Demand forecast in tourism is of great economic value both for the public and private sector. Any information concerning the future evolution of tourism flows is of great importance to hoteliers, tour operators and other industries concerned with tourism.
Our Client is a Tours and Travel agency who wants to efficiently plan its offerings according to different segments for the upcoming year and ensure adequate capacity and infrastructure.
Tour and Travel agents in the UK (eg. Thomas Cook) need to plan demand while making their advance bookings of bus, flights, hotels etc. that they can offer to their varied customers (business, Leisure, family etc.) as packages. Knowing the right demand can really help the travel agents maximize their revenues by not losing out on potential customers and minimize wasteful costs by not making unnecessary bookings in advance. Through this report, we are designing a model to forecast the number of tourists traveling to the UK with different purposes and from various parts of the world (i.e. Europe, North America or Other regions). This model can also be leveraged to price their packages, keeping in mind advance bookings made and demand forecasts to squeeze out the maximum margins from the limited resources they may have.
The data obtained includes the number of monthly inbound travellers to UK from regions travelling for different purposes from Jan’12-Sep’18 [Total: 81 months].
Source: Data is collected by UK government every quarter (source: www.ons.gov.uk) and is updated every quarter, leading to a lag of 3 months in data reporting.
Yearly Seasonality and trend can be observed in all the time series and 3 different peak seasons can be identified based on purpose of travel:
- Business travellers: Apr-Jun and Sep-Nov
- Leisure travellers: May to Sep
- Visiting friends/family: Jun-Sep
Surprisingly, growth in overall inbound travel in the UK is observed in 2017; it is probably due to currency depreciation post Brexit that made inbound travel to the UK cheaper.
There is a sudden dip in 2018 as well, probably because of following 2 reasons:
- People who would have originally travelled in 2018 already travelled in 2017
- Brexit to complete in Mar 2019, hence people preferring less to travel to UK for work purposes
Forecasting description: To provide a year’s forecast of the demand, we ran several analytical models on different series namely: Seasonal Naïve, Holt Winter’s (Additive & Multiplicative), Linear regression using dummy variables and exchange rate (to capture Brexit) and Ensemble models. For overall inbound UK travel, the best performing method is Multiplicative Holt-Winters with a MAPE of 5.7% in validation, as compared to seasonal Naïve MAPE of 5.8, hence performing slightly better than the seasonal naïve model. Holts’ Winter with high levels of Alpha and Beta was really able to capture the effect of Brexit well in the forecast. Linear Regression models for visitors travelling with different purpose were finally selected after comparing with seasonal Naïve benchmark and error plots. Through research we also found that COGS comprise 80% of revenues of travel agents, thus cost of over forecasting is much greater than the cost of under-forecasting.