Product Case · Data Analysis & Forecasting

A Glasgow taxi company had trip data — and no clear view of demand patterns, pricing anomalies, or fleet allocation.

This is a Product Strategy Growth problem.

Academic data science project completed in 2024 with Kin Karta Consulting at the University of Strathclyde Business School, United Kingdom, using Haggis Hopper's real historical trip data across Glasgow.

Demand wasn't the mystery — Haggis Hopper already had the trip volume. What was missing was where fleet capacity should sit by hour and postcode, and whether the pricing riders were charged actually held up against the data. I turned raw trip records into live dashboards and a forecasting model, so fleet deployment and pricing could be decided from evidence instead of instinct.

Streamlit analysis app Tableau dashboard

Who Haggis Hopper's operations team Decides where to position the fleet and how trips should be priced across Glasgow.

What See where and when demand happens — and act on it Turn raw trip records into a live view of demand, revenue, and fleet performance.

Why Demand mapped by hour and postcode Two live tools let stakeholders see where demand concentrates, instead of reading a static report.

$ Impact +3.8% revenue vs. current flat pricing Scales with peak-hour trip volume — ≈£1,800–£2,200/month at the current run rate.

Two things that were invisible

Fleet allocation was a guess

Demand across Glasgow shifts by hour and postcode, but without mapped patterns, vehicles couldn't be positioned where and when riders actually needed them.

Pricing wasn't checked against the data

Fares hadn't been validated against actual trips, leaving the fare-distance relationship and pricing anomalies unexamined.

What we changed

Mapped demand by hour and placeTemporal and geospatial patterns surfaced hotspots and under-served postcodes, not just a citywide average.
Built live dashboards, not static reportsTableau dashboards let stakeholders monitor demand, revenue, and fleet performance themselves, on demand.
Tested whether demand is predictableTime-series and machine learning models forecast short-term demand, moving allocation from reactive to planned.
Put pricing under the microscopeA dedicated pricing deep-dive checked tariff logic, fare-distance correlation, and flagged anomalies.

What it moved

3,068 trips

Modeled in the pricing deep-dive

February 2024 · tariff inference, fare-distance correlation, anomaly detection

8 demand views

Temporal and geospatial patterns mapped

hour, postcode, duration, fare — not a single summary chart

2 live tools shipped

Tableau dashboard + Streamlit app

built for stakeholders to explore, not just read a report

The business case

+3.8%

Projected monthly revenue lift from peak-hour pricing

≈£1,800–£2,200/month · fares were flat by time of day despite peak trips taking 60% longer to complete

11% of trips

Flagged in a cash-reconciliation gap

342 trips · amount charged ≠ amount recorded, concentrated entirely in cash/unlogged payments

4 zones

Demand zones segmented for a prioritized pricing rollout

clustered by location, trip volume, revenue/km, and peak share

Fares were fully explained by trip distance and duration — a rigid, formula-only system with no mechanism to price for demand. That gap, proven statistically, is what turned into the peak-pricing recommendation above. Full pricing deep-dive, anomaly detection, and recommendations: pricing analysis report →

Supporting visuals

Every chart and dashboard view behind the numbers above, pulled directly from the full analysis.

Quantitative analysis (Python)

Trip duration trends by weekday and hour — Duration Analysis

Drop-off postcode demand by hour — Drop-off Postcode vs Hour

Pickup demand by hour of day — Pickup Hour-of-Day

Trip counts per postcode — Trip Counts by Postcode

Median fare amount by hour and day of week — Fare and Tip Analysis

Outlier detection plot — Outlier Analysis

Interactive Tableau highlights

Demand temporal patterns dashboard — Demand Temporal Patterns

Trip volume analysis dashboard — Trip Volume Analysis

Passenger behavior dashboard — Passenger Behaviour