Different organizations operate at different stages of analytic maturity – from almost non-users or rudimentary users of data to very advanced users of predictive models and algorithms and data-driven decision optimization. Competitive advantage increases with this degree of data intelligence deployed – beginning at the level of Hindsight based reactive decision making (Reacting to what were the business outcomes), to Insights (why those outcomes occurred - understanding performance drivers post facto) to Foresight (what is likely to happen - predictive modeling based proactive Decision making) and Optimization (how best to act - prescriptive analytics - scenario analysis, simulation and Optimization).” QPS helps businesses at every stage of analytic maturity and assists them in progressing through the stages quickly.
Our solution offerings have capabilities to mine structured and unstructured data of any volume and variety to extract actionable insights in addition to developing, deploying and managing custom models and scorecards. Our solutions help clients make fast, data-driven, strategic and tactical decisions across key business functions like marketing and customer relations management, risk management and operations. We work with organizations across geographies, Industries, size, and analytic maturity.
Our solution offerings address many key business questions using data and help organizations make quick, informed decisions based on actionable quantitative insights.
- What is going on in my business? How are my key metrics looking for different geographies, branches, campaigns, etc.? How can I enhance revenue from existing customers?
- Who are my best customers? Which segments are the most profitable and which are draining my resources? Are my promotions and services well aligned to the needs of my best customers?
- Which customers are most likely to attrite? Which customers are most likely to stay loyal?
- How can I measure the effectiveness of my loyalty programs? How does it impact my revenue and profit?
- How can we respond more quickly to changes in the competitive landscape?
- Where can I cut my operating and marketing costs? How do I know what my customers are saying in social media and what’s my brand perception?
- What are my underwriting risks for any new applicant? How do I identify fraudulent applicants?
- Which customers are most likely to default?
Our analytics consulting group has been helping clients with all of these critical business questions and many more over the years. A broad overview of our most sought after analytics offerings are given below:
Predictive Modeling, Optimization and Other Advanced Analytics Solutions
QPS's solutions in this area enable businesses to make forward-looking decisions. This is powered by a discovery and simulation-based approach using data. QPS's most frequently implemented advanced analytics and predictive modeling solutions help organizations to predict outcomes, understand key performance drivers, quantify impacts of those drivers and make optimal decisions on those levers and monitor outcomes helping further dynamic fine-tuning and adjustments.
Model Development (predictive scorecards and other models)
It provides a business an immense advantage to be able to predict likelihood of an event of interest (models to estimate probability like logistic regression) or estimate expected value of certain business metrics of interest under different scenarios and/or at a future date (e.g., regression models and time series models), or how much time it will take for an event to take place (survival models) - all based on information available currently. To achieve this, different statistical models and ML (Mission Learning) algorithms are used. These models are developed using all relevant data available within the business and all other relevant and available external data. Our solutions in this area support all model development needs of any business across all functions, namely, marketing and CRM, risk management, operations, logistics and others.
Response and Conversion models (new acquisition campaigns): Models to rank-order customers in terms of their predicted likelihood to respond to a new acquisition campaign – developed using historical data from previous campaigns on similar prospect databases – helps improve ROI on new acquisition campaigns.
Cross-sell / Up-sell models: Models to rank-order and identify best targets (in terms of predicted likelihood of accepting offers) for cross-sell campaigns - improves ROI on cross-sell campaigns. These scorecards are usually built combining demographics, transactions and behavioral data coupled with outcomes from similar campaigns run in the past.
Attrition models: Models to predict risk of attrition by customers (predicted probability of attrition), basis their behavioral, transactions and demographic characteristics – helps portfolio managers in identifying retention targets and designing custom retention plans.
Transaction Scorecards (Marketing/CRM specific): Models to segment customers into different categories on the basis of their predicted probability to manifest a specific behavior (of suitability for marketing or CRM actions, e.g., spend, cross-sell/up-sell response, attrition, etc.) computed using their card transactions characteristics (merchant type, location, amount, frequency, recency, consistency, etc.) – great enabler in customized campaign strategies, loyalty management, retention management, spend triggering marketing strategies, etc.
Application Scorecards: Scorecards to predict risk of default within one year of card/loan approval and activation/disbursal, based on application data (mostly demographics and bureau information available at the time of application processing) and payment performance of approved cards/loans for about six months to one year after activation/disbursal – used in application screening and /or initial credit limit assignment and collection queuing.
Behavioral Scorecards: Scorecards to segment existing customers with long enough time on book into different credit risk categories on the basis of different behavioral characteristics (purchase, payments, limit utilization, etc.) - deployed in credit limit management, pricing, authorization, collection prioritization, etc. and any other risk management tasks.
Collection Scorecards: Collection scorecards predict likelihood of success (to collect) and/or proportion (of outstanding amount that may be collected) given default – these scorecards are built using historical data on collection, customer detail and customers’ application, transactions and behavior data.
Transaction Scorecards (Credit risk specific): Scorecards to quantitatively predict risk of default using their card transactions data (merchant type, location, amount, frequency, recency, consistency, etc.) – significantly bolsters the behavioral scorecards used in risk management functions by incorporating most recent characteristics of customers as revealed in their transactions.
Loss forecasting and LGD models: Apart from the credit risk scorecards (models of probability of default at different stages of customer lifecycle, using different types of customer data), we also provide loss forecasting and LGD (loss given default) model development solutions. Our loss modeling solutions include but are not limited to traditional approaches like net flow rate analysis and vintage loss curve analysis to advanced econometric/statistical models and machine learning algorithms as well as application of simulation and scenario analysis techniques.
QPS offers an efficient Model Implementation engine making it easy for the end users to use models to assign scores to customers. Using our simple GUI-driven model implementation engine, front end users can implement any model to compute customer scores only through a few clicks of the mouse.
Post-implementation validation and monitoring is an immensely important part of a model driven decision system. It is critical to ensure that the models developed using data from a certain time period from past, continue to stay relevant for the current data and performs with required level of prediction accuracy. In case a model doesn’t pass the tests, an appropriate remedial strategy needs to be implemented depending on the findings from the model evaluation exercise.
QPS provides a comprehensive model evaluation, validation, monitoring and management service (encompassing all available and recommended statistical tests and other evaluation methods) for both models developed by QPS experts as well as by any other provider or model developed internally by an organization.
Some customers are more valuable than others, and marketers cannot afford to treat them all alike. Customer segmentation and profiling exercise provides a data-driven efficient way to customize marketing, CRM and credit risk and collections management strategies according to important segment characteristics of business interest.
In the exercise the customer portfolio is segmented into a few smaller groups based on all available customer data, past campaigns and any other relevant information using suitable data mining algorithms (e.g., hierarchical and k-means clustering) and a thorough analysis and profiling is done on each group to know them closely and across all dimensions of interest.
Our Loyalty Analytics solutions provide a comprehensive coverage of all loyalty related business questions right from measurement of customer loyalty and analysis and modeling of loyalty drivers (using Structural Equation Modeling) to evaluation of existing loyalty programs and recommendations on modification, as well as deep customer insights to support designing of new loyalty programs.
Our Customer Lifetime Value analysis and modeling solution incorporates all key factors that determine the long term expected value of a customer to the business. The solution involves modeling and computation of attrition risk and survival likelihood over a given period of time (usually three to five years), all revenue flows generated by the customers and all relevant cross-sell/up-sell probabilities (for a multiproduct organization like a bank, insurance company, etc.). In case of organizations where individual customers are not tracked through a customer ID in a CRM system, our solution is modified by modeling and computing the key parameters at a segment level including any inter-segment movement dynamics.
Our solution offerings to help optimize marketing spend allocation across channels cover both traditional market mix modeling services (mostly applicable for off-line marketing) with channel and geography level aggregate time series data as well as cutting edge bottom-up attribution modeling with highly granular, often individual level data (mostly relevant for online/digital marketing).
Our fraud analytics solutions cover fraud at different stages of customer lifecycle, namely at the point of application (usually first party fraud) as well as transaction frauds (usually third party fraud). Our analytics experts apply state of the art techniques (several different supervised and unsupervised machine learning algorithms and statistical modeling techniques) for a comprehensive mining of all available data for anomaly detection, extraction of rules and patterns to flag potentially fraudulent behavior. With system integrated implementation, this consistently delivers a significant lowering in fraud rate while keeping false positive cases in check. Our fraud analytics solutions result in consistent early detection and prevention of fraud, reduction in fraud losses from high exposure activity, reduction in false positives by better classification of transactions queued for review, enhanced customer protection, satisfaction and reduced reputational risk.
There are scenarios where an organization has plenty of data collected over a period of time but no well-defined business question is posed to make full use of this data or the organization doesn’t have a well defined analytic strategy to make regular use of this rich information repository. In such cases, an open ended data mining exercise is carried out to search for any non-trivial, previously unknown actionable insights and patterns. Even when an organization’s data is used for certain specific purposes, such additional open-ended (KDD) exercise is strongly recommended. These projects often start with a semi-structured approach which is allowed to evolve based on findings as the analysis progresses.
Under this line of solution offerings, QPS analytics group regularly works on projects to extract actionable insights and patterns from data by intelligent mining and analyses.
Transactions Analytics, suitable for banking (cards), retail and any other business domain where a customer can have multiple transactions over time, is one of our highly appreciated offerings from KDD/PDD line of solutions.
Transaction Data is a goldmine of information about customer behavior - this data is the most dynamic, up-to-date and candid source of information about the customers – their preferences, activity patterns, spending habits, and any changes therein. If appropriately analyzed and mined, this data can reveal highly valuable actionable insights to facilitate data-driven, optimized strategic and tactical decision making in banks. In most banks this data remains underutilized in the sense that very little analysis is done and insights are not extracted beyond some basic risk indicators to inform collections.
This solution, which involves a cross functional data integration exercise followed by multidimensional mining and modeling, is another frequently sought after service under our KDD/PDD line of offerings.
In this, we integrate data from all different sources within an organization covering all business functions and customer dimensions. This integrated customer data provides a 360-degree information coverage on the customers’ information and serves as a single source of customer insights. This consolidated customer view can then be mined for actionable insights and patterns to support many different business activities like cross-sell and retention, offer alignment and campaigns, product designing and loyalty programs, credit risk management and collections, and any other customer related strategic and tactical decisions.
Big Data, includes (but is not limited to) structured and unstructured data of any volume and variety, external openly available data gathered from the web sources like social media, blogs, forums, etc., internal data like call center transcripts, customer on-boarding document, logs from web portals, click streams, etc. come under the broad definition of Big Data. And all of these can be analyzed in conjunction with or independently of the traditional internal data (like customer demographics, transaction history, etc.) stored by organizations.
The key aspects of Big Data and Social Media Analytics are data collection (external open data), data cleaning and management (both internal and external), analysis and mining to extract actionable insights, data visualization and quantitative analytics/modeling/machine learning using unstructured texts (through data vector extraction) as well as structured data. We handle the end-to-end process, through our multi-stage and modular Data Sciences engine, which includes advanced NLP (Natural Language Processing) and machine learning capabilities apart from web crawling tools. The engine resorts to Hadoop ecosystem and MapReduce based distributed computing whenever warranted and is thoroughly customized for different projects.
Our first level solution offerings in this space cover customer and market pulse monitoring, competitive intelligence on social media and customer interactions analytics, B2C lead generation based on customers’ expressed needs in online social platforms.
- Analyses and scores customer satisfactions(CSAT) on own brand across key demographic segments and geographies
- Identifies and wherever possible, quantifies, key drivers of CSAT and loyalty- which aspects and features do customers like and which ones they don’t – helps product design, communication and marketing strategy
- Creates attrition alarms from social media posts and trends
- Measure and Score brand sentiment and analyses major factors affecting sentiment
- Tracks economic sentiment and market trends across key segments and geographies
- Analyses and scores customer attitude towards competitors’ products and brand
- Analyses key customer sentiment (positive or negative) drivers for competitor(s) products/services – this also helps product design, communication and marketing strategy
- Tracks market sentiment trends for competitors’ products/services
- Customer Interactions through any channel generates a wealth of information mostly hidden in unstructured text data (or voice data easily convertible to text)
- Call center transcripts
- Email messages
- Chat transcripts
- Activity logs (from server) on web-site, click streams
- Customer expressions on social media, blogs, forums, review websites, etc. are also external sources of valuable information about customer experience – a large part of these are openly available
- Highly valuable actionable insights are extracted from these voluminous, varied and mostly unstructured data by using a combination of smart data collection and management, Advanced Text Mining/NLP methods, statistical and data mining algorithms
- Identification of ‘expressed needs/interests’ to boost sales through our optimized keyword based web crawling
- Customer need and direct B2C marketing opportunity identification
- Enhance overall marketing strategy and product design through a clear view of customers expressed needs and preferences
Turnkey Analytics Practice Set-up and Consulting
For organizations at an early stage of analytics use or data-driven decision making, QPS helps develop and execute a rigorous strategy framework to expedite and optimize the evolution. This involves steps to create an integrated analytics data mart, data retention and management plans and a comprehensive plan for each Business Unit to utilize data through fast actionable insights, predictive models and ad-hoc drill-downs. If required QPS experts may also help develop all of the components.
Business Intelligence (BI) Dashboards and Insights - Performance Monitoring
To monitor business performance and drill-down to understand how different factors drive performance is an extremely important enabler of data-driven decision management. Our BI and Insights solutions address this with detailed reports with multidimensional drill-downs and a mix of industry standard metrics and custom defined innovative ones.