12/04/2021

FGP Loyalty Program

UJJWAL MAHESHWARI (Z5281596)
XIYING HUANG (Z5215551)
YUCHONG ZHANG (Z5221854)
JACINTA LU (Z5255704)
Table of Contents

1. Executive Summary ....................................................................................................... 2

2. Research Background and Methodology ...................................................................... 3
2.1 Research Background………………………………………………………………………………..………..3
2.2 Research Methodology………………………………………………………………………..……………..3
3. Research Questions and Results ......................................................................... ….…….4
3.1 How to measure the value of the consumer?.........................................................4
3.2 Which consumer is the most valuable? What are their characteristics?................7
3.3 What factors determine customers to churn?........................................................8
3.4 How should customers participating in loyalty program be classified? What are
the characteristics of each category?......................................................................9
3.5 What are some interesting trends in customer behaviour?..................................10
4. Conclusion ................................................................................................................... 10
5. Recommendations ...................................................................................................... 11
6. Limitations ................................................................................................................... 13
7. References ................................................................................................................... 14
8. Appendices .................................................................................................................. 15
1. Executive Summary
This report seeks to provide insights and recommendations for the profitable and sustainable
expansion of FGP’s Loyalty program. To do so, Mint Consulting has conducted statistical and
data-driven analysis specifically focusing on customer behaviour at the individual level. More
specifically, Mint Consulting seeks to estimate the value of customers in the loyalty program,
anticipate customer churn and manage customers’ effectively. We believe that examining
consumer behaviour is of high importance as the growth and profitability of companies is
dependent on the actions of customers.

To conduct analysis, this report combines customer value analysis models as well as data
mining techniques. Models such as CLV, RFM, NPS and statistical approaches such as cluster
analysis, regression analysis and hypothesis testing were used to analyse customer behaviour
for program improvement and expansion.

The main findings of our report include:

⚫ The most valuable customer groups include the following characteristic: owning a credit
card, being a male, living in city B or E and being generation Y.
⚫ FGP programme has an alarmingly low and negative Net Promoter Score and high
customer churn rate. This is mainly caused by the low satisfaction level of the fast food
chain. Conversely, customers are more likely to be loyal to the FGP project because of the
petrol station.
⚫ Customers can be segmented according to RFM scores. Parts of the customers have not
redeemed points for a long time and some of them do not redeem points very often. In
general, most customers have a low RFM score.

Using this insights, it is recommended that:

⚫ FGP should seek to increase their customer satisfaction due to the many benefits this
marketing strategy brings about. Moreover, they need to develop strategies to increase
customer loyalty based on consumer segmentations.
⚫ Loyalty programme should be expanded according to the characteristics of the customer.

In addition, due to the lack of relevant data in the dataset and unknown collection methods,
there may be some bias and human error in the analysis process.
2. Research background and Methodology

2.1 Research Background

Mint Consulting, a marketing consulting firm based in Asia, has developed a comprehensive
report to provide insights, recommendations as well limitations to assist in a profitable yet
sustainable expansion for the FGP loyalty program.

A loyalty program is often designed and implemented by merchants as an effective marketing

strategy to engage and communicate with their target customers more frequently, in the
hopes of increasing customer retention and loyalty. The FGP program encourages customers
to accumulate points by spending in any of the three merchants (fast food chain, grocery store
chain and petrol chain) and reward them whereby their points can be redeemed at any of the
three merchants. Loyalty programs bring about many benefits such as increasing purchase
volume, reduced operating cost and larger profit margin, increase long term customer
involvement and decrease customer churn.

Though, despite the benefits of loyalty programs, it’s effectiveness depends on the behaviour
of their target consumers. Mint Consulting seeks to inform the merchant members of FGP’s
loyalty program and their loyalty program manager Jennifer of their customer’s behaviour at
the individual level - specifically examining customer value, valuable customer segments and
customer churn. Ultimately, we seek to expand the loyalty program by increasing it’s customer
volume and equipping additional merchant members, as well as address complaints by
existing merchant members.

2.2 Research Methodology

In recent years, data mining and statistical analysis have become very popular not only in the
research field, but also in the business world. Data mining and statistical analysis can help
companies to identify meaningful trends, characteristics and correlations in operational data
to assist in business management and decision making. This report uses data mining
techniques such as cluster analysis, regression analysis and hypothesis testing based on three
customer value analysis models, CLV, RFM and NPS, to analyse customer behaviour and give
assistance in improving loyalty programme.
3. Research Questions and Results

3.1 How to measure the value of the consumer?

ROMI: Return on Marketing Investment

ROMI is a measure that purely takes the monetary perspective in Marketing analytics. As the
name suggests, it calculates the Return on Marketing Investment by subtracting the Expenses
from the Revenue Earned.

ROMI = (Revenue - Marketing Expenses)/Expenses

Since we are not provided with the customer acquisition cost or the marketing cost therefore,
we presume the marketing expenses to be $100 per customer. By taking $100 as the
acquisition cost, we can calculate the ROMI in both Loyal Members (Active in both 2015 &
2016) and Non Loyal Members (Active in 2015 only).

The ROMI of FGP Program is

Mean Sales Amount $ Estimated ROMI %

Active Customers 680.46 580.46%
Inactive Customer 251.93 151.93%

The ROMI of the FGP Program shows that the ‘Loyal Customers’ provided about 3.5 times
more return on marketing investment as compared to ‘Non Loyal Customers’. This explains
that the loyalty program is beneficial to all three merchant members as it brings in more
revenue in the long term. We will now see the ROMI for each of the individual merchant
members presuming each of them have an equal stake and contribution in the FGP program’s
revenue and expenses (i.e 33% each)

Mean Sales Estimated Mean Sales Estimated Mean Sales Estimated

$ Amount ROMI % $ Amount ROMI % $ Amount ROMI %
Petrol Petrol FastFood FastFood Grocery Grocery
Active 784.55 685% 126.10 26% 2785.21 2685%
Customers
Inactive 251.93 152% 264.40 264% 45.06 -55%
Customers
 All 1036.48 936% 390.50 390% 2830.27 2730%
Customers

Individual Return on Marketing Investment Shows that Fastfood Firm has the lowest Individual
Return on Investment in the FGP Program (while it is spending equal money on customer
acquisition) whereas Grocery and Petrol have significantly higher ROMI in spite of spending
the same amount as fast food.

CLV: Customer lifetime value

Customer lifetime value is a projection of the monetary value that current customers will bring
to our program in the future. We use the following simplified CLV model:

(1 + 𝑟)
𝐶𝐿𝑉 = (𝑅 − 𝐶) − 𝐴
(1 − 𝑝 + 𝑟)

r=5%, p=retention rate=75%, R=Total Sales Amount, C=$0, A= $0

We assume a discount rate at 5%. The retention rate has been calculated to be 75%. The
revenue is the total sales amount. Since we do not have information on the cost of acquiring
new customers, we assume both acquisition cost and servicing cost to be $0. Therefore, it will
lead to overestimated CLV scores.

NPS: Net Promoter Score

NPS is an effective customer loyalty metric that assesses customer’s engagement in the loyalty
program whereby they recommend the program to their relatives and friends via word of
mouth (WOM). A customer is either a promoter, passive or detractor. For any company,
promoters are desired because such customers are loyal enthusiasts who will keep buying and
refer the loyalty program to others and thus, increasing company growth. (Reichheld 1996)

NPS = Promoters% - Detractors% = (397/1995) – (1071/1995) = -33.78%

Approximately 20% (397/1995) of customers are promoters compared to the significant

greater 55% (1071/1995) of detractors, hence explaining the large negative NPS value of 34%.
This indicates an unlikely change for customers to share their loyalty program experience with
others. Thus for FGP, WOM hasn’t been a strategy that has been leveraged favourably.
RFM: Recency, Frequency and Monetary

The RFM analytic model is an important tool for measuring the value of a customer. The model
reduces the complexity of the analyzing by describing the value of the customer through only
three key indicators, namely Recency, Frequency and Monetary (Kaymak, 2001). The RFM
model has the following characteristics (Wu & Lin, 2005).

1. Customers who have recently purchased are more likely to purchase again than those who
have not recently purchased.

2. Customers who buy more frequently are more likely to buy again.

3. The higher the purchase amount, the more likely the customer is to make a repeat
purchase.

This shows that RFM is related to customer loyalty. In this report, we will be discussing the
history of redemption points rather than consumer purchase, which is to determine how well
the loyalty program is being implemented among consumers. From the given data, we assign
the scoring system, where a higher final RFM score means a higher user value. The regression
analysis to test the model can be found in the appendix 1.

Recency Score Frequency Score Monetary Score

More than 12 Months 5 1-2 5 0-1000 5
Past 9-12 Months 10 3-4 10 1001-4000 10
Past 6-8 Months 15 5-6 15 4001-7000 15
Past 4-6 Months 20 7-8 20 7001-10000 20
Past 1-3 Month 25 9-10 25 >10000 25

A visualisation of the result can be found in appendix 2. The three indicators are directly
proportional, with more points being used by customers who redeem frequently and recently,
and fewer points being used by customers who redeem less frequently and did not redeem
for longer periods of time.
3.2 Which consumer is the most valuable? What are their characteristics?

We used average CLV scores to find out the most valuable groups of customers. Since we have
assumed both acquisition and serving cost to be zero, the CLV scores predict the potential
economic value customers will bring into the FGP program. We used pivot tables to calculate
the average CLV score among different groups and used independent t-test and one-way
ANOVA to test whether or not the average CLV differed among groups, including whether they
own a credit card or not, age, gender, home city and race. Our null hypothesis is that the
means of CLV among different groups are the same. It is tested at 5% level of significance.

The results (Appendix 3) shows a significant difference on whether or not the customer owns
a credit card, gender, home city and age respectively. As shown on the Appendix 4.1, the
customers with credit cards have a higher average CLV score ($2327.00) than those without a
credit card ($1756.48). From Appendix 4.2, males have a higher average CLV ($2412.67) than
female customers ($1612.37). From Appendix 4.3, the average CLV generates from city B and
City E are statistically higher than those from city C.

In terms of customer’s age, they are segmented into five generations. As shown on Appendix
4.2, it is noticeable that baby boomers and generation X have the highest average CLV. There
are two external factors that lead to higher average CLV as age goes up. Firstly, the older age
groups (baby boomers and generation X) might have more disposable income compared to
the younger generations. Second, they might be responsible for more family member’s
groceries, while younger generations purchase groceries only for themselves. On the other
side, from descriptive analysis of age distribution (Appendix 4.5), generation Y accounts for
nearly two third of all respondents. We suggest generation Y will take over the responsibilities
of the older generation and more profit will be generated from generation Y in the long run.

In short, the most valuable customer group include the following characteristic: own a credit
card, males, living in city B or E and generation Y.
3.3 What factors determine customers to churn?

Customer churn, otherwise referred to as defection or attrition rate, refers to consumers who
are no longer affiliated with a company over a given period of time. This report makes the
assumption that a customer has churned if their last interaction with a chain occurred over 12
months ago and hence, customers who are inactive in 2016 will be considered churned.
Customer churn rate
= (customer active in 2015 – customers active in 2016) / customers active in 2015
= (1995 – 1499) / 1995
= 0.2486 = 24.86%
(Appendix 7)

A high defection rate of 24.86% indicates poor customer retention and hence, we must seek
to determine potential factors that cause customers to churn. This allows managers to
anticipate customer churn, preventing it before it actually occurs and hence, increase
customer retention which ultimately increases the program’s profitability.

Why decrease customer churn?

• Decrease customer acquisition costs as it’s more expensive to acquire and market to a
new customer than an existing one

• Reducing customer defections by 5% means double profits for FGP (Reichheld 1996)

Potential factor that predict customer churn:

Hypothesis testing: Does customer satisfaction with the FGP program significantly predict
whether they will remain active in 2016?

• H0: A customer's satisfaction with the FGP program does not significantly predict
whether they will remain active in 2016

• Ha: A customer's satisfaction with the FGP program significant predicts whether they
will remain active in 2016

We have accepted the alternate hypothesis (Ha). This is because the mean sales amount in
active customers of the fast food merchant decreased from 2015 ($264.40) to 2016 ($126.10),
whereas the mean sales amount for the other 2 firms increased drastically (Appendix 6). Thus,
a decrease in sales amount suggests that customer dissatisfaction from the fast food chain
was a potential reason for customer churn. Consequently, Mint consulting advises FGP
program to seek to increase customer satisfaction by gaining a greater holistic understanding
of their target customers e.g. demographic variables of attractive customer segments.

3.4 How should customers participating in loyalty program be classified? What are the
characteristics of each category?

Market segmentation, introduced by Wendell Smith, refers to the process of classifying

customers according to their attributes, behaviors, demands, preferences and values within a
defined strategic business model and specific market, and providing targeted products,
services and sales models (Alan & Jim 2013). We conducted a hierarchical cluster analysis of
consumer behavior based on the RFM. Hierarchical clustering is a common method of market
segmentation. It allows entities to be divided into homogeneous groups according to a set of
observed characteristics. (Richard & Dean, 2002). Ultimately, we grouped customers into
seven categories, the specific results of the cluster analysis can be found in the appendix 7.

Based on the results of the cluster analysis, we grouped the consumers. Since both Groups 6
and 7 contain only one sample each, we removed them as outliers.

Groups Number of Average R Average F Average M Average

sample Value Value Value RFM
1 (Appendix 7.2) 311 19.47 5.24 6.17 25.88
2 (Appendix 7.3) 53 21.04 7.17 10.00 33.21
3 (Appendix 7.4) 18 23.89 10.83 13.89 43.61
4 (Appendix 7.5) 7 23.57 9.29 15.71 43.57
5 (Appendix 7.6) 2 25.00 20.00 22.50 62.50
6 (Appendix 7.7) 1 10.00 5.00 25.00 35.00
7 (Appendix 7.7) 1 25.00 20.00 25.00 65.00

On this basis, we defined six categories of customers.

1. VIP (Group 5): This type of consumer scores extremely high on recency, frequency and
monetary.
2. Potential VIP (Group 4): Potential VIPs have relatively high scores on recency,
frequency and monetary.
 3. Regular (Group 3): A regular customer has high recency and frequency score but does
not redeem many points.
4. Needing Attention (Group 2): Customers have low frequency and monetary score, but
high recency scores.
5. About to Sleep (Group 1): Customers are those who have low recency, frequency and
monetary score, some of whom may have already churned.
6. Potential Regular: Potential customers are those who have spent in merchants but did
not participate in the program in 2015.

Based on the segmentation, we visualised the percentage of each type of customer who have
participated in the program (Appendix 7.8). It can be seen that the highest RFM scoring VIP
customers had only two samples, representing 0.51%, while the smallest scoring "about to
sleep" had the largest sample size, with a percentage of 79%, indicating a high risk of losing
customers to loyalty programme.

3.5 What are some interesting trends in customer behaviour?

While going through the purchase records and redeem records we noticed some strange
trends.
Sales Table
FASTFOOD $68661.28 6.00%
PETROL $812541.94 70.97%
GROCERY $263759.80 23.04%
TOTAL SALE $1144965.02 100%

Points Redemption Table

FASTFOOD 331745 63.71%
PETROL 188992 36.29%
GROCERY 0 0
TOTAL POINTS 520737 100

We discovered that Fast Food accounts for about just 6% of the total revenue in the FGP
program but 63% of the point redemption is done at Fast Food which suggests that fast food
is worse off as it has to give out more freebies. Whereas Petrol which contributes to about
70% of the sales only has 36% points redemption and Grocery is contributing to about 23%
Sale but has 0 point redemption.
This means that the Fastfood stores are losing money at the store level as they have to give
significantly more freebies with respect to the amount of sales they are making whereas
Grocery Stores have to give out 0 freebies as they have no redemption history.

We also noticed that Active Customers spend the most $ amount on Petrol and Petrol also
has the highest Mean Satisfaction Rating of 8.07. Highest satisfaction with Petrol & the
highest sales in Petrol whereas significantly lower point redemption in Petrol shows that
customers are more likely to be loyal to the FGP Program because of the Petrol Stations.

4. Conclusion
In summary, most customers are very disloyal and have a high level of customer churn. The
reason for the high customer churn rate may be due to the operational problems of the fast
food chain. In terms of market segmentation, the majority of current customers are of low
value. Characteristics of customers with high economic value include having a credit card,
being male, living in a B or E city, and being a Y generation.

5. Recommendations

Customer loyalty is a consumer's biased buying behaviour towards a brand when making a
purchasing decision. High loyalty level will decrease the churn rate. Moreover, loyal customers
have a promotional effect. Therefore, increasing customer loyalty is an effective way to
achieve sustained profit growth. Based on the RFM analysis and the results of the Customer
churn rate, it can be found that both the program and the merchants have a critical churn
situation of their customers. Therefore, how to reduce customer churn and increase customer
loyalty has become the primary issue to be addressed by the loyalty program.

1. Improving customer satisfaction

It is generally accepted that customers who are highly satisfied are more loyal than those
who are less satisfied (Dimitriades, 2006). As mentioned earlier, the reason for the low
customer loyalty in this programme could be due to consumer dissatisfaction with the
service of the fast food chain.
 Therefore, it is suggested that positive WOM should be heavily leveraged because it is a
very effective marketing strategy operating on the notion of trust (Whitler 2014). To do
so, the loyalty programme can target passives which account for 26% of customers by
increasing their customer satisfaction. However, targeting those “sitting on the fence”
customers is still not effective because the total promoters and passives combined are
still lesser than the amount of detractors. Thus, FGP is advised to increase both their
promoter score and decrease the alarming detractor value. FGP will be more effective
when NPS is >75 (Reichheld 1996). Ultimately, increasing NPS is necessary to achieve
Jennifer’s aim of expanding the loyalty program.

Sometimes, the level of customer satisfaction does not fully reflect loyalty. Satisfaction
refers to the customer's satisfaction with past services and is not related to whether they
will buy again in the future. In this regard, Lafley and Martin suggest that companies need
to get their customers into the habit of buying (Lafley and Martin, 2017). This requires a
concerted effort by merchants participating in loyalty programs to improve the quality of
their products, improve the customer experience, etc.

2. Develop different marketing strategies based on market segments

In addition to developing a strategy covering all customers, loyalty programs need to pay
attention to market segmentation. According to the findings of the hierarchical cluster
analysis, customers participating in loyalty program varies greatly and different marketing
strategies need to be developed according to different customer segments.

A segment of customers with relatively average RFM scores (VIP customers and regular
customers) are the most stable customers and loyalty programs can maintain them by
offering additional benefits. For customers with low frequency scores (potential VIP
customers), loyalty program and merchants can offer products or services that are
superior to those of their peers and build trust with them, thereby increasing their repeat
buying. For customers with low recency scores (Needing Attention and about to Sleep
customers), loyalty program should use phone calls and advertising to bring them back
and prevent them from churning. Then, attempts can be made to build a deeper trust
with them.
Moreover, there is a category of customers who spend with the merchant but do not
participate in points redemption, which can be targeted for expansion of the loyalty
programme. In this regard, this report gives the following comments.

3. Expansion according to customer characteristics

For Jennifer's objective of expansion, we suggest to strengthen the publicity of members'

perceived benefits and mainly target high-value customers- people with high CLV. Since
part of members did not participate in the points redemption, we suggest the FGP
program should ensure that the interests of the members are displayed in a prominent
position in every merchant. From our CLV and age distribution analysis, we recommended
to target people who own credit cards and generation Y. To reach people with credit cards,
the strategy is cooperating with banks. For example, waiving the sign-up fee for those
people who have credit cards from cooperative banks or launching co-branded credit
cards like what frequent flyer programs did. In terms of generation Y, we recommend
optimizing the steps of redeeming points. For example, launching an application allowing
members to redeem the point for groceries.

6. Limitations
The analysis in this report is limited by the data set. Firstly, the first limitation is the lack of
data. Many of the data is not given, for example the acquisition and serving cost required to
calculate CLV, leading to potential errors between the CLV scores in this report and the true
values. In addition, the method of data collection is unknown and there may be some bias and
human error in the data collection process. The survey questions are very basic and indefinite,
which might cause omission on other possible factors. For the question about race, ‘other’ is
a very broad category and it might contain a multi-cultural background.
7. References

Kaymak, U 2001, ‘Fuzzy target selection using RFM variables’, in Proceedings Joint 9th IFSA
World Congress and 20th NAFIPS International Conference (Cat. No. 01TH8569), IEEE, pp.
1038–1043 vol.2.

Wu, Jing & Lin, Zheng 2005, ‘Research on customer segmentation model by clustering’, in
Proceedings of the 7th international conference on electronic commerce, ACM, pp. 316–318.

Zimmerman, Alan & Blythe, Jim 2013, ‘Segmentation, targeting, and positioning’, in Business
to Business Marketing Management, Routledge, pp. 145–167.

Johnson, RA & Wichern, DW 1982, Applied multivariate statistical analysis, Prentice-Hall,

Englewood Cliffs, N.J.

Dimitriades, Zoe S 2006, ‘Customer satisfaction, loyalty and commitment in service

organizations’, Management research news, vol. 29, no. 12, pp. 782–800.

Lafley, AG & Martin, RL 2017, ‘Customer Loyalty Is Overrated’, Harvard Business Review,
Available from: https://hbr.org/2017/01/customer-loyalty-is-overrated [accessed 1st April
2021]

Whitler, K 2014, ‘Why Word Of Mouth Marketing Is The Most Important Social Media’,
Forbes, Available from: https://www.forbes.com/sites/kimberlywhitler/2014/07/17/why-word-of-
mouth-marketing-is-the-most-important-social-media/?sh=65fef07a54a8 [accessed 1st April 2021]
8. Appendices
Appendix 1: Regression Analysis

In order to test the validity of the RFM model and to estimate future customer behavior, the
equation form has been provided below.

RFM= β0 + β1 Recency +β2 Frequency + β3 Monetary +ε

Where RFM is the RFM score, which measures the value of the customer. Recency is the time
interval in months between the last redemption of points up to 1 January 2016. Frequency is
the number of redemptions in the statistical period. Monetary is the total number of points
redeemed in the statistical period.

A multiple linear regression analysis was performed on the data and the following output was
obtained.

RFM = 29.03-1.66 Recency +2.49 Frequency +0.001 Monetary +ε

Assume that the Gauss-Markov assumptions hold, and holding other variables constant, the
estimated coefficients for β1 indicates that the RFM score is reduced by 1.66 for each
additional month that the most recent purchase is made from the statistical cut-off date (1
January 2016). And β2 indicates that for each additional purchase made during the statistical
period, the RFM score increases by 2.49 points. The last β3 indicates that every 1 dollar
increase in the purchase amount is associated with increased RFM score by 0.002. R2 of the
OLS estimates of the equation is 93.22%. It means 93.22% of sample variation in RFM can be
explained by regression.

To conclude, the results of the regression analysis are consistent with the model previously
assumed. Therefore, this RFM model can be used to roughly estimate customer value.

Appendix 2: RFM Visualization

Appendix 3: CLV SPSS Results:
3.1 T-test on creditcard

3.2 One-way ANOVA on Age-generation

3.3 T-test and NPar-test on Gender

3.4 One-way ANOVA on Homecity

3.5 One-way ANOVA on Race

Appendix 4: CLV Analysis

Appendix 4. 1 Appendix 4.2

 Appendix 4.3 Appendix 4.4

Appendix 4.5 Appendix 4.6: Descriptive Analysis

Appendix 5: Customer Churn

Frequency Percentage
Active 2015 1995 100%
Active 2016 1499 75%
Non-Active 2016 496 25%

Appendix 6: Mean Sales Amount and ROMI% Descriptive Analysis

Appendix 7: Cluster Analysis

Appendix 7.1

Appendix 7.2

Appendix 7.3 Appendix 7.4

Appendix 7.5 Appendix 7.6 Appendix 7.7 Appendix 7.8

Appendix 8: Customer Trends