Analytics and Data Mining - Goran's Blog
If you are interested in analytics - you are on a right place! Send me a comment, let me know what is happening in YOUR analytical world! For much more visit my portal: http://www.bigdatanalysis.com/ Hope to see you there! Goran
Tuesday, April 5, 2016
Sunday, August 9, 2015
Big-Data Analytics in hotel industry
Hotel industry is another industry where effective use of
analytics can change dramatically how business is run. It is another data rich
industry that captures huge volumes of data of different types, including
video, audio, and Web data. However, for most hoteliers data remains an
underused and underappreciated asset. Hoteliers capture loyalty information,
for example, but few go beyond loyalty tier in how they consistently view and
take action with their guests. With analytical exploitation of their data, hoteliers can
go beyond their traditional loyalty programs and deepen their knowledge of
guests in order to develop a more granular understanding of segment behavior,
needs, and expectations; identify profitable customer segments and their buying
preferences; and identify opportunities to attract new guests. But all that
starts with having clear customer-driven vision, before embarking on Integrating and standardizing guest data from
multiple channels, systems and properties into a unified, accurate view
of all interactions.
Next phase is using analytics to segment guests according
to booking trends, behavior and other factors in order to reveal their
likelihood to respond to promotions and emerging travel trends. It is vitally
important for hoteliers to be able to understand guest preferences (locations,
activities, and room types), purchase behavior (frequency, length of stay, time
of year) and profit potential in order to increase the brand loyalty and wallet
share of their most valuable guests. Focusing on the wrong guests reduces
profitability across the enterprise. For example, if a hotel targeted guests
who would likely take advantage of spa services, golf and restaurants, rather
than guests who only generate room nights, they could significantly increase
revenues and profitability.
To maximize profits, hotels need
to increase the loyalty and wallet share of their most valuable guests by
marketing to their preferences and encouraging repeat visits. Focusing on the
wrong guests reduces profitability across the enterprise. For example, if a
hotel targeted guests who would likely take advantage of spa services, golf and
restaurants, rather than guests who only generate room nights, they could significantly
increase revenues and profitability. Unfortunately, money often gets spent on
blanket campaigns that don’t target individual guests or segments with offers
they’re most likely to respond to. As a result, guests may feel that the hotel
doesn’t care about them, or simply doesn’t offer services designed to meet
their needs. It becomes easy for those guests to switch to a competing hotel.
Unfortunately,
money often gets spent on blanket campaigns that don’t target individual guests
or segments with offers they’re most likely to respond to. As a result, guests
may feel that the hotel doesn’t care about them, or simply doesn’t offer
services designed to meet their needs. It becomes easy for those guests to
switch to a competing hotel. For analytics to truly
be a game changer, hospitality organizations need to recognize the difference
between reactive and proactive decision making. Using your data to create
reports, drill-downs or alerts helps you to keep a finger on the pulse of your
business. But these things only show you what
happened. They will not tell you why the
problem is happening or what effect it will have in the future. Predictive analytics, like forecasting and optimization,
can help you figure out why things are happening, show you what will happen
next, or even lead you to the best alternative action considering all of your
operating constraints. Hoteliers are starting to use more and more predictive analytics to move from reactive to proactive
decision making, which would enable them to stay
one step ahead of trends, set strategy and achieve goals. They gain advantage over the competition, increase value to
shareholders, and continue to surprise and delight their guests. Following
are areas where analytics can play essential role:
Customer Segmentation
For the hotel industry, a more useful
approach might be to identify the unique cluster groups and to then conduct a separate value segmentation
exercise for each cluster. For example, for a given hotel we identify 4 basic
clusters or distinct customer groups such as tennis groups, ski group, pampered
group (e.g. use spa and valet type services) and the nighthawk group ( fine
dining and theatre goers). The segmentation approach might look as follows:
initial learning from this type of segmentation could be used in developing a
marketing strategy that is data-driven. The hotel could examine its current business customer base and once
again establish unique groups of business customers. For example, we know that
there are groups of business customers that simply use the hotel for overnight
stays, while others are there for longer term events held at the hotel. It may
be possible to further segment these groups based on industry sector. We would
certainly expect that local events held by the oil and gas industry might be
more appropriate in one city, while financial services type events may be more
prevalent somewhere else. Of course, all this supposition on what might define
unique business segments would need to be determined quantitatively through
clustering routines. By using the data and mathematics rather than intuitive
judgment to define key customer segments, we can develop unique programs that
are appropriately geared to different groups of unique business travelers.
Customer experience has always been the overriding customer philosophy within
the travel industry, long before the advent of data analytics. Yet, with data
analytics, the travel industry can now use information to make better decisions
regarding its customers. This enhanced decision-making capability enables hotels
to be more proactive with its customers. Traditionally, success in the hotel
industry has always been determined by superior customer service actions that
address the immediate requested needs of the customer. The competitive
advantage in today’s hotel industry is driven more by the ability to anticipate
and proactively meet the needs of customers; an ability that can only be
exercised through data analytics.
Customer Profiling Customer
profiling is accomplished through in-depth analysis of guest demographics and
lifestyle characteristics. Attributes such as income levels, family status, age
and sports and cultural interests, if known, can be appended to model guests.
Customer profiling can be used to create an e-mail listserv for targeted
marketing of current as well as prospective clientele. Prospect profiles can be
especially useful in identifying those folks most likely to respond to
marketing and/or promotional offers. Profiling can also be important in
determining which market segments are most productive and profitable.
Site Selection Data
mining can also be essential to determining sound criteria for restaurant site
selection given an index derived from an analysis of high-volume, successful
units. Such items as demographics (customer profile) and psychographic (buying
patterns), and related customer descriptors are used to delineate highly
probable factors for site modeling. As a result, evaluation data and analytical
profiling qualify companies to be better able to identify candidate sites.
Forecasting Customer
transactional data (segmented by menu item and day part) can be useful in the
development of a forecasting model that accurately produces meaningful
expectations. Regardless of whether a restaurant company relies on moving
average or time series forecasting algorithms, data mining can improve the
statistical reliability of forecast modeling. Estimating in advance how much
and when menu items will need to be prepared is critical to efficient food
production management. Data mining can provide a prognostication of product
usage by day part given available sales data. In addition, knowing how much
product was sold during any meal period can also be helpful in supporting an
effective inventory replenishment system that minimizes the amount of capital
tied up in stored products.
Customer
Relationship Management An effective CRM program can be a direct
outcome of data mining applications. The ability to enhance CRM given rapid
accessibility of more comprehensive management information should lead to
satisfied clientele and improved sales performance. The ability to anticipate
and affect consumer behavior (influence menu item sales and other promotions)
can provide the restaurant with a competitive advantage. Having a signature
item, for example, can be found to be a driver of improved relations while
providing a product that customers do not perceive as having an equivalent
elsewhere.
Menu
Engineering An analysis of menu item sales and
contribution margins can be helpful to continuous, successful restaurant
operations. While menu engineering deals with menu content decisions, data
mining can produce reports to indicate menu item selections, by customer
segment, as a basis for operational refinement. For example, Applebee’s has
been described as employing data mining expressly for the purpose of
determining ingredient replenishment quantities based on a menu optimization
quadrant analysis that summarizes menu item sales. Through such analysis the
company then decides which menu items to promote.
Productivity
Indexing By
correlating order entry time (POS time stamped) with settlement time, data
mining is able to provide a reliable estimate of elapsed production and service
times. This data provides insight into average service time relative to customer
turnover as well as waiting line statistics. While productivity data is
difficult to ascertain, this analysis provides factual data to assist
management in fine tuning operations (heart of the house and dining room
staff).
Customer Associations and Sequencing Data mining can uncover affinities between isolated events. For example, a guest purchasing the restaurant house specialty is likely to also purchase a small antipasto salad and glass of Chardonnay. Paired relationships provide a basis for bundling menu items into a cohesive meal that simplifies ordering while ensuring customer satisfaction. Menu design can also be manipulated to feature such combinations as unique opportunities for customers. Data associations are often credited with a means for influencing customers to spend more than anticipated or upselling.
Customer Associations and Sequencing Data mining can uncover affinities between isolated events. For example, a guest purchasing the restaurant house specialty is likely to also purchase a small antipasto salad and glass of Chardonnay. Paired relationships provide a basis for bundling menu items into a cohesive meal that simplifies ordering while ensuring customer satisfaction. Menu design can also be manipulated to feature such combinations as unique opportunities for customers. Data associations are often credited with a means for influencing customers to spend more than anticipated or upselling.
Forecasting As
mentioned previously, forecasting is one of the strengths of data mining and
enables restaurants to better plan to exceed the needs of its clientele.
Forecasting enables more efficient staffing, purchasing, preparation and menu
planning.
Customer Value Within the travel industry,
customers have always considered their time at a hotel as an experience rather
just a visit. Activities such as fine dining, nightly entertainment, spas,
corporate seminars / meetings nurture this notion of ‘customer experience’.
This range of activities is going to have varying levels of appeal among a
given clientele. The role of data mining and analytics can be quite significant
in helping us to better understand these varying client needs. Our first task
might be to conduct a basic customer value exercise in order to ultimately
identify our best customers. As with many analytical exercises, the concept of
seasonality needs to be considered here. Seasonality is a very significant
factor within the hotel industry. Most analysts would agree that for the travel
industry, the issue of seasonality can potentially have a significant impact on
travel behaviour. For example, one traveler may spend $1,000 annually as a
casual traveler throughout the year and is considered an “average customer”.
Another traveler spends $1,000 annually, but on a tennis package for one week
period. Both customers spend the same amount but are in fact very different
types of customers. This notion of seasonality is significant when conducting
any analytics exercise particularly if we consider that many hotels will offer
tennis and golf packages in the summer and ski packages in the winter. In
addition to the issue of seasonality, there are various services that may have
more appeal to certain groups of customers. Fine dining and theatre may appeal
to one group of customers while spas and perhaps valet type services appeal to
another group. With varying interests amongst travel clientele, a “cluster
type” segmentation exercise would be a very useful way to identify different
groups of customers. Experts in the travel industry would certainly agree that
there are distinct or homogenous customer segments. Using the data being captured
on travel customers, we can apply some Science to identify truly distinct
customer segments. How do we integrate the notion of customer value within the
cluster segmentation approach? Typically we might conduct a value segmentation
exercise on the entire customer base and then overlay the cluster segments to
see how they align with customer value.
Personalized Marketing and Website Optimization By tracking and processing your customer’s
behavior and actions, you can provide them with personalized offers that are
more effective and give a personal touch. Let say for example, you have a
client that visits your hotel restaurant on a frequent basis due to business.
When you are planning your next promotional campaign, make it targeted and
personal. Send an email to this client saying “We know you have enjoyed our
great restaurant in the past, so when you visit next week, here’s a coupon for
a free appetizer and drink”. There are various marketing automation tools out
there that facilitate this process and allow you to deploy an effective and
personalized cross channel marketing strategy. Another area where you can use data in order to
boost business is to optimize your website or landing pages through A/B
testing. Are you implementing a marketing campaign, but the conversion rates of
your landing pages are not as anticipated? An easy solution is to resort to A/B
testing. A/B testing is the act of running a simultaneous experiment
between two or more pages to see which performs or converts the best.
Energy Consumption
In the
hotel industry world, analytics can also be used for internal operations.
Energy consumption accounts for 60 to 70% of the utility costs of a typical
hotel. However, costs can be controllable, without sacrificing guest comfort,
by using energy more efficiently. At present times, smart data can help
managers to build energy profiles for their hotels. There are modern software
solutions that gather data from multiple sources, including weather data,
electricity rates and a building’s energy consumption to build a comprehensive
‘building energy profile’. Through a cloud-based, predictive analytics
algorithm, the software can fine-tune whether power comes from the grid or an
onsite battery module.
Investment Management
Another
way to use analytics for the hotel industry is for financial performance and
investment. When managers want to proceed to make capital investments, like
refurbish the lobby or the rooms or renovate their restaurant, they can
consider implementing a “Randomized Testing” strategy. How does this work?
Basically the hotel chain would refurbish the lobby and rooms in only two or
three “test” hotels. Then, they would monitor if there has been a difference in
bookings and customer satisfaction. The data obtained from the test hotels can
then be compared to the data of the other hotels that were not refurbished.
Thus, managers can take a data driven decision and clearly see if it’s profitable
to make the investment throughout the whole chain. In conclusion, data
analytics can be a powerful force in transforming the hotel industry. From
taking evidence based actions to developing customer centered marketing and
pricing strategies, increasing the ROI of capital investments and generally
empowering hoteliers to make bigger and better decisions. There
are however also some great examples of hotel chains moving in the right
direction in respect to use of analytics. This can result in improved customer
satisfaction, personalized marketing campaigns and offers so that the right
guests book the right room at the right moment and at the right rate. In
addition, it can boost in employee productivity and more efficient operations.
The advantages of using analytics and data mining the hotel industry are
enormous. Deep
customer insights can lead to improved guest satisfaction and an unforgettable
experience. Making these insights available to all levels and departments
within the hotel is crucial. It allows the concierge to know which local tours
to recommend that fit your preferences. It allows the restaurant departments to
predict which menu items are likely to be ordered, based for example on the
local weather. It allows the reservations department to predict the optimal
rate for a room and sales and marketing to create tailored messages across
different (social) networks and send truly personalized email campaigns. Let’s
dive a bit deeper in some possibilities:
The right room at the right rate
Yield
management is nothing new in the hotel industry. Providing different rates to
different customers has been done for ages and with success. Big Data offers
hotels the possibility to take revenue management a giant leap forward and
start offering truly personalized prices and rooms to guests. The massive
growth in booking websites, hotel review websites such as TripAdvisor and Yelp
and the ever growing list of social media networks offer a lot of potential.
Combined with hotels’ own CRM systems and/or loyalty programs there is a lot of
data that can be used to optimize revenue management. According
to some industry studies, the hotel chain Marriot has been using Big Data
Analytics to start predicting the optimal price of its rooms to fill its
hotels. They do this by using improved revenue management algorithms that can
deal with data a lot faster, by combining different data sets and making these
insights available to all levels to improve decision-making. The American hotel
chain Denihan goes even a
step further. They used Analytics software to maximize profit and revenue
across thousands of their rooms by combing their own data sets and data from
for example review sites, blogs and/or social network website. They understand
the likes and dislikes of their guests, optimize their offering and adjust the
room rates accordingly.
Mobile Big Data throughout the Hotel
More
and more hotels have developed mobile Apps that guests can use to book a hotel
room. These apps however offer vast more possibilities for guests if developed
correctly. It could serve the key to your hotel room; it can be used to make
reservations in restaurants and spa’s and for example to order room service. If
hotels start using the vast possibilities of mobile application they can
generate massive amounts of data that can be analyzed. So, from a guest
perspective, mobile offers a lot of convenience. From an employee perspective,
it can make life a lot easier for the staff while at the same time increase
customer satisfaction. Providing the housekeeping department with smart devices
for example will allow them to know in real time, that you prefer an extra pillow or
an extra light. Kempinski and Hyatt in Dubai already use such applications for
their hotels. Most of the staff within hotels do not have an office or a
computer so providing them with real-time guest information should be done
on-the-go. Although this requires a different approach and a different way of
presenting the insights, placing user-friendly analytics in the hands of guest
facing employees will definitely improve customer satisfaction.
More efficient hotel operations
From a
hotel operations point of view, big data offers also many different solutions.
Big Data can be used to reduce your energy bill for example. By combining data from 50 different sources, including
electricity rates, weather data and a building’s energy consumption, two
InterContinental hotels in San Francisco managed to reduce their energy costs
by 10-15%. They created detailed energy profiles for their buildings and using
a predictive algorithm they decided whether to use an onsite battery module or
receive power from the grid. Hotels
should also use analytics more to help more efficiently running their IT operations, which is especially relevant for
chains that operate their own booking engine. A server that breaks down or a
booking engine that is inaccessible could result in lost bookings and therefore
lost revenue. IT operations analytics monitors a hotel’s complete IT
environment, including the different relations between applications and
hardware and can predict when things are about to go wrong. Advanced IT
operations analytics can even solve problems automatically before they occur.
This could save a lot of money because IT that’s not working will results in a
bad customer experience. Of course the examples given here are just a few of
the massive possibilities that analytics has to offer for the hotel industry. Data mining technology can be a
useful tool for hotel corporations that want to understand and predict guest
behavior. Based on information derived from data mining, hotels can make
well-informed marketing decisions, including who should be contacted, to whom
to offer incentives (or not), and what type of relationship to establish. Data
mining is currently used by a number of industries, including hotels,
restaurants, auto manufacturers, movie-rental chains, and coffee purveyors.
Firms adopt data mining to understand the data captured by scanner terminals,
customer-survey responses, reservation records, and property-management
transactions. This information can be melded into a single data set that is
mined for nuggets of information by data mining experts who are familiar with
the hotel industry. However, data mining is no guarantee of marketing success.
Hotels must first ensure that existing data are managed—and that requires
investments in hardware and software systems, data mining programs,
communications equipment, and skilled personnel. Affiliated properties must
also understand that data mining can increase business and profits for the
entire company and should not be viewed as a threat to one location. Since data mining is in its initial stages in the hotel
industry, early adopters may be able to secure a faster return on investment
than will property managers who lag in their decisions. Hotel corporations must
also share data among properties and divisions to gain a richer and broader
knowledge of the current customer base. Management must ensure that hotel
employees use the data-management system to interact with customers even though
it is more time consuming than a
transactional approach.
Big-data analytics for lenders and creditors
Credit today is granted by
various organizations such as banks, building societies, retailers, mail order
companies, utilities and various others. Because of growing demand, stronger
competition and advances in computer technology, over the last 30 years
traditional methods of making credit decisions that rely mostly on human
judgment have been replaced by methods that rely mostly on statistical models.
Such statistical models today are not only used for deciding whether or not to
accept an applicant (application scoring), but also to predict the likely
default of customers that have already been accepted (behavioral scoring) and
to predict the likely amount of debt that the lender can expect to recover (collection
scoring). The term credit scoring can be
defined on several conceptual levels. Most fundamentally, credit scoring means
applying a statistical model to assign a risk score to a credit application or
to an existing credit account. On a higher level, credit scoring also means the
process of developing such a statistical model from historical data. On yet a
higher level, the term also refers to monitoring the accuracy of one or many
such statistical models and monitoring the effect that score based decisions
have on key business performance indicators.
Credit scoring is performed,
because it provides a number of important business benefits, all of them based
on the ability to quickly and efficiently obtain fact based and accurate
predictions of credit risk of individual applicants or customers. So, for
example, in application scoring, credit scores are used for optimizing the
approval rate for credit applications. Application scores enable the
organization to choose a optimal cut-off score for acceptance, such that market
share can be gained while retaining maximum profitability. The approval process
and the marketing of credit products can be streamlined based on credit scores:
High risk applications can, for example, be given to more experienced staff or
pre-approved credit products can be offered to selected low-risk customers via
various channels, including direct marketing and the Web.
Credit scores, both of prospects
and existing customers, are essential in the customization of credit products.
They are used for determining custom credit limits, down payments or deposits
and interest rates. Behavioral credit scores of existing customers are used in
the early detection of high risk accounts and enable the organization to
perform targeted interventions, for example by pro-actively offering debt
restructuring. Behavioral credit scores also form the basis for more accurate
calculations of the total consumer credit risk exposure, which can result in a
reduction of bad debt provision.
Other benefits of credit scoring
include an improved targeting of audits at high-risk accounts, thereby
optimizing the workload of the auditing staff. Resources spent on debt
collection can be optimized by targeting collection activities at accounts with
a high collection score. Collection scores are also used for determining the
accurate value of a debt book before it is sold to a collection agency. Finally, credit scores serve to
assess the quality of portfolios intended for acquisition and to compare the
quality of business from different channels, regions and suppliers.
Building credit models in-house
While under certain circumstances
it is appropriate to buy ‘ready-made’ generic credit models from outside
vendors or to have credit models developed by outside consultants for a
specific purpose, maintaining a practice for building credit models in-house
offers several advantages. Most directly, it enables the lending organization
to profit from economies of scale when many models need to be built and to
afford a greater number of segment specific models for a greater variety of
purposes.
Building up a solid, re-usable
and flexible data, knowledge and skill base of its own also makes it easier for
the organization to stay consistent in the interpretation of model results and
reports and to use a consistent modeling methodology across the whole range of
customer related scores. This results in a reduced turnaround time for the
integration of new models, thereby freeing resources to more swiftly respond to
new business questions with new creative models and strategies.
Finally, in-house modeling
competency is needed to verify the accuracy and analyze the strengths and
weaknesses of acquired credit models, to reduce access of outsiders to
strategic information and to retain competitive advantage by building up
company specific best practices.
Larger credit scoring process
Modeling is the process of
creating a scoring rule from a set of examples. In order for modeling to be
effective, it has to be integrated into a larger process. Let’s look at
application scoring. On the input side, before the modeling, the set of example
applications has to be prepared. On the output side, after the modeling, the
scoring rule has to be executed on a set of new applications, so that credit
granting decisions can be made.
The collection of performance
data is at the beginning and at the end of the credit scoring process. Before a
set of example applications can be prepared, performance data has to be
collected so that applications can be tagged as ‘good’ or ‘bad’. After new
applications have been scored and decided upon, the performance of the accepts
again has to be tracked and reports created, so that the scoring rule can be
validated and possibly substituted, the acceptance policy be fine-tuned and the
current risk exposure be calculated.
Choosing the right model
With available analytical
technologies it is possible to create a variety of model types, such as
scorecards, decision trees or neural networks.
When you evaluate, which model type is best suited for achieving your
goals, you may want to consider criteria such as the ease of applying the
model, the ease of understanding it and the ease of justifying it. At the same
time, for each particular model of whatever type, it is important to assess its
predictive performance, i.e. the accuracy of the scores that the model assigns
to the applications and the consequences of the accept/reject decisions that it
suggests. A variety of business relevant quality measures, such as concentration,
strategy and profit curves are used for this (see section Model Assessment in
the case study section below). The best model will therefore be determined both
by the purpose for which the model will be used and by the structure of the
data set that it is validated on.
-----------------------------------------------------------------------------------------------------------
Scorecards
The traditional form of a credit
scoring model is a scorecard. This is a table that contains a number of
questions that an applicant is asked (called characteristics) and for each such
question a list of possible answers (called attributes). One such characteristic may, for example, be
the age of the applicant, and the attributes for this characteristics then are
a number of age ranges that an applicant can fall into. For each answer, the
applicant receives a certain amount of points – more if the attribute is one of
low risk, less vice versa. If the application’s total score exceeds a specified
cut-off amount of points, it is recommended for acceptance. Scorecard model, apart
from being a long established method in the industry, still has several
advantages when compared with more recent ‘data mining’ types of models, such
as decision trees or neural networks. A
scorecard is easy to apply: if needed the scorecard can be evaluated on a sheet
of paper in the presence of the applicant. It is easy to understand: the amount
of points for one answer doesn’t depend on any of the other answers and across
the range of possible answers for one question the amount of points usually
increases in a simple way (often monotonically or even linearly). It is
therefore often also easy to justify a decision that is made on the basis of a
scorecard to the applicant. It is possible to disclose groups of
characteristics where the applicant has a potential for improving the score and
to do so in broad enough terms not to risk manipulated future applications.
Scorecard development process
Development sample
The development sample (input data set) is a balanced sample
consisting of 1500 good and 1500 bad accepted applicants. ‘Bad’ has been
defined as having been 90 days past due once. Everyone not ‘bad’ is ‘good’ , so
there are no ‘indeterminates’. A
separate data set contains the data on rejects. The modeling process,
especially the validation charts, require information about the actual good/bad
proportion in the accept population. Sampling weights are used here for
simulating that proportion. A weight of 30 is assigned to a good application
and a weight of 1 to a bad one. Thereafter all nodes in the process flow
diagram treat the sample as if it consisted of 45 000 good applications and
1500 bad applications. Figure 3 shows the distribution of good/bad after the
application of sampling weights. The bad rate is 3.23%. A Data Partition node
then splits a 50 % validation data set away from the development sample. Models
will later be compared based on this validation data set.
Classing
Classing is the process of automatically and/or
interactively binning and grouping interval, nominal or ordinal input variables
in order to
- manage the number of attributes per characteristic
- improve the predictive power of the characteristic
- select predictive characteristics
- make the Weights Of Evidence - and thereby the amount of points in the scorecard - vary smoothly or even linearly across the attributesThe amount of points that an attribute is worth in a scorecard is determined by two factors:
- the risk of the attribute relative to the other attributes of the same characteristic and
- the relative contribution of the characteristic to the overall scoreThe relative risk of the attribute is determined by its ‘Weight of Evidence’. The contribution of the characteristic is determined by its co-efficient in a logistic regression (see section Regression below).The Weight of Evidence of an attribute is defined as the logarithm of the ratio of the proportion of goods in the attribute over the proportion of bads in the attribute. High negative values therefore correspond to high risk, high positive values correspond to low risk. Since an attribute’s amount of points in the scorecard is proportional to its Weight of Evidence (see section Score Points Scaling below) the classing process determines how many points an attribute is worth relative to the other attributes of the same characteristic.After classing has defined the attributes of a characteristic, the characteristic’s predictive power, i.e. its ability to separate high risks from low risks, can be assessed with the so called Information Value measure. This will aid the selection of characteristics for inclusion in the scorecard. The Information Value is the weighted sum of the Weights of Evidence of the characteristic’s attributes. The sum is weighted by the difference between the proportion of goods and the proportion of bads in the respective attribute. The Information Value should be greater than 0.02 for a characteristic to be considered for inclusion in the scorecard. Information Values lower than 0.1 can be considered weak, smaller than 0.3 medium and smaller than 0.5 strong. If the Information Value is greater than 0.5, the characteristic may be over-predicting, meaning that it is in some form trivially related to the good/bad information.There is no single criterion, when a grouping can be considered satisfactory. A linear or at least monotone increase or decrease of the Weights of Evidence is often what is desired in order for the scorecard to appear plausible. Some analysts would always only include those characteristics where a sensible re-grouping can achieve this. Others may consider a smooth variation sufficiently plausible and would include a non-monotone characteristic such as ‘income’, where risk is high for both high and low incomes, but low for medium incomes, provided the Information Value is high enough.
Regression analysis
After the relative risk across
attributes of the same characteristic has been quantified, a logistic
regression analysis now determines how to weigh the characteristics against
each other. The Regression node
receives one input variable for each characteristic. This variable contains as
values the Weights of Evidence of the characteristic’s attributes. (see table 1
for an example of Weight of Evidence coding). Note that Weight of Evidence
coding is different from dummy variable coding, in that single attributes are
not weighted against each other independently, but whole characteristics are,
thereby preserving the relative risk structure of the attributes as determined
in the classing stage
A variety of further selection methods (forward, backward,
stepwise) can be used in the Regression node to eliminate redundant
characteristics. In our case we use a simple regression. These values are in
the following step multiplied with the Weights of Evidence of the attributes to
form the basis for the score points in the scorecard.
Score points scalling
For each attribute its Weight of
Evidence and the regression co-efficient of its characteristic could now be
multiplied to give the score points of the attribute. An applicant’s total
score would then be proportional to the logarithm of the predicted bad/good
odds of that applicant. However, score points
are commonly scaled linearly to take more friendly (integer) values and to
conform with industry or company standards. We scale the points such that a
total score of 600 points corresponds to good/bad odds of 50 to 1 and that an
increase of the score of 20 points corresponds to a doubling of the good/bad
odds. For the derivation of the scaling rule that transforms the score points
of each attribute see equations 3 and 4. The scaling rule is implemented in the
Scorecard node (see Figure 1), where it can be easily parameterized. The
resulting scorecard is output as a table in HTML and is shown in table 2. Note, how the score points of the various
characteristics cover different ranges. The score points develop smoothly and,
with the exception of the ‘Income’ variable, also monotonically across the
attributes.
Reject Inference
The application scoring models we
have built so far, even though we have done everything correctly, still suffer
from a fundamental bias. They have been built based on a population that is
structurally different from the population to which they are supposed to be
applied. All the example applications in the development sample are
applications that have been accepted by the old generic scorecard that has been
in place during the last two years. This is so because only for those accepted
applications it is possible to evaluate their performance and to define a
good/bad variable. However, the
through-the-door population that is supposed to be scored is composed of all
applicants, those that would have been accepted and those that would have been
rejected by the old scorecard. Note that this is only a problem for application
scoring, not for behavioral scoring . As a partial remedy to this
fundamental bias, it is common practice to go through a process of reject
inference. The idea of this approach is to score the data that is retained of
the rejected applications with the model that is build on the accepted
applications. Then rejects are classified as inferred goods or inferred bads and
are added to the accepts data set that contains the actual good and bad. This
augmented data set then serves as the input data set of a second modeling run.
In case of a scorecard model this involves the re-adjustment of the classing
and the re-calculation of the regression co-efficients.
------------------------------------------------------------------------------------------------------
Decision Trees
On the other hand, a decision
tree may outperform a scorecard in terms of predictive accuracy, because unlike
the scorecard, it detects and exploits interactions between characteristics. In
a decision tree model, each answer that an applicant gives determines what
question he is asked next. If the age of an applicant is for example greater
than 50 the model may suggest to grant a credit without any further questions,
because the average bad rate of that segment of applications is sufficiently
low. If, on the other extreme, the age of the applicant is below 25 the model
may suggest to ask about time on the job next. Credit would then maybe only
granted to those that have exceeded 24 months of employment, because only in
that sub-segment of youngsters the average bad rate is sufficiently low. A decision tree model thus consists of a set
of if .. then … else rules that are still quite straightforward to apply. The decision
rules also are easy to understand, maybe even more so than a decision rule that
is based on a total score that is made up of many components. However, a decision rule from a
tree model, while easy to apply and easy to understand, may be hard to justify
for applications that lie on the border between two segments. There will be cases where an applicant will
for example say: ‘If I had only been 2 months older I would have received a
credit without further questions, but now I am asked for additional securities.
That is unfair.’ That applicant may also be tempted to make a false statement
about his age in his next application. Even if a decision tree is not
used directly for scoring, this model type still adds value in a number of
ways: the identification of clearly defined segments of applicants with a
particular high or low risk can give dramatic new insight into the risk
structure of the population. Decision trees are also used in scorecard
monitoring, where they identify segments of applications where the scorecard
under performs.
Finally, decision trees often can achieve a similar
predictive power as a scorecard with much fewer characteristics. Models that
only require few characteristics, sometimes called ‘short scores’, are becoming
especially popular in the context of campaigning and marketing for credit
products. However, there is a fundamental problem associated with short scores:
they diminish the richness of information that the organization can collect on
the applicants and thereby erode the basis for future modeling.
--------------------------------------------------------------------------------------------------------
Neural Nets
With the decision tree, we could
see that there is such thing as a decision rule that is too easy to understand
and thereby invites fraud. Ironically speaking, there is no danger of this
happening with a neural network. Neural networks are extremely flexible models
that combine combinations of characteristics in a variety of ways. Their
predictive accuracy can therefore be far superior to scorecards and they don’t
suffer from sharp ‘splits’ as decision trees do. However, it is virtually
impossible to explain or understand the score that is produced for a particular
application in any simple way. It can
therefore be difficult to justify a decision that is made on the basis of a
neural network model. In some countries it may even be a legal requirement to
be able to explain a decision and such a justification then must be produced
with additional methods. A neural network of superior predictive power
therefore is best suited for certain behavioral or collection scoring purposes,
where the average accuracy of the prediction is more important than the insight
into the score for each particular case.
Neural network models can not be applied manually like scorecards or
simple decision trees, but require software to score the application. Then,
however, their use is just as simple as that of the other model types.
Model Assessment
After building both a scorecard
and a decision tree model we now want to compare the quality of the models on
the validation data. One of the standard Enterprise Miner charts in the
Assessment node is the concentration curve and is shown in Figure 9. It shows
how many of all the bads in the population are concentrated in the group of 2%
(4%, 6%, …) worst applicants as predicted by the model. Sorting applicants
based on the scorecard scores will result, for example, in around 30% of all
the bads being concentrated in the 10% applicants that are considered the worst
by the scorecard model. The decision tree is only able to concentrate about
half as many bads in the same number of what it calls the worst applicants (the
10% decile is marked by the vertical black line in In summary, the
scorecard is assessed to be superior, because its curve stays above that of the
tree.
Defining decision rules for application approval and risk
management
Application approval and risk management do
not rely on scores alone, but scores do form the basis of a decision strategy
that groups customers into homogenous segments. These segments can then be
treated with the same action. For example, in the case of approval decisions,
customers are often classified using appropriate cutoff scores as approved,
referred for examination or rejected. Other segmentation strategies can
determine the limit amount that is assigned to a segment or the collection
actions taken. An important type of segmentation is the division of customers
into risk pools for the purpose of calculating
certain risk components: probability of default (PD), loss given default (LGD)
and exposure at default (EAD). These risk components are required by the risk
weighted assets (RWA) calculation mandated by the Basel II and III capital
requirements regulation. Analysts apply the scorecard and the pooling
definition to a historical data set. The long-run historical averages of the
default rate, losses and exposures can then be calculated by pool and used as
input into the RWA calculation. There are various ways to group customers into
segments using a scorecard. Often segmentation involves the setting of
thresholds. Sometimes analysts define these thresholds manually, and sometimes
they use an algorithm to automatically find a decision rule that is optimal in
a specific way. The way multiple thresholds are combined further characterizes
a decision rule. Typical examples of decision rules include policy rules
(exclusions), single score bins, multiple score bins and decision trees.
Deploying scores and decisions
Execution of decision rules can be done in
batch for all customers so that the assignment of each customer to a group and
an action is available in an operational data store for instant retrieval by
front-office software. Or, alternatively, the front-office software can
initiate execution of the decision rule to make a decision on an individual
customer, possibly using new or updated information supplied by the customer at
that time (online). The decision is then passed back immediately to the
front-office software. In either case, the decision rule is not executed by the
front-office software but through middle-layer software on a central server.
For existing credit customers, the batch option will be most commonly used,
since behavioral information derived from the customer transaction history and
other stored customer characteristics is typically more predictive than
information a customer might supply in the front office.
Is big-data analytics ultimate solution for airlines?
If the airline industry could be
described in two words, it would be "intensely competitive". The airline
industry generates billions of dollars every year and still has a cumulative
profit margin of less than 1%. The
reason for this lies in this industry’s vast complexity. Airlines have a multitude
of different business issues that need to be solved at once, such as globally
uneven playing field, revenue vulnerability, an extremely variable planning
horizon, high cyclicality and seasonality, fierce competition, excessive
government intervention and high fixed and low marginal cost. The low profit to turnover ratio of airlines
have been further exacerbated by growing low-fare competition, increasing
security costs, and frequent dynamic shifts in air travel consumer behavior.
The historical business model of many network airlines now appears to be unable
to support sustained profitability under any but the most favorable economic
conditions. The industry is at a turning point.
The market dictates an “adapt or die” policy, and the airlines that wish
to survive will face the challenge of having to make significant changes to
their current archaic business model. To do this requires far more allowance for
analytical technologies that would allow flow of consistent, repeatable and
reliable enterprise wide intelligence needed to tackle all the challenges the
industry is facing.
To ensure the best chance for
full economic recovery, airlines should fully leverage their most prolific
asset - data. Data used in conjunction
with innovative technologies that would allow the creation of an Enterprise Wide
Intelligence Platform, will provide the capabilities for a comprehensive intelligent
management and decision-making system throughout the enterprise. The ultimate
benefits of implementing and using an enterprise wide intelligence platform, together
with airline business acumen and experience would include timely responses to
current and future market demands, better planning and strategically aligned
decision making, and clear understanding and monitoring of all key performance
drivers relevant to the airline industry. Achieving these benefits in a timely
and intelligent manner will ultimately result in lower operating costs, better
customer service, market leading competitiveness and increased profit margin
and shareholder value.
Business challenges in airline industry
Key to successful deployment of technological advances in
airline industry is to be able to anticipate how the current business model
will change to survive in tough market conditions.
Some of the challenges that can be successfully addressed by
Enterprise intelligent Platform are:
- The need for accurate daily and weekly performance measurement reports (e.g. “flash/estimated” revenue, operating costs and net contribution reports for every aircraft’s actual flight per sector/route).
- The Need to better manage all aspects of risk.
- The Need for better impact analysis and more effective optimization of all resources as well as being able to produce accurate passenger-revenue forecasts,
- The Need for a holistic, 360 degrees view of the airline industries customers, suppliers, service providers and distributors.The Need for expense verification models in order to better control all industry cost aspects.
Performance Measurements
Airlines usually operate in a
globally competitive environment and therefore require prompt and accurate
enterprise performance measurements. Furthermore, airlines are volume driven
and small variations (passengers flown, fuel spent/bought, load carried) can
multiply into major effects – therefore appropriate and timely action is
critical. They also suffer substantial difficulties to produce daily/weekly
reliable performance measurements. Current airlines “legacy” IT systems such as
Revenue Accounting, require several weeks after a month end to generate revenue
results for every flight per sector/route.
Enterprise Intelligence Platform can automate production of daily
activity reports such as number of passenger flown per flight/sector, distance
flown, etc which can be used to provide estimated performance measurements such
as daily or weekly revenues for specific routes or sectors.
Risk Management
The
global airline industry has been subjected to major catastrophes over the past
years. It is accordingly imperative for
airlines to develop various risk management models and strategies to protect
themselves from negative impact of these types of events. Furthermore, due to
the global playing field, airlines often earn its revenues and pay its costs in
different baskets of currencies (e.g. USD, Euro, GBP etc). As a result there is
frequently a mismatch between the flow of revenue receipts and expenses of each
basket of currency - creating risk exposure reports.
Control and Verification
Airline
carriers require a number of control and verification models to be able to
control costs arising from its various operational activities. To enable this,
airlines have a pressing need for a complete and integrated repository of
flight information data gathered from all its disparate business units. This
will enable computation of various efficiency analytics - e.g. planed fuel
usage compared with actual fuel usage per aircraft, crew utilization (roster
optimization). These issues could also be fully addressed by consolidating and
analyzing relevant flight and aircraft data. In turn this would help to create
a 360 ° view of each flight and aircraft, allowing the business users to
dramatically improve their control and verification systems.
Airlines
require the development of an effective and holistic forecasting model to regularly
assess the impact of options and alternatives such as increasing aircraft seats
available, adjusting fares, introducing new routes etc. Forecasts should also
take account of actual statistical trends and results e.g. actual passengers
carried and actual average fares earned. Such forecasts should then
be compared against budgets and prior year performance.
Holistic customer view
Airlines would greatly benefit
from knowing and understanding its business environment along some of the key
business issues, such as performance, behavior, risk, profitability, etc. Using
customers as an example - the
main objective would be to enrich the knowledge about individual customers
leading to new strategic customer segments. This intelligence would
allow airlines to reap the host of benefits such as successful, targeted
customer promotions, cross-selling and up-selling campaigns for different
flights and booking classes leading to improved yield and revenue. For example,
it would give airlines the power of knowing to limit discounts on flight routes
which are usually over-booked, allowing the large number of passengers to
compete for high profit seats immediately prior to departure. Such
multidimensional views of the business can help the airline to better serve its
customers through more effective, efficient and personalized service, receiving
in return customer loyalty, support and market share, all leading to higher
profitability.
RFM Segmentation
Even though RFM segmentation is well known in retail industry, and
basic premise is that by knowing recency, frequency and value of the purchase
you can be in good position to start figuring out specific customer in terms of
its value, purchasing behavior and its loyalties. However, same logic can be
applied for any phenomena that we trying to predict. Therefore, knowing how
often something happens, how recently its happened and its voracity – has same
type of predictive power as it has in retail context. And whenever I used it
for predictive modeling- RFM would always come as one of the top predictors.
So, let me delve deeper in explaining basic principles of RFM method.
RFM segments the customer base based on recency of purchase (R), frequency
of purchase (F) and monetary value (M). Recency
parameter is the most powerful of the 3. In forecasting models latest time
series often has the highest weighting and is the most predictive of the next
forecasting value. Second most powerful is the frequency as long as the definition of the frequency is limited to last month or quarter and not over entire
life-span of customer relationship. Least powerful is the monetary value. Since the total value in the period of time is
directly correlated with frequency it
is advisable to use an average value.
There are several different ways to calculate RFM groups and scores
and below is the classic approach:
First create 5 segments based on the recency, dividing the data file
into 5 exact quintiles, where the contacts with the most recent Transactions
(i.e. in the top 20% of the file) are given a
recency value of 5, then the next 20% are given a recency value of 4 and
so on. Then, each of those quintiles, segmented into 5 further quintiles based
on the frequency value for each
contact where the contacts with the highest transaction frequency value are of
5, then the next 20% is given a frequency value of 4 and so on. Finally, each of these segments is then
segmented into 5 further quintiles, based on the monetary value of each
contact; i.e. the total amount which all that contact’s transactions add up to.
Those contacts with the highest monetary values (i.e. in the top 20%), are
given a monetary value of 5, then the next 20% are given a monetary value of 4
and so on.) At the end of this process,
you will have 125 segments with a RFM group between 111 and 555 with the same
number of contacts within each segment; and each contact will have a RFM score
of between 3 and 15.
An alternative approach is to still calculate RFM Groups/Scores using
quintiles, but by using the Independent RFM Quintile approach, not just the recency but also the frequency and monetary values for each contact are calculated across the whole
data file and are not dependent on any of the other values/RFM factors or any
other quintile. Another approach is to use user-definable bands for each
criterion (i.e. each RFM factor) in order to determine what recency, frequency and monetary value that should be given to
each contact. Even-though RFM segmentation can be used on “stand-alone” basis,
I always tend to incorporate it with other demographic and affinity variables
in order to have more holistic view of the segment's make-up.
I have coined my own approach
that I often use which is somewhat different of the classic approach and it
goes in following way:\
1.) Create variable Total Spend for for
each customer
2.) Create variable Total number of
visits for each customer
3.) Divide both variables into 3 equally spaced bins, based on frequency
– 1st bin would be lowest 30% of all
customers in regard to spending (and visits – separate variable)
4.) Evaluate
each customer in terms of in which group he belonged (for that time) in terms
of his total spending, and total visits, and label him for that group (Example:
variable “FRM_Spend_label” would have
values “L”, “M” and “H”. If amount of his total customer spending for 12m is
within threshold fits within second bin – give him a value “M” (medium) in
variable “FRM_Spend_label”
5.) Do the
same thing for visits, creating a new variable “FRM_visit_variable”.
6.) Do
slightly different thing for “Recency” – starting from the same endpoint as it
has been done for “spending” and visits – go behind only 3 months and not 12.
Then, do the following: if customer did purchase in month 1 (the most recent
month) give him a value “H”, if the most recent purchase was in month “2” –
give him a value “M” and if the most recent purchase was in month “3” – give him
value “L”.
Note – it might happen that most of a customers have some
sort of purchase in all months in which case it would be advisable to raise
threshold above “0”. In other words call the recent purchase only if monthly
total is above some specified amount bigger than “0”.
7.) Combine
all three FRM dimensions together into single variable where values would be
combinations of “H”, “M” and “L”. If value is “HLH” it would mean that customer
falls in the top group of customers in terms of their number of visits to the
stores, it means that customer wasn’t in the store (with purchase larger than…)
for a month and it means that customer falls in the top group of customers in
terms of their total monetary value that they bring to the company.
8.) In last
step I deploy “19 +1” rule, where i retain top 19 combinations based on its frequencies
and all the other combinations I drop into “other” category, so that my FRM
variable doesn’t have more than 20 distinct values.
Hope this helps!
Subscribe to:
Posts (Atom)