CHALLENGE: A B2B sales team faced inconsistent closes, poor follow-up, and emotionally reactive selling under pressure. Leadership saw a need to improve emotional intelligence (EI).

SOLUTION: We used an AI tool to analyse sales calls for tone, pace, empathy, and interruptions. Reps received personalised feedback and targeted EI micro-coaching on impulse control, empathy, and stress management. Weekly call reflections reinforced learning.

IMPACT: 93% of reps reported higher confidence and emotional awareness

Close rate ↑ 29%, Follow-up rate ↑ 42%, Sales cycle ↓ 18%

TECH STACK: AI call analysis, Python, EQ coaching app
METHODS: Voice emotion analysis, behavioural feedback, EI coaching

CHALLENGE: An airline aimed to reduce customer churn, improving on an existing model with 75% precision and recall.

SOLUTION: We used an ensemble approach: four logistic regression models captured different behavior patterns, feeding into a neural network for final prediction.

IMPACT: The new model achieved 93% precision and 97% recall on future data, enabling earlier, more accurate retention efforts.

TECH STAC: R, Scala Spark | METHODS: Neural Networks, Regression

CHALLENGE: As a scale-up grew from 50 to 150 staff, communication friction, low emotional intelligence, and declining client satisfaction emerged. Leadership needed better insight into internal dynamics.

SOLUTION: We deployed an AI sentiment analysis tool across internal communications—email, chat, and meetings—to detect emotional tone and engagement. Anonymised insights were shared via dashboards, paired with targeted EI coaching for managers and teams.

IMPACT: The initiative led to stronger collaboration, fewer conflicts, and better client outcomes: Collaboration ↑ 41%, Conflicts ↓ 60%, Client satisfaction ↑ 23%. We also noticed a positive shift in client reviews, and 87% of staff reported improved emotional awareness.

TECH STACK: Python, GPT-based sentiment engine, Power BI
METHODS: Sentiment analysis, EI coaching, real-time feedback loops

CHALLENGE: The client faced serious issues with actively disengaged staff who disrupted team morale and performance—one case even involving police intervention.

SOLUTION: We conducted confidential “Personal Interventions” with the individuals, resolving deep-rooted resentment through one-on-one coaching and hypnotherapy. Each staff member either re-engaged or resigned—eliminating toxicity without conflict.

IMPACT: Team harmony was restored, culture improved, and performance increased. The interventions acted as “emergency acupuncture,” preventing further organisational damage.

TOOLS: Human-led intervention, Hypnotherapy
METHODS: Anonymous coaching, behavioural turnaround strategy

CHALLENGE: The client needed a new loyalty tiering framework aligned with updated KPIs, shifting focus from customer profit to profitability. Stakeholders also disagreed on segmentation definitions, requiring a more flexible, data-driven approach.

SOLUTION: We built a CLV-based ranking system using Bayesian modelling to estimate lifetime value, even for new customers—improving as more data became available.

IMPACT: Stakeholders could dynamically segment and analyse customers, leading to better personalisation, insights, and a more strategic loyalty model.

TECH STACK: R, Python
METHODS: Bayesian Statistics

CHALLENGE: Globalisation and COVID-19 brought an urgent need to build leadership team cohesion virtually, as in-person facilitation came to a halt.

SOLUTION: Thomas and his partners designed deep, interactive online sessions that encouraged vulnerability and trust. Surprisingly, virtual settings allowed for even deeper emotional breakthroughs when facilitated intentionally.

IMPACT: The approach became a core offering. Since 2021, hybrid delivery—blending online and in-person sessions—has become the standard for fostering leadership team cohesion across global clients.

TOOLS: Zoom, Miro, custom collaboration tools
METHODS: Virtual facilitation, emotional safety, hybrid transformation design

CHALLENGE: The client aimed to boost email engagement by personalizing send times, moving beyond static scheduling methods.

SOLUTION: We built a Bayesian Send Time Optimizer (STO) using Thompson Sampling to balance proven send times with exploration of new options. The model continuously refined optimal delivery windows.

IMPACT: The STO delivered a 50%+ higher clickthrough rate vs. historical best times, and 2x the rate of random scheduling.

TECH STACK: R | METHODS: Thompson Sampling, Multi-Armed Bandit

CHALLENGE: A new procurement leader was rejected by her team after sabotage by her predecessor. Her boss avoided involvement, leading to workflow collapse and union escalation.

SOLUTION: We ran fact-finding interviews, EQi assessments, and coaching for key leaders. A full-day offsite with mediators enabled open confrontation, resolution, and the creation of team rules with a WHWNH accountability protocol.

IMPACT: The team rebuilt trust and transformed its culture. Purchase lead time ↓ 22%, Inventory ↓ 18%, Sick leave ↓ 68% and Zero staff turnover the following year.

TOOLS: Interviews, EQi assessment, mediation
METHODS: EI coaching, team confrontation, cultural reset

CHALLENGE: Fragrance choice is deeply personal. The goal: build a “Spotify for perfumes” — recommending scents based on personality traits and past ownership.

SOLUTION: We built five distinct recommendation engines plus a random model for comparison. Models were tested on 150 consumers (UK & US), evaluating likeability, wearability, and purchase intent.

IMPACT: Three models outperformed the control, effectively matching users with fragrances they were more likely to enjoy.

TECH STACK: Python | METHODS: Random Forest, Association Rules

CHALLENGE: Automotive dealers lacked insight to prioritise leads, risking lost sales and inefficient outreach.

SOLUTION: We built a machine learning–based propensity scoring system (1–3 scale) tailored by region and contact type. Historical data was used to predict purchase likelihood, validated through simulated past scenarios.

IMPACT: High-scoring leads were 5x more likely to convert. The latest model improved performance by 16–25% over the first version, boosting dealer efficiency and sales.

TECH STACK: R, Python | METHODS: Random Forest, Experimental Design

CHALLENGE: Evaluate live chat impact during peak events, identify pain points, assess agent performance, and prove chat value.

SOLUTION: An NLP model analyzed 40K chats and 500K lines of text, combining transcripts, surveys, and transactions. Real-time insights were shared with leadership.

IMPACT: Chat sentiment drove a 30% rise in conversions. Agent reports improved performance; strategic findings guided decisions.

TECH STACK: R, Python
METHODS: Text mining techniques

CHALLENGE: A client’s customer database contained duplicate records with different IDs, making accurate analytics and insights difficult.

SOLUTION: We developed an algorithm that used email, physical addresses, and preferences to identify and merge duplicates. It preserved the most accurate details in a unified master record.

IMPACT: 11% of the database was successfully deduplicated with no known errors. This led to cleaner data, better insights, and more effective marketing.

TECH STACK: Python
METHODS: Association Rules, tf-idf

CHALLENGE: Gain audience insights across four key markets to shift from anonymous followers to an engaged community.

SOLUTION: We aligned research goals in an on-site session, then used machine learning to segment audiences and surface behavioural insights via an interactive dashboard.

IMPACT: Identified six key segments, enabling targeted marketing. Insights were featured at Advertising Week Europe to inform Gen Z engagement strategies.

TECH STACK: R, Sisense
METHODS: PCA, Self-Organising Maps, Hierarchical Clustering

CHALLENGE: Manual, spreadsheet-based reporting was slow, error-prone, and time-consuming.

SOLUTION: We built an automated data pipeline and a custom Power BI dashboard for real-time, unified customer insights.

IMPACT: Reporting time dropped from hours to minutes. The client gained faster decisions, improved scalability, and reduced manual effort.

TECH STACK: Python, SQL, Power BI, AWS Lambda
METHODS: Simple statistics

CHALLENGE: Unpredictable call volumes led to long wait times and idle staff. The client needed accurate forecasts to optimise staffing.

SOLUTION: We built a SARIMAX model to predict daily call volumes based on seasonality and campaign data, delivered via an interactive Excel calendar.

IMPACT: Improved staffing efficiency, reduced wait times, and enabled a data-driven approach to call volume management.

TECH STACK: R, Excel
METHODS: Time Series Analysis (SARIMAX)