpectedly high frequency of the word "Amazon" in public conversation.
• Revealed that the decline in in-store sales was actually a shift to online purchasing via discounted Amazon listings — a root cause the P&G team could not identify on its own.

455,327 posts analyzed · 1.64M reach · 1-week research window.

• Traditional media coverage of the "We Believe" campaign quoted mostly negative reactions.
• Social listening told a different story: broad public sentiment was considerably more positive than the press narrative suggested.
• Reach of 43.1M, engagement of 1.25M, and 7,138 mentions tracked and segmented by sentiment across the full campaign window.

The gap between media narrative and public sentiment was the actual finding.

• Comparative sentiment and marketing-performance analysis ahead of Easter; Cadbury's identified as the leader in both categories.
• A second, unplanned finding surfaced through word-trend analysis: significant negative sentiment linked to Lindt's palm oil sourcing and habitat destruction.

20.9M reach across 745 posts — the reputational signal that mattered wasn't the one anyone was looking for.

• Led a team of ~40 people building and delivering a real-time analytics platform for election campaigns — sentiment, media networks, donations, and state-level geography.
• Platform used across multiple U.S. election cycles (including the 2016 and 2019/2020 presidential races) and regional elections.
• Accuracy example: a comparative analysis of three candidates for a 2018 Florida State House District 72 race correctly predicted the final placement of all three parties with 100% accuracy, based on just 3 days of research.

The system combined sentiment analysis, media-network mapping, and state-level geo-analytics in real time.