Youth Involvement in an Innovative Coconut Value Chain by Mwalimu Menza
dff
1. ABOUT COMPANY
[24]7 Inc. describes itself as a "Predictive Customer Experience Solutions Company‖.
Headquartered in Silicon Valley, USA, [24]7 Inc. is a global online technology and operations
company with over 9,000 employees worldwide.
DETAIL
Type Multinational
Industry Predictive Customer Experience Solution
Founded April 2000
Headquarters Campbell, California, United States
Key People PV Kannan, CEO & Co-founder, S. Nagarajan, COO & Co-Founder
Employees 9000
[24]7 Inc. 9 contact centers are located in India (2 delivery centers), Philippines (4 delivery
centers) and 1 each in Guatemala, Nicaragua and China, which deliver services in nine different
languages to customers across North America, Europe, Asia and Australia. [24]7 manages over
2.5 Billion interactions through its 9 global contact centers and has transformed over 20 million
of these interactions into predictive experiences using its Px platform. [24]7 Inc. develops online
predictive technologies through integration of SaaS and contact center operations which enables
them to deliver customer service in real time. [24]7 transforms customer interactions of large
telecom, financial services, retail, technology and travel companies from traditional contact
channels such as phone and email to predictive and personalized online interactions.
LOCATIONS
USA - Campbell, California
UK- Covent Garden, London
Australia – Sydney Delivery Centers
India – Bangalore & Hyderabad
Philippines – Makati City (4 facilities)
Guatemala – Calzada Atanasio Tzul, Guatemala City
Nicaragua - Ticuantepe, Nicaragua
China – Shanghai
2. INTUTIVE CONSUMNER EXPERINCES
At [24]7, goal is to make it simple for consumers to connect with companies to get things done.
Our software platform and consumer-focused design leverage ―big data‖ to help companies
anticipate, simplify, and learn—and by so doing, provide smarter and more effective multi-
channel customer service. A new framework is needed. Three pillars define our unique approach
to reinventing consumer focused customer service:
ANTICIPATE
We make consumer interactions intuitive by predicting what people want and giving them an
easy way to get things done in the channel of their choice. Our predictive technology leverages
big data to provide companies with real intelligence about their consumers based on identity,
location, behavior and analysis of massive amounts of data from historical interactions. Our
technology delivers this predictive power to your online, speech, and mobile self-service
channels, as well as to your chat and voice agents. The result is a truly personalized approach—
one that‘s completely relevant to the context of the moment and the tasks your consumers are
trying to complete.
SIMPLY
We make things easy for consumers by seeing the world from their perspective. Our
sophisticated technology delivers a simplicity that allows people to engage in an intuitive way
3. with the companies they do business with and get the answers they need. Our software initiates a
set of interactions that provide an immersive experience across channels, devices and modes of
input—from online, speech, and mobile self-service to chat and beyond. It‘s all about removing
potential pain for consumers and replacing it with magical efficiency.
LEARN
Our software learns by using big data that enables enterprises to continuously understand more
about their consumers. It‘s an interactive intelligence that powers insights and improves
consumer experiences by building on the past—and ensuring that every new interaction, with
each individual, gets better all the time.
PLATFORM
Px Online
Px Online powers web self-service apps that maximize the effectiveness of your online
investments. Px Online monitors visitor behavior, detects difficulties, and leads
customers step-by-step through even the most challenging journeys. Through advanced
statistical models of visitor intent and engagement; Px Online anticipates customer
behavior and determines the optimal interaction strategy. By simplifying the experience,
resolving issues and completing transactions in web self-service, Px Online dramatically
reduces the need for live chat assistance and calls into the contact center. Px Online also
learns and drives improvements in your web site design and content by pinpointing the
journeys and customer segments that are under-serviced. Px Online delivers a distinctive
user experience that lowers customer effort, increases loyalty, and reinforces your brand.
Px Speech
Px Speech redefines IVR by enabling speech self-service that‘s conversational, intuitive
and easily integrated with your online and mobile channels. Through advanced statistical
models of caller intent and experience, Px Speech anticipates caller requests and
proactively adapts the conversation flow. Px Speech links real-time interaction data and
customer profiles across all channels to build the full context of every customer journey.
Px Speech drives shorter call durations, increased self-service adoption, and a higher
level of customer service for your phone callers.
Px Mobile
Px Mobile powers apps on Smartphone and tablet devices that automate customer
journeys using a combination of speech and touch modalities. Px Mobile delivers an
experience that leverages enhanced capabilities such as speech commands and location
information while adapting to the limitations of small screen sizes. Through advanced
statistical models of user intent and engagement, Px Mobile anticipates customer
behavior and determines the optimal interaction strategy. Mobile apps integrate with
4. enterprise data sources and third-party systems enabling location-aware, information-rich
solutions.
Px Assists
Px Assist provides the intelligent interaction platform for live chat and voice assistance.
Through predictive analytics and smart routing, Px Assist targets the customers most
likely to benefit from live assistance and connects them with the agents best able to
service their needs.
For your customers, Px Assist:
Provides instant access to skilled agents
Maintains information entered or derived from self-service
Supplements live interactions with rich content pushed to smart devices
For contact center managers and agents, Px Assist:
Enables collaboration, workflow management and performance management.
Combines chat analytics, speech analytics, and customer surveys to measure
interaction quality, track agent skills, and improve predictive models
Improves efficiency by transferring all relevant interaction information to the
agent in real time which also drives significant improvements in customer
satisfaction.
Helps managers to create training plans that drive agent performance,
productivity, and skills development.
5. SOLUTIONS
INDUSTRY SOLUTION
COMMUNICATION
BILL INQUIRY
FINANCIAL SERVICES
FRAURD INQUIRY
TRAVEL
FLIGHT RECOMMODATION
RETAIL
INVENTORY CHECK
TECHNOLOGY
TECHNICAL TROUBLESHOOTING
6. CASE STUDIES – NEW CUSTOMER ACCQUSITION
INTRODUCTION
A customer profile is a snapshot of who your customers are, how to reach them, and why they
buy from you. Deeper understandings of new and existing customers by collecting customer
profile increases sales. This information is used to segment prospects more effectively, as well as
send targeted communications more precisely to them.
Accurate market segmentation is essential in successfully acquiring customers. Making the right
strategic and operational decisions can result in higher lifetime customer value and increased
customer satisfaction.
Two main goals of market segmentation have remained constant over the years. At a strategic
level, segmentation should help an organization rapidly evaluate new opportunities. At an
operational level, segmentation efforts should yield information to help craft successful
marketing offers to acquire prospects.
The statistical scoring model generates a score value that indicates the probability of conversion
based on segmented group characteristics. This provides a foundation for more advanced
selection of potential customers. Statistical score value predict prospect behavior more accurately
than judgmental score value, but require much more data, and staff allocation.
Page
TASK
To increase Sales per Hour (SPH) by applying analytics.
CHALLENGES
Customer Acquisition based on demographic and geographic characteristics, and time-
periods optimization management for different segmentation approaches to define hot
prospects.
Substantial difference in prospects behavior across various geographical parameters -
time zone, province and postal code.
Identify hot acquisition targets by evaluating future potential of each segment.
METHODOLOGY
I. Selecting the Data Source
Data for modeling can be generated from number of sources. Those sources are two types:
Internal and External. Internal data sources are those that are generated through activities
undertaken by contact center solution providers like dialer transaction, agent performance etc.
External data sources includes lead information, large prospect database etc.
The data sets used to arrive at Analytical model for analysis are:
Lead Information
Agent Background
Dialer Transaction
Job / Campaign Session
7. II. Preparing the Data for Analysis
Data preparation is the most important step in the analysis and model development process. In
many cases it was necessary to combine data from several sources to create the final data set.
Data cleansing is most important and time consuming process. This involves looking for and
handling data errors, outliers and missing values.
Data Errors and Outliers: An outlier is a single or low frequency occurrence of the value of a
variable. Determining whether a value is an outlier or a data error is an art as well as a science.
Having process knowledge of data is best strength.
Missing Values: Missing values are present in almost every data set. The fact that a value is
missing, however can be predictive. It is important to capture that information.
III. Transforming the Variables
Before making segmentation, there is a need to understand the data. For this purpose, variety of
numerical summaries (including descriptive statistics such as averages, standard deviations, pivot
tables and so forth) and study of distribution of the data is needed. Applying graphical and
visualization tools to find new insights of different variables in the database and their importance
for effective data analysis is useful. Sometimes, the format of a variable in the raw database is
not sufficient for analysis. Hence, transform variables in accordance with the requirements of the
algorithm chosen to build the model.
IV. Selecting the Contributing Variables
In practice, all variables were not taken into consideration in the segmentation approach. One of
the reasons being, the time it takes to build a model increases with the number of variables and
another reason would be, blindly including extraneous columns which can lead to models with
less rather than more predictive power. Some techniques are used to reduce the number of
variables that eliminate marginal and unproductive variables.
V. Segmentation Based on Contributing Variables
Prospect segmentation is a process that divides prospects into smaller groups called segments.
Segments are to be homogeneous within and desirably heterogeneous in between. In another
words, prospects of the same segments possess the same or similar set of attributes. For example,
forecasting on the basis of agent cluster rather than individual agent as a predictor variable. This
yields more accurate results, which are parsimonious and are also easy to understand.
VI. Training and Testing Dataset Generation
A subset or sample of the database needs to be selected to build models, which are capable of
representing the whole population. Various techniques are available for generating the training
and testing datasets, such as, stratified sampling, cluster sampling, sampling without
replacement. Method selection depends on selection according to the nature of data. Generally, a
stratified sampling technique is used to split into training and testing data set.
8. VII. Scoring Model Building
Model building is an iterative process. There is a need to explore alternative models to find the
one that is most useful in solving the business problem. Searching for a good model may require
us to go back and make some changes to the data or even modify the problem statement. We
used logistic regression model. The logistic regression model computes the probability of the
selected response as a function of the values of the predictor variables.
VIII. Scoring Model Testing
We used different techniques to evaluate the model accuracy on test data set. One of the key
measures is the ‗Confusion Matrix‘. This shows the extent of type I, type II and overall errors.
We consider a model is perfect if it shows more than 80% accuracy during testing. We also use
different methods to test the model, such as, chi-square, p-value, odds ratio, and ROC curve.
IX. Model Evolution
The model generated from the testing phase would not give results for future lead files as
expected. If the expected results generate a certain percentage of error beyond the acceptable
threshold, we rebuild the model considering the recent dataset along with the earlier dataset. The
process was repeated till a stable model was produced.
RESULT
SAS Statistical Analysis Software was chosen as profiling, segmenting and scoring a
prospect for this analysis. Predictive segmentation methods based on statistical analysis
produced three optimal clusters on agent performance. These three different clusters
classify agents as Very Good, Good and Average.
People living within the same geographical boundaries exhibit similar buying patterns.
Segmenting markets along geographical boundaries can lead to more specialized and
focused marketing approaches.
Each segment‘s propensity to buy the services was used to evaluate the future potential of each
segment.
Past call transaction data was used to know the day of a week conversion trend and hour of a day
conversion trend. This trend was utilized to call a prospect at proper time. A time zone mapping
exercise was conducted to guide the dialing strategy.
A standard definition of segments was developed which could be used for selecting potential
prospects. Various groups could now more effectively collaborate around tactical dialing
strategies.
SCORING MODEL
The scoring method was used to find out a score value of a prospect based on a collection of
evidence as a whole while considering numerous dimensional groups. This provides a foundation
for more advanced selection of potential prospects. Statistical score value predict prospect
9. behavior more accurately than judgmental score value. Logistic Regression statistical model was
used to score a prospect.
A decile analysis was done to know the top scorer where conversion is high. It also helped to
make an overall segment a prospect as hot, warm and cold. A graphical plot of decile chart is
given below to visualize hot, warm and cold.
STRATEGY DEVELOPMENT
Whom to call: Prospect selection strategy
Predict prospects score (based demographic variables, geographic variables) using
scoring model. Find out which decile it belongs to and classify as hot, warm or
cold prospect.
When to call: Dialing Strategy
Select weekday using ―Weekday time zone wise conversion trend‖ table
Select hour of a day using ―Hourly time zone wise conversion trend‖ table
Initial stage makes ‗X‘ number of dials to a particular prospect phone number to
make maximum 3 connected dial
If prospect is not connected then make second stage ‗Y‘ dials after keeping an
interval of time gap
Who will handle the call: Agent allocation strategy
Initial stage selects Skill (Very good, good and Average) based agent allocation to
convert a prospect to customer
Second stage allocates mixture proportion of very good, good and average agent
group to hot, warm and cold prospect.
The product strategy group to define the product specifications and develop an
acquisition plan for a new product or service.
The campaign group to identify the best geographical area and partnership strategy.
The Analytics research team can drill down deeper into the segment for new insights
about prospect attitudes towards support.