HOTEL ROOM PRICE DETERMINATION

BASED ON DYNAMIC PRICING MODEL

USING NONLINEAR PROGRAMMING
METHOD TO MAXIMIZE REVENUE
Muhammad Fadly Ari Yanuar Ridwan Mohammad Deni Akbar
Fakultas Rekayasa Industri Fakultas Rekayasa Industri Fakultas Rekayasa Industri
Telkom University Telkom University Telkom University
Bandung Indonesia Bandung, Indonesia Bandung, Indonesia
mmdfadly@gmail.com ariyanuar@telkomuniversity.ac.id denimath@telkomuniversity.ac.id

Dynamic pricing strategy is a pricing strategy that is often In Figure 1.1, it can be seen the ratio of the ratio of hotel
applied by hotels to maximize their income. The purpose of this revenue to prisoners and desired hotel hotel revenue targets
study is to provide a pricing policy proposal in accordance with in recent years. Seen there are a number of years where
dynamic pricing models. This study provides a dynamic pricing hotel targets have not been reached. Like in 2015 and 2017
model that has been adapted to the problems that exist in hotels
with various types of rooms. This research consists of three
where the comparison between targets and income is quite
stages, the first phase is predicting future demand. The second far apart.
stage applying revenue management tools, namely dynamic
pricing, to model the effect of prices on demand. Prices change
dynamically based on the number of requests available. The third
stage is using a nonlinear programming approach to maximize
revenue. Forecast parameters and dynamic pricing models are
estimated using historical sales data for one of the hotels in
Bandung, West Java, Indonesia. The results suggest a pricing
policy for superior and deluxe room types that can increase
revenue by 27% more than the existing pricing policy. The
proposed pricing policy is able to complete the method void in Figure 1. 2 Demand Superior
determining prices. In addition, the results of this study provide
optimal room rates every day along the planning horizon.
Keywords— dynamic pricing, revenue management, hotel
room pricing policy

I. INTRODUCTION
In recent years, the number of hotels that grew in Indonesia
is rising. This is due to the increasing demand of people for
temporary housing. Because these hotels are very useful for
people who are traveling on business or just for vacation.
Due to the increasing needs of these, Narapati Hotel requires
more attention to price management in order to get the
maximum benefit in each period. The size of the profits
obtained by the company can not be separated from the
company's strategy in determining the price of services to be
provided to consumers. When the company is not good
enough in determining the price that will be offered, then
most likely the benefits to be gained are not optimal. Policy
in determining service prices is a basic strategy in daily
operations for the company, because with prices we can
determine various objectives whose function is to increase
the competitiveness of companies, one of which is to
increase revenue.
Figure 1.1 Actual VS Target Revenue
Figure 1. 3 Demand Deluxe
Figures 2 and 3 show the occupancy level of "superior" and
"deluxe" room types in January 2017-2018. It can be seen
that in certain months the demand increases which is called
high season such as the school holidays month and in certain
months the demand falls called the low season. When
demand is going down empty rooms still cost a lot such as
electricity, cleaning etc. while the room does not provide
income.
The hotel uses a fixed or static pricing strategy, so that
under any circumstances the price given to customers will
remain the same. Prices that are fixed are considered less
relevant for the hotel industry because it will eliminate the
company's opportunity to get a higher income. Therefore the
company must be careful and precise in determining the
price of the room, if the price is too low then the company
will lose the opportunity to get higher profits, whereas if the
room price is too high, then the company will lose
customers, causing many vacant rooms with operational
costs which still exists.
The last few years RMS using dynamic pricing has proven
to be able to increase revenue in the hotel industry [1].
Although in reality the company's target is to get the

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

maximum profit, but the company cannot sell the room price Analysis | Regression) will be used. After getting a trend
at a high price all the time. Prices given must be dynamic value of 0 and level 0, then a deseasonalized demand value
and change daily according to forecasting from customer will be searched after the regression with the formula:
demand.
To solve this problem, RMS will be used by using a (2.2)
dynamic pricing model which will then be solved by
Ds = Deseasonalized demand after regression
nonlinear programming methods to get the right price and
h = day “h”
optimal and dynamic nature that adjusts to demand to
After that find the initial seasonal factor with:
maximize hotel revenue.
(2.3)
II. LITERATURE REVIEW & METHOD
= Seasonal Factor Initial
A. Revenue Management
After getting the Initial Seasonal Factor value, then the
According to Kimes & Wirtz (2003) revenue management seasonal factor value will be searched using the formula:
(RM) is a collection of tools and actions taken to achieve
optimal profit levels and crude profits by offering the right SF = (2.4)
product to the right consumer through the right distribution
channel and the right price with the right communication SF = Seasonal Factor
[6]. Similar like Kanban in factory [10]. r = Seasonal Cycle (number of repeated p)
After getting the seasonal factor value, then it will look for
B. Data Pattern levels and trends in the next period with the formula:
According to Hanke & Wichern (2009) [4], data patterns are L = α(D/SF)+(1-α)*(Lh-1+Th-1) (2.5)
divided into 4 types, namely horizontal patterns, seasonal T = β(Lh –Lh-1)+(1-β)*(Th-1) (2.6)
patterns, trend patterns and cyclical patterns. α = Level smoothing constant
1. Horizontal pattern β = Trend Smoothing Constant
Occurs if the data fluctuates between a constant average L = Level
value (stationary to the average value). T = Trend
2. Seasonal pattern After getting the level and trend values, then will look for
Occurs if a data is influenced by seasonal factors such as seasonal factor values for the period to be predicted by the
quarter of the year, monthly or days of a particular week. formula:
3. Trend patterns SFf = γ(Dh-t / Lh-t)+(1-γ)*(SFh-t) (2.7)
Occurs if there is a long-term increase or decrease in data. SFf = Seasonal Factor for forecasted periods
4. Cyclic data patterns In this method 3 smoothing parameters are needed, namely
Occurs if the data are affected by long-term economic alpha, beta and gamma which can be valued from 0 to 1, so
fluctuations such as those related to the business cycle. that many combinations can be tried to get the most optimal
value.
C. Winter’s Exponential Smoothing After getting the value, then you can predict the future
This method is suitable when the data has a seasonal demand by using the formula:
component [5]. The first thing to look for is level 0 and F = (La+(h – ha)*Ta)*SFf (2.8)
trend 0 by way of regression deseasonalized demand to get F = Forecasted demand
level 0 and trend 0 which will then be used as a forecast a = Last value
component. Then after regressing the intercept value will be
used as a level and X Variable will be used as a trend. D. Dynamic Pricing
To find the value of deseasonalized demand with odd In revenue management [5], dynamic pricing is a pricing
periods, you can use the formula: strategy in which companies adjust prices for products and
services as a function of demand at different times. While in
 p −1 
t+
the science of supply chain management, dynamic pricing is

 2  a tactic for determining prices varying from time to time that
Dt =  D / p (2.1) requires the presence of different customer segments, where
 p −1  some customers are willing to pay higher prices for these
t− 
 2  products. The idea of dynamic pricing is to model the effect
Information: of product prices at different times on the demand for a
product. The dynamic pricing model is explained by the
= Deseasonalized Demand equations below.
R= (2.9)
D = Hotel room demand Subject to:
p = Period
t = Number of days in one period
After getting deseasonalized demand values, then these Where:
values will be regressed to get a trend value of 0 and level 0. (2.10)
To regress that value, Microsoft Excel software (Data | Data
 Based on the results of historical data patterns, it was found
(2.11)
that the data patterns have seasonal data trends. For this
reason, the Winter’s Exponential Smoothing method is used
da,l = Predicted demand value (demand expectation) to predict room sales. Forecasting with the Winter's
a = First night of the stay (date) Exponential Smoothing method uses 3 parameters by
L = Length of stay selecting the values α, β, and γ that produce the minimum
Pnominal = Hotel room price (Average prices of rooms sold) MSE and MAD values, the value of each parameter with the
l = night l minimum MSE and MAD values is α = 0.1, β = 0.1, and γ =
e = Elasticity between price and demand 0.
E. Nonlinear Programming Demand forecasting will be done for the next 1 month and
based on the length of stay and then summed to get the
Research using heuristics methods has been done for demand forecast on that day, for example will be forecasting
optimization problems, such as travel time optimization [3], for the first day with L = 1 day.
time delay [8], and reducing product blocking [7]. The First of all, we will find the value of deseasonalized demand
method used in this research is nonlinear programming. according to equation 2.1, which is:
According to Rama [9], the general form of non-linear
programming problems is to determine x = (x1, x2, ... xn) so
as to achieve the goal of:
Maximize/Minimize :
The result is = 1.94521
Constraints : and Then the normalized demand will continue to be searched
: from day 183 to day 548. Then the value will be regressed
with Microsoft Excel (see appendix) to produce T0 (X
F. Conceptual Model Variable) and L0 (Intercept) values. The values obtained are:
T0 = -0.000148757
Nominal Price of the
Hotel
L0 = 1.905512119
Based on equation 2.2 then the first day's Ds value will be
Room Demand
Room Price at Night l
found using the formula:
Price and Demand
Dynamic Pricing Model
Elasticity

Room Price Reserved Room at Night

l
)
Ds value obtained is: 1.905363362
Hotel Capacity
Then equation 2.3 is used to find the initial seasonal factor
value ( ).
Length of Stay Forecasted Demand

Historical Demand

Figure 2.1 Conceptual Model value obtained: 3.673839929

In this research, the first step taken is observation in order to
identify pricing in this company. Collecting data needed for The next step is to find the Seasonal Factor (SF) value for
research and finding solutions is done after making the first day with the formula:
observations. The solution can be implemented with a
Dynamic Pricing Model approach with several SF =
characteristics that suit the circumstances. These data are
room demand data in the previous period, data on room
SF =
prices that have been sold, number of rooms in the hotel,
average length of stay, price class, and future demand
SF =
forecast.
Then the thing to do next is to make a mathematical model The SF values obtained are: 3.187492237
using the data that has been obtained which consists of The next step is to determine levels and trends using the
objective and limiting functions in accordance with the formula in equations 2.5 and 2.6.
characteristics of the problem at hand. L = 0.1(7/3.187492237) + (1-0.1) x (1.905512119 + (-
0.000148757))
III. DATA PROCESSING & DISCUSSION
The level value (L) obtained is: 1.934435404
A. Data Processing T = 0.1(1.934435404 - 1.905512119) + (1-0.1) x (-
0.000148757)
In this study, researchers determine the price of rooms per The trend value (T) obtained is: 0.002758447
day that adjusts to the predicted demand. researchers Then the calculation that has been carried out until this stage
consider sales data and historical prices for 24 months or 2 is continued to the next day as long as the data is available.
years. the first stage is determining future demand by using In this case the data held is 2 years or 730 days, therefore
winter’s exponential smoothing. The equation below will the above calculation is done until the 730th day.
show the process of forecasting the demand.
After calculating these values, SFf will be calculated, which Table 3.2 Forecasted Deluxe Demand
is seasonal factor for forecasting period. The calculation will Day D a,1 D a,2 D a,3 D a,4 D a,5
be carried out for day 731 with the formula contained in 1 0.994522 0.5 0.5 0 0
equation 2.7, namely:
SFf = γ(Dh-t / Lh-t)+(1-γ)*(SFh-t) 2 0.997503 0 0 0 0
SFf = 0(5 / 1.845271454 )+(1-0)*( 3.187492237) 3 0 0 0.5 0 0
The value obtained is equal to: 3.187492237 4 0.501733 0 0 0 0
The next step is to predict demand for day 731 by using the
formula in equation 2.8 as below: 5 0 0 0 0 0
F = (La+(h – ha) x Ta) x SFf 6 7.151195 0 0 0 0
F = (1.838321051+(731 – 730) x ( -0.00399512)) x 7 0.506207 1.989829 0 0 0
3.187492237
8 3.590616 0 0 0 0
The result of forecasting demand for 731 days or the first
day of forecasting after a series of processes is 5.8469, this 9 0.509191 0 0 0 0
value is the first day's room demand and staying for 1 day 10 1.544363 0 0 0 0
(D1,1). 11 0.512175 0 0.963582 0 0
Table 2.1 and 2.2 will show the forecasted demand for
Superior and Deluxe Room. 12 4.158895 0 0 0 0
13 1.042746 0 0 0 0
Table 3.1 Forecasted Superior Demand
14 1.033306 0.5 0 0 0
Day D a,1 D a,2 D a,3 D a,4 D a,5 D a,6 D a,7
15 0.518146 0 0 0 0
1 5.8469 1 0 0 0 0 0
16 1.571454 0 0 0 0
2 0.974786 0 0 0 0 0 0
17 0.521133 0 0 0 0
3 3.397543 0.80504 0 0 0 0 0
18 1.580489 0 0 0 0
4 3.882739 0 0 0 0 0 0
19 1.060887 0 0 0 0
5 1.923262 1.610911 0 0 0 0 0
20 3.191736 0.994917 0 0 0
6 2.899745 0 0 0 0 0 0
21 0.527108 0 0 0 0
7 1.92906 1.611739 0 0 0 0 0
22 1.069963 0 0 0 0
8 2.399234 0.5 0 0 0 0 0
23 2.676076 0.5 3.854678 0 0
9 1.92085 0 0 0 0 0 0
24 0 0 0 0 0
10 8.139232 0 0 0 0 0 0
25 0.533087 0 3.854735 0 0
11 4.296518 0 0 0 0 0 0
26 3.24621 0 0 0 0
12 1.424496 0 0 0 0 0 0
27 1.085098 0 0 0 0
13 1.904423 0.807101 0 0 0 0 0
28 2.176251 0 0 0 0
14 3.318677 0 0 0 0 0 0
29 1.091154 0 0 0 0
15 1.896206 0 0 0 0 0 0
30 0 0 0 0 0
16 0 0 0 0 0 0 0
31 1.097212 0 0 0 0
17 4.241137 0.5 0 0 0 0 0

18 1.883875 0 0 0 0 0 0
B. Optimization Using Dynamic Pricing
19 2.342901 0 0 0 0 0 0
After the expected demand have been forecasted, the next
20 0.462125 0 0 0 0 0 0 thing to do is optimize the price using the dynamic pricing
21 0.935769 0 0 0 0 0 0 model. below will show the optimized price to escalate the
demand so the income will increase.
22 1.867424 0 0 0 0 0 0

23 0.931655 0 0.45 0 0 0 0

24 0.929597 0 0 0 0 0 0

25 0.92754 0 2.221941 0 0 0 0

26 1.850963 0.809715 0 0 0 0 0

27 1.846846 0 0 0 0 0 0

28 0.921364 0 0 0 0 0 0

29 1.838611 0 0 0 0 0 0

30 3.668983 0 0 0 0 0 0

31 0.915186 0 0 0 0 0 0
 Table 3. 3 Room Price and Demand Before Optimization
l P1l O1l l P2l O2l
1 500000 7 1 600000 2
2 500000 2 2 600000 2
3 500000 5 3 600000 1
4 500000 5 4 600000 2
5 500000 4 5 600000 1
6 500000 5 6 600000 8
7 500000 4 7 600000 3
8 500000 5 8 600000 6
9 500000 3 9 600000 1
10 500000 9 10 600000 2 Figure 3.2 Solver Parameter for Deluxe Room

11 500000 5 11 600000 2
Figure 3.1 and 3.2 is a display of the solver in the Microsoft
12 500000 2 12 600000 6 Excel application. In the set objective column, the objective
13 500000 3 13 600000 3 function is revenue from superior rooms obtained from the
total multiplication between price (Pl) and number of rooms
14 500000 5 14 600000 2
booked (Ol). In the changing variable is the decision
15 500000 2 15 600000 2 variable that is Pl and the constraints part is the limitation in
16 500000 0 16 600000 2 the dynamic pricing model that is Pl ≥ 250000 for Superior
17 500000 5 17 600000 1 Room and Pl ≥ 400000 for Deluxe Room because this
value is the lowest price that can be given by the company
18 500000 3 18 600000 2
for Superior and Deluxe rooms and the number of rooms
19 500000 3 19 600000 2 booked does not exceed the capacity of available rooms.
20 500000 1 20 600000 5 The hotel capacity is 26 rooms for Superior Room and 12
21 500000 1 21 600000 2
rooms for Deluxe Rooms.
22 500000 2 22 600000 2 Table 3.4 Optimization Result
23 500000 2 23 600000 8 l P1l O1l P2l O2l
1 250000 26 565964.6 3
24 500000 2 24 600000 5
2 368700 5 565913.3 3
25 500000 4 25 600000 9 3 250000 17 583022.1 2
26 500000 5 26 600000 8 4 250000 19 583035.6 2
5 250000 15 591537.7 1
27 500000 5 27 600000 5
6 250000 19 471390.1 12
28 500000 1 28 600000 3 7 250000 15 556729.4 4
29 500000 2 29 600000 2 8 250000 18 502018.6 8
9 347896 6 591402.4 1
30 500000 4 30 600000 0 10 279753 26 573722.3 2
31 500000 1 31 600000 2 11 250000 18 574715 2
12 423027 2 510355.9 7
Table 3.3 shows the number of rooms booked on the ninth 13 321251 8 565617.9 3
night for Superior Room (O1l) and Deluxe Room (O2l) type 14 250000 16 573929.5 2
if the hotel uses a static pricing system. It can be seen in 31 15 392575 4 582728.5 2
16 500000 0 573255.8 2
days that 107 rooms have been sold for Superior Room and
17 250000 19 591200 1
101 rooms sold for Deluxe Room.
18 350387 6 573100.1 2
19 360046 5 582014.8 2
20 476905 1 527332.6 6
21 451525 2 573921.3 2
22 394534 3 581859.8 2
23 423134 3 490497.7 11
24 423254 3 537178.6 7
25 250000 14 474975.7 12
26 250000 20 488921.4 11
27 250000 20 527083.8 8
28 452327 2 562784.8 3
29 396481 3 581497.6 2
30 250000 15 600000 0
31 452670 2 581394 2
Figure 3.1 Solver Parameter for Superior Room

Table 3.4 shows the room price and the number of Superior
and Deluxe rooms booked after optimization using
nonlinear programming using a solver in Microsoft Excel.
After price optimization, the number of rooms sold
increased to 331 rooms from 107 rooms for Superior and
127 rooms from 101 rooms for Deluxe.
C. Income Comparison Analysis
Based on the optimization objectives, researchers compare
historical sales profits with sales profits from the proposed
model. Historical gains are derived from profits for the
corresponding month. The table below shows the increase in
sales profits after optimization.
Table 3.5 Income Comparison
Old Price New Price
Superior Deluxe Superior Deluxe
Revenue 53500000 60600000 90945536 66677850
Total 114100000 157623386
% 28%
Optimization can increase sales profits by 28% from
historical sales profits. The results of this study can provide
options for front office managers in determining room
prices, where the existing pricing policy is determined based
on the intuition of the front office manager. This is
potentially bad if the front office manager no longer works
in a hotel. The hotel needs a method to establish a pricing
policy without relying on front office managers to increase
sales profits or at least be the same as the previous horizon.
The proposed model complements the method gap in the
existing pricing policy.
IV. CONCLUSION
Based on the results of data processing in this study, it can
be concluded into several points, including:
1. In the forecasting process, winter's exponential smoothing
is suitable for predicting hotel room demand because hotel
demand has seasonal components that are characteristic of
seasonal data such as weekends, school holidays, national
holidays and other seasonal components.
2. Dynamic pricing model produces an optimal room rate
every day and the room rate can change every day in
accordance with the high and low demand. Sales revenue
obtained using this dynamic pricing model is Rp.
157,623,386; where the value has an increase of 28%.
greater than compared to sales revenue at the old / fixed
price.
ACKNOWLEDGMENT
This research was supported by Narapati Hotel Management
for providing their data and time to help with this research.
We would also like to show our gratitude to our family,
friends, and colleagues who help us through thick and thin
so that this paper could be done.
REFERENCES
[1] Aziz, H. A., Saleh, M., Rasmy, M. H., & ElShishiny, H. (2011).
Dynamic room pricing model for hotel revenue management
systems. Egyptian Informatics Journal, 177-183.
[2] Chopra, S., & Meindl, P. (2016). Supply Chain Management
Strategy, Planning, and Operation (6th Edition). Upper Saddle
River: Prentice Hall.
[3] Desiana, A., Ridwan, A. Y., & Aurachman, R. (2016).
Penyelesaian Vehicle Routing Problem Untuk Minimasi Total
Biaya. e-Proceeding of Engineering, 2566-2574.
[4] Fani, E., Widjajati, F. A., & Soehardjoepri. (2017).
Perbandingan Metode Winter Eksponensial Smoothing dan
Metode Event Based Untuk Menentukan Penjualan Produk
Terbaik di Perusahaan X. Jurnal Sains dan Seni ITS, 6(1).
[5] Hanke, J. E., & Wichern, D. W. (2009). Business Forecasting.
New Jersey: Prentice Hall.
[6] Ivanov, S. H. (2014). Hotel Revenue Management: From
Theotry to Practice. Varna: Zangador.
[7] Kimes, S. E., & Wirtz, J. (2003). Has Revenue Management
Become Acceptable? Findings from an International Study on
the Perceived Fairness of Rate Fences. Journal of Service
Research, 6(2), 125-135.
[8] Nofal, R., Ridwan, A. Y., & Chulasoh, B. S. (2016). Design of
Layout for Chemical Warehouse PT. XYZ with Heuristic
Approach and Shared Storage Method to Improve Capacity and
Reduce Out of Block Product. roceedings of the International
Conference of Logistic and Supply Chain Management System.
[9] Putra, A. K., Ridwan, A. Y., & Aurachman, R. (2018). Design
of Storage Allocation Using Interaction Frequency Heuristic-
Order Oriented Slotting Approach to Reduce Delay Time on
Traveling Searching Activities in Refinery Unit V Warehouse
PT XYZ. Jurnal Rekayasa Sistem Industri, 1-10.
[10] Razafuad, R., Ridwan, A. Y., & Chulasoh, B. S. (2018).
Development of E-Kanban Application Using Stock-Needs Rule
Prioritizing Policy to Reduce 0-Ick for Pharmaceutical
Warehousing. 6th International Conference on Information and
Communication Technology (ICoICT), 310-318.
[11] Shadiqurrachman. S., Ridwan, A.Y., & Kusuma, PGA. (2019).
Online Travel Agency Channel Pricing Policy based on
Dynamic Pricing Model to Maximize Profit Using Nonlinear
Integer Programming Approach. 1st International Conference on
Engineering and Management in Industrial System.