Problem description:
Expedia is in the process of setting up its branch in the New Zealand market, called
Expedia.NZ, to compete with the existing big guns such as Black Sheep, ExperienceNZ, and
Kiwiway.
It is looking for a niche edge over the others in the already crowded travel space and wants to
capitalize on the open data on the tourists’ and visitors’ pattern to come up with key insights
on total addressable market.
It has past data for over 8 years and has information on what its competitors are doing; it has
also completed its primary and secondary research in terms of what it is that its target
audience is looking for. It knows what has already been implemented in the industry.
Its key objective is to develop its product and service offerings which would give it a head
start both in terms of reducing cost and increasing revenues.
Thus, Expedia.NZ came to us to help them out in forecasting tourist and visitor inflow to
New Zealand to tailor their offerings and price them optimally.

The data:
The historical data available to us ranges from October 2008 to November 2016 representing
the number of people based on their purpose of visit. The source for this data is New Zealand
government Immigration statistics
The charts below represent the different segments based on the purpose of the visit of the
tourists inflow to New Zealand. As can be seen from the pattern below, the segments follow
an additive seasonality with a linear trend except for the visiting friends category which is
following multiplicative seasonality pattern. Education seems to follow a bi-yearly
seasonality while Holiday/vacation, visiting friends, and Business seem to follow a yearly
seasonality.
In terms of level, the biggest chunk of tourists falls under the “holiday/vacation” category at
~150,000. The peaks across all categories fall during the holiday season from November to
January
High level description of the final forecasting method and performance
Out of the five segments, we decided to target four of the five segments which fit the criteria
for our client’s requirements.
1. Time series plot for “Visiting friends/relatives” data:
- The data set contains trend, and 12 month seasonality
2. Time series plot for “Holiday/vacations” data:
- The data set contains trend, 12 month seasonality
3. Time series plot for “Business” data:
- The data set contains no trend has 12 month seasonality
4. Time series plot for “Education” data:
- The data set contains trend has 6 month seasonality
Below chart highlights the model performance on various data series
The methods highlighted in green were used for the corresponding series for generating
forecasts after analyzing sufficient fit with the actual and predicted value for the forecasting
period and chances of overfitting.

Conclusions and recommendations:
The model accommodates the seasonality that coincides with the climatic conditions of New
Zealand. December to January is the best time to visit the Trans-Tasmania. The tourists
flocking the land of the Maori’s include the neighbors from Australia that want to escape the
dry Australian summers. Thus Expedia.NZ must introduce differential offerings and split
their services in to on season and off season. The months of November to February must
include special bundled packages for air fares and hotels along with customized tour options.
While the dry off season months must include smart discounts and a points system for the
frequent visitors who can encash the said points during the lean months.
The standard deviation of errors for April is considerably higher than for December,
indicating significant impulse tourism travel during April than a more planned travelling
during the holiday month of December. The forecast for the next five months as required by Expedia.NZ is given in the table alongside