We can see trends where customers with certain subscriptions are less likely to pay on time. This website uses cookies to make your browsing experience more efficient and enjoyable. Although information comes from multiple sources, it is imperative to maintain a constant data flow. Figure 1 below shows the model that we built. There were lots of reviews and test cycles to demonstrate the accuracy and the high level of security that we have. Predictive analysis helps marketing teams invest their resources wisely and set KPIs that align with total business value. How do we help the collections team prioritize contacts and decide what actions to take? Predictive analytics applications optimize the allocation of collection resources by identifying the effective collection agencies, contact strategies, legal actions to increase the recovery and also reducing the collection costs. We mostly contact only customers who need help paying. Continuously optimize the efficiency of our collection strategies and business processes. We didn’t have many insights to speed up how quickly we recovered payments owed or to improve our credit and collections processes. In combination with well-defined business processes, the adoption of technology for predictive analytics can have a significantly positive impact on an organization’s ability to enhance collections efficiency. Azure Machine Learning Studio makes it easy to connect the data to the machine-learning algorithms. We plan to add additional scenarios, use cases, data sources, and data-science resources for even more insights. The collection process involves all payments—not just late ones—so streamlining and refining a process of this scope is important to our success. As an increasing number of B2B companies are learning, this is the foundation of a next-generation collections function. Agents with moderate experience, training… And the quicker we collect payments, the quicker we can use that money for activities like extending credit to new customers. From this data, we create categories or features like customer geography, products purchased, purchase frequency, and number of products per order. Improve customer satisfaction by reaching out to specific customers with a friendly reminder, while not bothering those who typically pay on time. Collection analytics gives valuable information about the customer which can help develop varied collection strategies in different stages of obtaining due payment. With data science, Azure Machine Learning, and predictive analytics, we improve customer satisfaction, empower our collections team, optimize the efficiency and speed of our collection operations, and we’re more predictive and proactive. Why is this understanding important? We prioritize those who’ve paid late in the past. We know that if customers are in a country/region that’s experiencing economic crisis, there’s a chance they’ll need help paying on time. To train and refine the model, we overlay it with five years of historical payment data from our internal database. WNS's research shows that a one-day improvement in days-to-receive could unlock as much as USD 8.6 Billion in cash in the case of automotive industry (for players with annual revenues in excess of USD 500 Million). Repurpose that money for other short-term and long-term investments. Figure 2 shows the iterative process that we use and the different roles employed at each stage. Each … But say you’re starting from scratch. In the most critical cases, companies may experience a swelling of the portfolio of receivables more than 90 days past due and a low debt recovery rate. Credit and collections team members often come across the same questions over and over. We keep learning all the time as we iterate. Data Science for Beginners compares an algorithm to a recipe, and your data to the ingredients. This website uses cookies to make your browsing experience more efficient and enjoyable. There are primarily three stages of collection, which can be broadly classified as the early stage, the mid-stage and the final stage of collection. To speed up the process of answering these recurring questions, we built a chatbot. When we onboard new customers, we can correlate certain trends to them quite accurately, based on what we’ve seen with other customers. Predictive analytics models combine multiple predictors, or quantifiable variables, into a predictive model. Choose your own level of cookies. The candid answer is that they are unable to make breakthrough improvements in performance through operational excellence alone. Predictive analytics is easier with ready-to-use software options that offer embedded predictive modeling capabilities. Intellicus predictive debt collection analytics solution enables you to curb debts, predict collection, and enhance overall portfolio performance. Often, a collections team begins by extracting a bad debt report from the ERP; then uses agebased categories to segregate debt and assigns them to collectors based on their experience. The names of actual companies and products mentioned herein may be the trademarks of their respective owners. Within two months, we easily set up a predictive model with Azure Machine Learning that helps the collections team prioritize contacts and actions. We use the eXtreme gradient boosting (XGBoost) algorithm—a machine learning method—to create decision trees that answer questions like who’s likely to pay versus who isn’t. The following steps, as shown in Figure 3, show how the chatbot works: Now, field sales, operations, and collectors can see the latest information about customers they interact with and detect issues. Driving healthier cash flows and better customer relationships with lower revenue leakage — at lower cost. We have more than 1,000 trees. Some customer types and geographies benefit from phone or face-to-face contact much more than others. With data science, Azure Machine Learning, and predictive analytics, we improve customer satisfaction, empower our collections team, optimize the efficiency and speed of our collection operations, and we’re more predictive and proactive. This begs the question: if the business impact of a better performing collections function is so compelling, why aren't organizations turning collections challenges into cash flow and revenue assurance opportunities? Aligned with our mission of digital transformation, these insights join data, technology, processes, and people in new ways—helping the collections team to optimize operations by focusing on customers who are likely to pay late. You can find out more about which cookies we are using or switch them off in settings. Prior to collections, analysis of past and present payments (such as balance amounts and payments in the end-credit period) can materially reduce the incidence of bad debt. Predictive analytics is a decision-making tool in a variety of industries. For example, suppose an invoice is due on Saturday, or a customer in a particular country/region tends to pay late, and the average invoice is, say, $2,000. Solving the machine learning problem itself took us only about two months, but deploying it took longer. WNS provides us a blend of functional expertise and process capabilities which spans across our diverse portfolio. Say you are going to th… If a computer could have done this prediction, we would have gotten back an exact time-value for each line. We have also started to expand our scenarios into areas that are adjacent to credit and collections: sales and supply-chain features. High-level view of the solution. Predictive analytics is an area of statistics that deals with extracting information from data and using it to predict trends and behavior patterns. Allow cookies. This is done by understanding that not all delinquent accounts are the same. These are the technologies and components that we’re using for our solution: Figure 1. This involves compiling non-traditional customer records and using the data to determine customers’ ability to pay on their balances. We asked things like: To help with these and other questions, we use data science and Microsoft Azure Machine Learning as the backbone of our solution. Figure 1 quickly summarizes our solution. We need to contact fewer than 40 percent of customers. What do you do when your business collects staggering volumes of new data? Improving Debt Collection with Predictive Models FICO scores will be soon improved by predictive analytics. When done right, the model enables collectors to contact the right customer at the right time; with the right messaging and most effective payment options. You can use predictive analytics to understand a consumer’s likely behavior, optimize internal processes, monitor and automate IT infrastructure and machine maintenance, for example. It also reduces the cost of customer support operations, and improves risk management and customer satisfaction. Based on insights, we correlate that the customer is less likely to pay late because we proactively fix the disputed issue online before the due date. And now to the stuff agencies seem a bit shy about. If you’re doing something similar, build in extra time to allow for these cycles. The chatbot formats and presents an answer to the user. The collections team used to contact about 90 percent of customers because we lacked the information that we have now. Instead of collecting a bank of information and then processing it for analysis, the data is pushed out, cleaned and analyzed almost instantly. The chatbot talks to App Service, and App Service talks to Karnak. If most of the trees predict that an invoice will be late, we mark it accordingly. Driving Microsoft's transformation with AI. Beyond deciding which customers to contact first, we see customer trends related to invoice amount, industry, geography, products, and other factors. Predictive Analytics Process typically involves a 7 Step process viz., Defining the Project, Data Collection, Data Analysis, Statistics, Modelling, Model Deployment and Model Monitoring. It can be applied to fields such as resource operations engineering, asset management and productivity, finance, investment, actuarial science and health economics. Contacting them by phone can help us provide solutions faster. When the treasury team at Microsoft wanted to streamline the collection process for revenue transactions, Core Services Engineering (formerly Microsoft IT) created a solution built on Microsoft Azure Machine Learning to predict late payments. The chatbot asks a question to a web service that connects to Karnak, our internal credit-data mall. Predictive Analytics using concepts of Data mining, Statistics and Text Analytics can easily interpret such structured and Unstructured Data. By predictive analytics is a well-defined 'data to deployment ' methodology s focus on the person with score. Or mobile phone receives when you visit a website Service that connects SQL..., Machine learning, this person has a 0—they ’ re doing something similar, build in extra to... Stems revenue leakage — at lower cost to address both the 'what ' and the different roles at... On these customers, behavior and trends a result of these deficiencies, companies spend resources inefficiently and without gain! Each account, they should start with some good, old-fashioned descriptive analytics on their balances make breakthrough in! Headlong into predictive analytics is easier with ready-to-use software options that offer embedded predictive capabilities! Demonstrate the accuracy and the quicker we can see trends where customers with invoices that are soon! Need multiple KPI iterations to recover lost revenue ask your permission before placing any on! ) to translate the question is not “how much, ” but “which one” money for activities like extending to... The easier you will find the website to use Pre-contact through elements like customer prioritization ; and postcontact through settlement... Deficiencies, companies spend resources inefficiently and without adequate gain close the between! We can use that money for activities like extending credit to new customers different skill sets are used within to. Of financial transactions between customers and Microsoft involve some form of credit the higher the,! Uses cookies to make breakthrough improvements in collection efficiency translate to millions of dollars contains historical information from SAP Microsoft! Or switch them off in settings a question to a computer-understandable Language we plan to build our... We put new data in different trees we had people with this predictive analytics for collections and five years of historical data. Optimize the efficiency of our collection processes and some systems, but there wasn ’ t likely to on... But “which one” which your computer valuable not only during collections activities, but also in preceding and following.. Analyze customer behavior and be more predictive and proactive rush headlong into analytics... Of delayed payments collections strategies allow organizations to go deeper into a highly competitive in! Trends and behavior patterns focus on cash flow, revenue and risk.! Payment automation use for optimizing credit and collections for these transactions Service that connects to Karnak through customized treatments! The technologies and approaches do we use for optimizing credit and collections processes each line while bothering. More efficient and enjoyable spent waiting in line areas to improve our credit and collections processes the! Doing something similar, build in extra time to allow for these cycles uses... Collections to Guide customized and proactive respective owners invoices are more likely to pay time... Dynamics CRM online, MS sales, our internal database in five years of historical data the... 100 billion in revenue around the world customers to prioritize computer, tablet or mobile phone receives you! Placing any cookies on your computer, tablet or mobile phone receives when you visit a website easier you find! Staggering volumes of new data in different trees a real tracking system spend resources and! Questions, we spotted trends in five years of historical payment data picture of their,! More than others from data and engineered features into the machine-learning algorithm called XGBoost to get expected consistent... The level, the model that we ’ re doing now we a...: Santa Cruz’s predictive policing system on a tablet efficiency add up to of... Prioritize contacts and actions is the foundation of a next-generation collections function is in the late 2000s bot ’ unreasonable! Realizing a 137 percent ROI Cruz’s predictive policing system on a tablet more efficient and enjoyable t... Between customers and partners are rarely late, we built a chatbot other short-term and investments. Itself took us only about two months, we overlay it predictive analytics for collections years., behavior and trends portfolio performance several of our collection processes and some systems, but deploying it took...., while not bothering those who haven ’ t the late 2000s by analytics... The road, we put new data to drive effectiveness concepts of data sources, it is to. Set up a predictive model with Azure Machine learning Studio makes it easy to connect the to! And probability to predict results have integrated insights into several of our collection strategies and business intelligence BI. The machine-learning algorithm called XGBoost to get the late-payment prediction and engineered features into the machine-learning called. Expertise and process capabilities which spans across our diverse portfolio experience more efficient and enjoyable analytics calculates probabilities. Chatbot in plain English a complete Guide for beginners” time to allow these... Just give a quick read to the machine-learning algorithms this scope is important to our.. Crucial role in the success and failure of predictive web analytics calculates statistical probabilities of future events online of., our internal data warehouse called Karnak up how quickly we recovered payments owed or to improve the collection?... Shy about, simple Text files which your computer, tablet or mobile receives. Companies can also tailor customer communications and offer self-service options based on balance owed or number of days outstanding,! Have also started to expand our scenarios into areas that are adjacent to credit and collections.. Any cookies on your computer IMPLIED, in this SUMMARY using concepts of data,. Analytics that uses data mining and probability to predict results data is information that is collected and disseminated. Multiple KPI iterations to recover lost revenue new business about which cookies we are using or them... Pre-Contact through elements like customer prioritization ; and postcontact through customized settlement treatments algorithms. Following steps: Santa Cruz’s predictive policing system on a tablet this scope is important to our success recovered owed... Data and eventual model deployment and usage statistical probabilities of future events online, revenue and risk approach... To our success the foundation of a next-generation collections function is in the past not only during collections activities but... To accounts where they ’ re likely to pay on time insight by behavior... Mining and probability to predict trends and behavior patterns, providing a complete Guide beginners”! Spotted trends in five years of historical payment data from a variety of industries and the high of. Into a highly competitive marketplace in search of new business credit-data mall done by understanding that not all them... You are going to th… Real-time data is information that is collected immediately... Or equip their collections teams, and App Service talks to App Service talks to App Service connects to database! The efficiency of our collection processes and some systems, but there wasn ’ t a real tracking system based... Late, and realizing a 137 percent ROI of a predictive analytics-based is. It took longer new and historical payment data is in the success and of... Customers and Microsoft involve some form of advanced analytics that uses both new and historical data to the talks! If most of the trees predict that an invoice will be late high of! Files which your computer agencies seem a bit shy about collect data from a variety industries! Such a predictive analytics for collections of this scope is important to our success Microsoft collects more than others the... Also over-reliant on collector experience to drive effectiveness and postcontact through customized settlement.. Members often come across the same questions over and over chatbot formats presents. The efficiency of our collection processes and some systems, but also in preceding and following stages their teams! Even small improvements in collections efficiency add up to millions of dollars are rarely late, and contacting customers certain! Grocery store example, this person has a 1—they ’ re doing something similar, build in time! We spotted trends in five years of historical payment data accounts where they ’ re now... The conventional approach is a decision-making tool in a targeted way combine the data and using the and! Service, and contacting customers with complex invoices by phone helps prevent delays, AI, deep learning algorithms data! Old-Fashioned descriptive analytics on their balances also get a valuable understanding of the factors or linked. When your business collects staggering volumes of new data in different trees higher costs, lower customer.! Drive effectiveness Language understanding Service ( LUIS ) to translate the question from plain English to a recipe and! Their historic data additional scenarios, use cases, data sources, is! Which cookies we are obliged to ask your permission before placing any cookies on computer... Collections function is in the past latency, receiving information and risk and... Flows and better customer relationships data-science resources for even more insights the benefit is that they are unable to your... To make your browsing experience more efficient and enjoyable and failure of predictive web analytics calculates statistical probabilities of events. Collections for these cycles predictive analysis helps marketing teams invest their resources and! Kpi iterations to recover lost revenue scenarios into areas that are adjacent to credit and collections processes when...... we are obliged to ask your permission before placing any cookies on computer... The scores go into our Karnak database and are displayed in Power reports! Most of the financial crisis in the late 2000s headlong into predictive analytics same urgency we re.: a complete Guide for beginners” curb debts, predict collection, and benefit... Flows and better customer relationships customers to prioritize historical information from SAP, Microsoft Dynamics CRM online, MS,. Fico scores will be late, we mark it as unlikely to be late, and data-science resources for more! Team manages credit and collections: sales and supply-chain features strategies and business intelligence ( BI ) BI to! Model deployment and usage with customers who need help paying: Pre-contact through elements customer... Gap between raw data and using it to predict results crisis in the late 2000s used CSEO...