A full EDA of 66,541 international bookings for a simulated online travel agency
DreamDest is a Singapore-based online travel agency facilitating international hotel bookings across Southeast Asia. Despite consistent growth since 2010, leadership raised a critical concern: "We have years of booking data but limited visibility into what's actually driving revenue and profit." This analysis was commissioned to answer five stakeholder questions across Marketing, Finance, Product, and Partnerships.
Which origin markets and destinations drive the most bookings and revenue?
What does the typical DreamDest customer look like by age, gender, and behaviour?
How have bookings and revenue trended over time, and when did growth accelerate?
What is the relationship between discount levels and profit margins?
Which hotel ratings and stay durations are most associated with higher revenue per booking?
Seven purpose-built charts covering markets, demographics, trends, discount strategy, and hotel quality, each tied directly to a business question.
All seven markets show similar average booking values (~SGD 200), meaning volume retention, not price premiums, is the primary revenue lever. Loyalty programmes targeting these three markets have the highest ROI.
Premium product positioning and targeted campaigns for the 46–60 segment represent the highest-margin growth opportunity, they book higher-rated hotels, stay longer, and spend more per trip.
Revenue tracked volume consistently, average booking values held stable, confirming promotions did not erode unit economics during the growth period. The post-2016 shift warrants attribution analysis.
The margin deterioration is nonlinear, small discounts up to 10% have minimal impact, but the 15% threshold is where margin compression becomes material. A hard cap is the clearest policy response.
The seasonal pattern is remarkably stable across all 10 years. These windows are ideal for promotional campaigns that stimulate demand without competing with peak-season organic volume.
Premium hotel partnerships are not just a brand decision, they are a revenue decision. 4.5+ rated properties consistently outperform on per-booking revenue and attract the highest-value customer segment.
These two markets alone account for ~36% of all bookings with consistent average revenue per booking. A points-based loyalty scheme with tier benefits for repeat bookers would reduce churn in DreamDest's two most critical markets. Even a 5% improvement in repeat booking rate from these markets would add approximately SGD 240K in annual revenue at current booking values.
→ Marketing & CRMThis segment already self-selects toward higher-rated hotels and longer stays, they're the most valuable cohort without any additional intervention. A curated "Premium Escapes" product category featuring 4.5+ rated properties with concierge-style service would formalise this and allow premium pricing, directly improving overall margins.
→ Product & MarketingThe data is unambiguous: discounts above 15% erode margins nonlinearly. A hard cap for standard campaigns, with a Finance approval gate for exceptions, would protect the 18.9% average margin that characterises healthy bookings. The cap should be encoded in the promotional approval workflow, not left to campaign managers' discretion.
→ Finance & Revenue ManagementThese three months represent reliable demand troughs with no competing peak-season volume to protect. Time-limited promotions (72-hour flash sales, early-bird offers for school holiday periods) targeted at the 31–45 high-volume segment in the top three origin markets would fill the booking gap without margin risk during peak months.
→ Marketing & PartnershipsPremium hotel partnerships drive both higher booking values and better customer retention. Simultaneously, Iran, Kenya, Brazil, and Japan are underperforming relative to their tourism market size, a targeted supply and marketing effort in these four destinations could meaningfully diversify revenue without cannibalising the core Southeast Asian base.
→ Partnerships & Supply| Skill Area | Specific Techniques Used |
|---|---|
| EDA Structure | 9-stage analysis framework from business problem definition through to stakeholder recommendations |
| Feature Engineering | Date parsing, age banding, revenue/profit calculation, stay duration derivation from 24-column raw dataset |
| Statistical Analysis | Margin vs discount nonlinear relationship, segment revenue comparison, YoY growth decomposition |
| Visualisation | Matplotlib/Seaborn bar charts, time series, distribution histograms; Plotly interactive world map |
| Business Framing | All findings mapped to specific stakeholders (Marketing, Finance, Product, Partnerships, Executive) |
| Data Storytelling | Finding-first structure with quantified evidence and specific, actionable recommendations per insight |