Practical Applications of Statistics in Tourism & Business – Case Studies

1. Why Statistics Matter in Tourism & Business

Tourism operators collect massive amounts of data: daily arrivals, hotel occupancy, average spend, online bookings, and customer feedback. Statistics converts this “noise” into information that supports:

Understanding these tools is essential for any tourism manager or business analyst.

---

2. Case Study 1 – Forecasting Monthly Tourist Arrivals

2.1 Data Collection

A regional tourism board recorded monthly international arrivals (in thousands) for the past 3 years.

2.2 Time‑Series Decomposition

We decompose the series into Trend (T), Seasonal (S), and Irregular (I) components using the Classical Additive Model:

$$Y_t=T_t+S_t+I_t$$

Step 1 – Centered Moving Average (CMA) for Trend

A 12‑month centered moving average smooths the data. The CMA values (in thousands) are:

Step 2 – Seasonal Indices

Seasonal index for month m = (average of (Y / CMA) for that month) ÷ overall average of ratios.

Calculated indices (rounded):

Step 3 – Deseasonalized Data

$Y_t^{\text{*}}=Y_t/S_t$. For example, for July Year 3:

$$Y_{Jul,3}^{\text{*}}=\frac{61.7}{1.12}=55.1\text{ (thousands)}$$

2.3 Forecast for Next Year (2025)

Assume the trend continues linearly. Fit a simple linear regression to the CMA values (time = 1…36). The estimated trend equation:

$${\hat{T}}_t=12.3+0.95t$$

For month m of the next year (t = 37 + m‑1), compute ${\hat{T}}_t$ and re‑apply the seasonal index.

Example – Forecast for July 2025

• t = 43 → ${\hat{T}}_{43}=12.3+0.95(43)=53.8$ (thousands)
• Seasonal index for July = 1.12
• Forecast ${\hat{Y}}_{Jul}=53.8\times 1.12=60.3$ thousand arrivals.

2.4 Interpretation

• Trend shows a steady increase of ~0.95 k per month, reflecting growing international interest.
• Seasonal peaks (July‑August) coincide with the monsoon trekking season and festivals like Rato Machhindranath.
• The forecast helps hotels in Pokhara allocate extra staff and negotiate bulk linen contracts before the high‑season surge.

---

3. Case Study 2 – Determining Factors of Tourist Expenditure

3.1 Objective

Identify which variables most strongly influence average daily spend (NRS) of foreign tourists visiting Kathmandu.

3.2 Variables Collected (sample = 120 tourists)

3.3 Correlation Matrix

Key observations

• Strongest positive correlation: Stay ↔ Spend (r = 0.62) – longer stays raise daily spend.
• Moderate positive correlation with Festival period (r = 0.28) – festivals boost spending.
• Age and exchange rate show weak relationships.

3.4 Simple Linear Regression (Stay → Spend)

Model: $Y={\beta }_0+{\beta }_1{X}_2+\epsilon$

Using least‑squares:

$${\beta }_1=\frac{\sum (X_2-\overline{X_2})(Y-\bar{Y})}{\sum (X_2-\overline{X_2}{)}^2}=850\text{ NRS/day per extra night}$$ $${\beta }_0=\bar{Y}-{\beta }_1\overline{X_2}=4,200\text{ NRS}$$

Interpretation – A tourist staying one extra night is expected to increase average daily spend by NRS 850.

3.5 Multiple Regression (All predictors)

$$Y=3,950+720X_1+820X_2+0.45X_3+1,150X_4$$

• $R^2=0.71$ – 71 % of variance explained.
• Significant predictors (p < 0.05): Stay, Festival, Exchange rate. Age is not significant.

3.6 Business Application

A travel agency (e.g., Pathao Tours) can use the model to price bundled packages. If a package includes a 5‑day stay during Dashain, the expected daily spend rises by:

$$\Delta Y=820\times 5+1,150\times 1=5,250+1,150=6,400\text{ NRS}$$

Thus, the agency can set a premium of NRS 6,400 per tourist for the festival bundle, ensuring profitability while matching market willingness.

---

4. Sampling Techniques for Tourist Satisfaction Surveys

4.1 Types of Sampling

4.2 Worked Example – Stratified Sample for Kathmandu Visitor Survey

• Population: 150,000 foreign tourists per year.
• Strata: Europe (30 %), Asia (45 %), America (25 %).
• Desired sample size: n = 600.

$$n_{Europe}=600\times 0.30=180,n_{Asia}=600\times 0.45=270,n_{America}=600\times 0.25=150$$

Randomly select tourists from each stratum using passport data at the airport. This ensures the final satisfaction index reflects the true mix of source markets.

4.3 Presentation of Survey Results

• Frequency distribution of satisfaction scores (1‑5).
• Ogive (cumulative frequency) to identify the percentile at which 80 % of tourists are “satisfied” (score ≥ 4).
• Box‑and‑whisker plot to compare satisfaction across regions.

Interpretation – Asian tourists show the widest spread (more variability), suggesting targeted service improvements.

---

5. Comparative Summary of Statistical Tools for Tourism Decision‑Making

---

6. Real‑World Applications

6.1 eSewa & Khalti – Transaction Volume Forecasting

Both e‑payment platforms experience seasonal spikes during festivals (Dashain, Tihar). They apply time‑series forecasting similar to the tourist arrivals example to allocate server capacity and schedule promotional discounts.

6.2 Daraz – Demand‑Driven Inventory

Daraz uses regression analysis to predict product demand based on variables such as search volume, price discounts, and holiday periods. The same regression logic helps a tour operator forecast bookings for a new trekking package.

6.3 Ncell – Customer Churn Prediction

Ncell analyses call‑detail records with correlation coefficients to identify factors (e.g., low data usage, high bill amount) that correlate with churn. Tourism firms can adopt the same approach to predict which repeat visitors are likely to stop using a travel agency’s services.

---

7. In the real world

• Daraz’s “Flash Sale” algorithm uses a time‑series model to predict the optimal discount level that maximizes sales without eroding profit margins. The model incorporates seasonal indices similar to those calculated for tourist arrivals.
• NTC’s network traffic monitoring applies standard deviation to detect abnormal spikes that may indicate network attacks; tourism websites (e.g., Booking.com Nepal) use the same technique to spot sudden surges in booking requests during a festival, prompting auto‑scaling of cloud resources.
• A Nepali bank’s loan‑interest calculator employs a regression equation where the dependent variable is the interest rate and predictors include credit score, loan amount, and prevailing market rate—mirroring the multiple regression used to estimate tourist daily spend.

---

8. Worked Example Tied to a Real Situation

Scenario – A boutique hotel in Pokhara wants to set a seasonal room rate for the monsoon trekking season (July‑August). Historical data shows:

Step 1 – Compute Coefficient of Variation (CV) for each month:

$$CV=\frac{SD}{\text{Mean}}\times 100\%$$

Assume SD of daily rate for July = 800 NRS, for August = 750 NRS.

$$C{V}_{Jul}=\frac{800}{6,800}\times 100=11.8\%$$ $$C{V}_{Aug}=\frac{750}{6,500}\times 100=11.5\%$$

Step 2 – Decide Pricing Strategy

• Lower CV indicates more stable demand; the hotel can raise rates with less risk of unsold rooms.
• Since July’s CV is slightly higher, set a modest increase of NRS 200 over the historical average, yielding NRS 7,000.
• For August, apply a larger increase of NRS 300 (NRS 6,800) because demand is still high but variability is marginally lower.

Result – Projected revenue for the two months:

$${\text{Revenue}}_{Jul}=7,000\times 0.92\times 30=NRS193,200$$ $${\text{Revenue}}_{Aug}=6,800\times 0.88\times 30=NRS179,520$$

The hotel can compare these figures with the baseline (no price change) to justify the new rates.

---

9. Advantages & Disadvantages of Key Techniques

---

In the real world

• eSewa & Khalti use time-series forecasting to predict transaction volumes during festivals like Dashain and Tihar, ensuring sufficient server capacity and cash reserves. For example, during Dashain, transaction volumes spike by 30–40% compared to regular days, requiring dynamic load balancing.

• Daraz applies multiple regression analysis to forecast demand for products like trekking gear and festival-related items. For instance, during the monsoon season, demand for raincoats and hiking boots increases by 25–30%, allowing Daraz to optimize inventory levels and reduce stockouts.

• Ncell leverages customer churn prediction models (using logistic regression) to identify subscribers likely to switch operators. For example, if a customer’s call drop rate exceeds 5% and data usage drops by 20% in a month, Ncell proactively offers promotions to retain them.

Exam tip

• Know the formulas for mean, median, mode, range, quartile deviation, standard deviation, CV, correlation coefficient $r$, simple linear regression (slope & intercept), and the additive time‑series model. Memorise the steps; the exam often asks for a complete calculation (e.g., “Compute the seasonal indices for the given data”).
• Practice interpreting graphs: be able to read a line chart of tourist arrivals, identify peaks, and explain what they represent (festival, weather, etc.).
• Remember the hierarchy: descriptive → dispersion → relationship → forecasting. Many short‑answer questions ask you to state why a particular tool is chosen for a tourism problem.
• Use the “5‑point checklist” when tackling a case study: (1) Identify the decision problem, (2) Choose the appropriate statistical tool, (3) Show calculations step‑by‑step, (4) Interpret the numeric result in tourism terms, (5) State a practical recommendation.
• Time management: allocate ~5 min for data cleaning, ~10 min for calculations, ~5 min for interpretation. Write the final answer in clear bullet form; examiners reward neat, structured presentation.