FICO

Next year Canada will launch the Real-Time Rails – this is a significant step in Payments Canada’s modernization program. Real-time payments offer significant benefits to consumers and businesses, but they are not risk-free. Experience from other countries that have real-time payments demonstrates that criminals will take advantage in a variety of ways. Countries that already have real-time payment schemes have had to manage the related crime. Currently there is little to suggest that Payments Canada has taken steps to stop this happening with the Real-Time Rails. The ‘Modernization Target State’ lays out Payments Canada’s vision for the future and it does talk about fraud prevention. The references are vague, however: “centralized fraud and risk management capabilities” are referenced but the nature of this protection is not defined. It is difficult to understand what format they’ll take, which types of fraud they’ll address or what access payment service providers will have to centralized fraud protection. Additionally, there is nothing publicized about the liability model that will be taken, particularly in the case of authorized push payment fraud. In the UK, the thorny issue of who should bear the losses when people are tricked into making payments to criminals lead to consumer group Which! launching a super complaint. The regulators are now looking to shift some of the liability from the customers to the banks, and a similar move has been made in Sweden. In Canada if banks are pushed to take liability then losses could be significant; if they don’t take liability they’ll still have issues related to bad publicity and reputational damage. Fortunately, other countries have had real-time payment schemes for a while – the most widely adopted is the UK’s Faster Payments scheme, which is a decade old. This means that Canada can learn from early adopters about the fraud that happens and the protection that can be deployed. Read our white paper ‘Canadian Payments - Adapting Fraud Strategy to the Real-Time Rails’ to understand more about fraud in a world of real-time payments and the solutions that can help. The post Will Real-Time Rails in Canada Bring More Fraud? appeared first on FICO.

Consumer awareness of their credit and FICO® Scores has never been higher. Whether planning to buy a car, home or head to college, millions of consumers know and manage their credit score. Now financial inclusion innovation is on the horizon that will change the dynamic of the lender and customer relationship.  Empowering consumers to have greater control over the information that is being used in credit risk management and enabling a deeper dialogue to help both parties make better financial decisions. Earlier this week from Money 20/20, FICO, Experian, and Finicity announced a new credit score called UltraFICO™ Score.  With UltraFICO™ Score, a consumer securely grants permission or ‘opts-in’ to contribute information such as checking, savings and money market account data, which can be compiled by Finicity and combined with consumer credit information from Experian to provide an enhanced view of positive financial behavior. The UltraFICO™ Score includes factors such as length of bank account history and consistency in maintaining positive balances.  In addition, to traditional credit report information. Consumers who are relatively new to credit with limited history or those with previous financial distress that are getting back on their feet stand to benefit the most from this new score. The UltraFICO™ Score will launch as a pilot program in early 2019. The pilot is designed to validate and assess the score with lenders as well as consumers across a broad spectrum including banks and FinTechs. For more information regarding this groundbreaking partnership and new score, read the recent Wall Street Journal cover story: FICO Plans Big Shift in Credit-Score Calculations.  Watch Good Morning America give the highlights of the new score for consumers.  Hear firsthand the excitement surrounding the new UltraFICO Score on YouTube. [embed]https://www.youtube.com/watch?time_continue=96&v=TEZhRwjYcHA[/embed] Lastly lenders and consumers can learn more by visiting: www.fico.com/en/products/ultrafico-score or www.fico.com/ultrafico. The post New UltraFICO™ Score Innovation Now Available for Lenders to Pilot appeared first on FICO.

It’s August and it’s hot in Washington DC.  Yet in the nation’s capital, fintech policy may actually be hotter than the temperatures outside.  Much of the latest buzz has come from the recently released Treasury report which focuses largely on the Trump administration’s outlook for financial innovation. This document is the fourth in a series of reports in response to Executive Order 13772 that details the President’s core principles for financial regulation. The 222 page report contains more than 80 recommendations. The issues that grabbed the headlines included recommendations urging the OCC to move forward with a special purpose bank charter for fintechs (it must have worked  since the OCC made a major announcement just a few hours after the report was released), encouraging U.S. regulators to develop a system similar to a regulatory sandbox, establishing a national data security and breach notification standard and calling for the rescission of the Bureau of Consumer Financial Protection (BCFP) payday/small dollar lending rule. Below, I highlight a few items in the report that may have gone under the radar. These emerging areas offer significant promise for both consumers and market participants alike and can be accelerated by a balanced and thoughtful policy approach. Fintech Policy in the US: New Credit Models and the Promise of Alternative Data The report devotes a section to the promise of new scoring models designed to leverage alternative data sources in credit decisions.  Alternative data in this context is frequently defined as information that currently resides outside of the credit bureau files.  Treasury discusses the opportunities that alternative data presents in expanding access to credit, especially for thin file and credit invisible consumers. The report also highlights FICO’s recent innovation in bringing an alternative data score to market that utilizes telecom and utility payments as well as select public records data not found widely in credit bureau files today. The score provides new credit insights for those who currently cannot be scored without this additional information. The results have been impressive and for the latest findings see a recent announcement regarding our work in this area. Quick Take:  The report was on the mark in encouraging regulators to help facilitate the development of new solutions utilizing alternative data. Treasury is also correct to emphasize the need that any new innovations ensure data accuracy and comply with existing consumer protection laws and regulations including the Fair Credit Reporting Act, Equal Credit Opportunity Act, the Federal Trade Commission Act as well as compliance with model validation and management guidelines. Fintech Policy in the US: Data Aggregation and Consumer Financial Account Data In the fintech world, consumer access to financial account data has attracted great interest. The report provides a short primer on how digitization has sparked new opportunities for fintech firms involved in data aggregation. The report also discusses the challenges with screen-scraping and the importance of improving API methods to address issues such as standardizing data elements and liability concerns linked to data breaches. Yet despite some remaining challenges, Treasury remains very pro-consumer access. It recognizes the BCFP’s prior efforts to promote consumer access to financial account information and encourages the Bureau to provide further clarity to ensure that “third parties properly authorized by consumers, including data aggregators and consumer fintech application providers” are able to obtain financial account information. FICO couldn’t agree more. In fact, we are already working with our partners on the development of a new score that would facilitate the inclusion of consumer-permissioned checking, savings and money market account information to provide additional insights on credit applicants that might not otherwise be approved. Many more d

Expenses Abuse Gets Media and Tax Payers In A Froth Recent news articles out of Ohio ($4 beer dings Butler County Visitors Bureau in state audit) highlights the reputational risks that can impact an organization when it fails to monitor spending for non-compliance or expenses abuse.  The audit uncovered a single $4 Beer which was part of a payment for $11,554. While the applicable statues clearly prohibit this purchase from being reimbursed, it is important to put this in perspective. This was a single $4 purchase and it represents 0.035% of the total amount. There are many ways this reimbursement request could have been identified. If the expense request had been stopped because a supervisor had noticed the purchase during the approval process, or an automated system had identified the non-compliant item then the requestor would simply have removed the item, and the remaining amount would have been reimbursed.  In this case, no audit finding would have been recorded, no blemish on the organization's reputation would have occurred, and no news articles would have been written. Expenses Abuse - Analytics Needed to Deal With Complexity in Large Organizations However, for most jurisdictions, there are not enough resources available review all purchases, especially low dollar purchases. That is why automated analytics are needed. Unfortunately, most ERP, procurement and expense systems don’t contain all of the capabilities needed to identify improper payments. Whereas the technology to provide analytics to truly evaluate and interpret expenses is available to enhance an organization, lower cost and protect their reputation. Analytics can also be tied to optical character recognition (OCR) that scans receipts, invoices and contracts. These analytics can score transactions for risk, fraud, or abuse, but they can also identify transactions that carry risk, or transactions, where the price paid, was too high. These analytics can supplement standard approvals, even scoring transactions prior to the approval to help streamline that process.  (I have discussed this in more detail in a previous post "Expenses Fraud or Honest Mistake? Predictive Analytics Will Tell" if you are interested in reading more.) Without these analytics, jurisdictions leave themselves vulnerable to audit findings, damage to their reputation and surprises (and none of us like these types of surprises!). As we think back to the original story and the initial question, it would be interesting to define the full “cost” of this $4 error.  In this case, it is quite significant, especially for a large organization when you look at the long-term impact. Expenses Abuse - The Hidden Cost Management overhead dealing with the press. The State’s audit is a public document, available on the Internet, and it was picked up by the press.  Of course, the article doesn’t highlight the relative magnitude of the issue or the fact that virtually all expenses were proper.  The organization’s management likely had to put together press releases, field questions from lawmakers, etc. over this one purchase. Management overhead dealing with the state’s audit. The State’s audit showed minimal findings and only this single exception.  However, one can imagine the time spent supporting the audit and responding to the findings, including working with their employee to get reimbursed for the erroneous $4 change before the audit was published. New agency policies and training. In order to be fiscally responsible, the jurisdiction likely wrote new policies to assure this type of erroneous expense did not occur again.  Policies were likely already in place that forbid this transaction, but new policies might have been created for additional manual reviews for these reimbursements.  These actions could have included new training and new formal guidelines.  The guidelines alone might have included numerous meetings and reviews by internal auditors, senior management and/or attorneys. Ongoing new man

If you are a fan of the Bachelor reality TV series, you know that the most anticipated episode of the season is the one after the bachelor has awarded his last rose to the woman of his dreams (or so he hopes).  The After the Final Rose show is intended to provide the inside scoop – the essential takeaways – from what has been a dozen episodes filled with zigs and zags as well as many unanswered questions. That’s the kind of wrap-up that’s long overdue for the industry’s challenge to the  Federal Communications Commission’s (FCC) 2015 Declaratory Ruling and Order  addressing the Telephone Consumer Protection Act (TCPA), which often seemed like a reality show in desperate need of an ending. The finale arrived in March as the DC Circuit Court of Appeals finally issued its long-awaited decision. Now that action has been taken, it’s time for a Bachelor-style breakdown. As I discussed in my last post on this topic, the FCC’s 2015 Declaratory Ruling was viewed by many as creating additional uncertainty in an area that has witnessed the chilling of the use of modern communications technology.  This has been especially apparent when it comes to non-marketing messages delivered by automated voice or text to customers’ mobile phones.  While the DC Circuit decision addressed four main issues, most commentary has focused on three areas: (1) Automatic Telephone Dialing Systems (ATDS); (2) Reassigned Numbers; and (3) Revocation of Consent. I will limit my discussion to these topics. ATDS: Reform Efforts Are Underway The Down and Dirty:  The TCPA defines ATDS as “equipment which has the capacity to store or produce telephone numbers to be called using a random or sequential number generator and to dial such numbers.” Industry wanted the Court to interpret “capacity” as present capacity and reject the 2015 Order that declared that if the equipment has the potential to be an ATDS, it is one. The DC Circuit declined to reinterpret the term “capacity”, but instead found that 2015 Order was arbitrary and capricious because it was confusing and contradictory in its explanation of the basic features of an ATDS.  Among the examples that the court highlighted was that under the 2015 Order a smart phone could be deemed an ATDS - clearly, a bridge too far. As a result, the court sent the issue back to the FCC for further review. Expected Action:  With a new President elected in 2016, the FCC has experienced a change in leadership as Chairman Ajit Pai now heads the agency under a Republican majority. Hopes for some meaningful reform is running high and industry wasted no time in filing a petition on May 3, 2018, asking the FCC for clarification around the issue of ATDS.  Specifically, 18 industry organizations are calling for the FCC to issue a declaratory ruling that would “(1) make clear that to be an ATDS, equipment must use a random or sequential number generator to store or produce numbers and dial those numbers without human intervention, and (2) find that only calls made using actual ATDS capabilities are subject to the TCPA's restrictions.”  The petition will be subject to a comment and reply period that will likely span 45 days in total.  After that, some legal experts have indicated that the FCC could issue a declaratory ruling on this topic as soon as the beginning of 2019. Reassigned Numbers:  Relief Will Require Patience The Down and Dirty:  The TCPA prohibits a caller from making non-marketing calls (including texts) using an ATDS unless the caller has the prior express consent of the “called party”. The FCC’s 2015 Order defined the “called party” as the current subscriber to the phone number called. However, since there are about 100,000 mobile phone numbers reassigned each day, there is no foolproof method of determining whether the consent received from the original subscriber was still valid. The industry requested the term current subscriber be interpreted as the intended recipient of the call. The FCC Order acknowledged the reassigne

The AI Revolution - Are You Ready? With most of the world's innovation created in just the last 150 years, human progress has occurred in what is essentially a historical blink of an eye. The steam engine created a force multiplier of human strength, transforming agriculture and manufacturing and kicking off the industrial revolution. Computers and the internet forever changed our capacity to build and access knowledge, producing everything from smartphones and wikipedia to FaceTime and Facebook. Now we are looking at the next historical change, where the ability for software to think and learn, like a human is upon us. Sebastian Thrun, the lauded computer scientist, innovator and entrepreneur best known for his work on Google's self driving car team, was our guest keynote speaker on day two of FICO World 2018.  More rockstar than computer geek, clad in a leather jacket, he outlined his vision of artificial intelligence and gave us four examples to illustrate the immense changes that are coming. "75 percent of the American working population works in offices. We are now defined through our mental strength. Its not our physical agility anymore, it is our brain," said Thrun. "Most of us do repetitive work, even if you are a doctor, lawyer or financial analyst. Now let me talk about AI,  because AI is about to change all of this.....AI is a way to program computers. Prior to AI we had very smart programmers and very dumb computers. And to instruct a computer to do the right thing, software engineers, would sit down and compose this incredible complicated kitchen recipe of what to do and when......so now AI is trained on examples of what we do and how we think, and can then infer its own rules. It is a huge game changer for all of computer science." AI Revolution: What Is Unique to Human Beings? Chess used to be the pinnacle of human intelligence, until Deep Blue played against grandmaster Gary Kasparov and won in 1997. This was the moment when an AI system equipped with relatively shallow intelligence but massive computing power was able to out calculate Garry Kasparov. Deep Blue was able to play 1,000,000 moves ahead while Kasparov was probably working at about 1,000,000 moves ahead. Then in 2016 a deep learning program beat the reigning Go champion in the world. This AI program played itself many thousands of time. It found all the rules by playing itself. It played so many times, so many more games than any of us could ever play it eventually got better than any human being. There is something systematic going on here. We are taking people who are extremely highly trained and then the AI system is able to beat them at their own profession. AI Revolution: Is Selling to People a Predictive Process? Standford University looked at repetitive work in the domain of sales, specifically online, inbound sales. The study looked at how consumers would land on the websites of multiple fortune 500 companies, inquire about the product and the sales agent's job is to convince them to purchase. It turns out if you look at sales agents, about 1 in 4 are really good and the others are fair to bad. The team monitored the sales agents for one month to discover what lead to a sale or non-sale. By examining the successful transactions, the team could program an AI to deliver a 'next best action' recommendation. Over time with many data points the system continued to improve, transforming itself into a 'salesperson's best adviser' and turn every salesperson into a superstar performer. Over about a six months period, we were able to triple the sales conversion of the team, which is a very significant outcome for this type of technology. AI Revolution: Will Dermatologists Be Trusted with Initial Skin Cancer Diagnosis in the Future? We asked the question, can we learn from these highly skilled people? Dermatology is a highly repetitive job when examining for skin cancers. Our research team compiled 129,000 images of about 2,000 different skin conditio

FICO WORLD 2018, MIAMI, Florida FICO today unveiled a host of new retail banking solutions that allow lenders to manage the credit lifecycle in a way that is smarter, faster and simpler. The innovations are designed to help support lenders with their digital transformations as retail banking continues to be shaped by the forces of mobile, cloud, analytics and big data. “We have seen remarkable changes in banking in the last two years as lenders accelerate their digital transformation and reinvent the customer experience,” said Tim VanTassel, vice president of FICO's credit lifecycle business line. “Lenders are engaged in an analytic arms race to find, delight and retain the right customers by determining and deploying differentiated offers and experiences.” Read the full media release here Originate Loans Using Optimized Deal Structures Providing consumers with multiple options for the best-fitting terms in real time has been an industry goal for some time, and with its latest innovation FICO has made it a reality. FICO® Origination Manager now interacts seamlessly with Alternative Deal Structure (ADS) to evaluate thousands of alternative offers and present a set of differentiated options in seconds. Manage Customers with Greater Agility FICO® Strategy Director, which delivers faster, easier decision authoring and management with plug-and-play industry configurations, builds on the proven decision process flow of FICO® TRIAD® Customer Manager and now offers the same business value with configuration for the Telecommunications industry. The Telco Service Management solution uses analytically driven decisions to help make relevant and profitable offers, reduce churn and improve device financing decisions. At the same time, new configurations for lenders bring the capabilities of TRIAD Customer Manager to Strategy Director and provide a simple migration to speed cloud adoption and time to value. Make Better Decisions Collecting Payments FICO® Debt Manager FitAgent can now sift through massive real-time datasets from internal or external sources, such as dialers and other communication systems, using FICO® Decision Management Platform Streaming to rapidly generate powerful insights that assist with the collection and recovery process. FICO has further improved the FitAgent user experience with new controls, views and insights around payments, customers, accounts and communication that put the customer at the center of a lender’s operations. To reduce the cost of deploying industry-leading analytics, Debt Manager also easily consumes optimization analytics from FICO’s suite of decision optimization capabilities. Improve Customer Communications with AI and the Cloud FICO® Customer Communication Services (CCS), provides analytically driven, scalable, two-way, automated voice, text, email and mobile app notifications. This solution helps banks, telcos, government agencies and utilities around the world connect more effectively with their customers and resolve suspected fraud cases and overdue payments automatically. FICO has collaborated with AWS to bring Amazon’s AI-driven text–to-speech service to CCS. [video width="1280" height="720" mp4="http://www.fico.com/en/blogs/wp-content/uploads/2018/04/AWS-FICOCCS-Final-041418-proxy.mp4"][/video] FICO is also using Amazon Web Services to deliver CCS Voice. “We make millions of contacts per day around the world and AWS allows us to scale with the high performance, security and reliability needed to match the growth of our business. In growing regions like Asia-Pacific, where we currently serve six countries, it has been the perfect match for customers who need speed, simplicity and scale,” said VanTassel. FICO® World 2018 runs through April 19 in Miami Beach, Florida. More than 1,000 business leaders from 50+ countries are attending to explore new ideas, technologies and solutions in AI, advanced analytics and decision management. The post Banking Solutions: Reduce Complexity, Increase Soph

Social Network Analysis (SNA) tools hold the potential to help tax and revenue agencies identify non-compliance and tax fraud.  These tools have been proven to be effective in fraud detection¹, but to date haven’t been utilized to any large degree at tax agencies.  This blog describes what an SNA tool is, and how it can be utilized at a tax agency. Social Networking Analysis Explained SNA is an analytic approach of correlating people, entities and relationships to determine how tightly an individual or business is related to others who have known compliance issues.  These relationships can be from shared phone numbers, physical addresses, bank accounts, credit cards, or any other connection that is available through data capture.  The results of the SNA can show the risk that a specific individual or business presents, based on their relationship with others who have known issues within the network.  As can be seen below, using graph theory and analytics, we gain can valuable insights into specific members of the network. Opportunities for Tax Agencies There are many opportunities for tax and revenue agencies to take advantage of SNA tools, across registration, audit and collections business areas. Improper registrations.  Tax agencies have always faced the problem where they close a business and the business re-opens (typically now owned by a relative of the original owner) at the same or a nearby location.  In this case, a business may have been closed by the state or municipal government for non-payment of taxes.  The owner tries to stay in business by opening a similar (or identical) business with a different legal name and a different legal owner (e.g., a spouse, parent or another relative).  Because of the manual efforts that would be needed to detect the issue, the relationships will almost always go undetected.  The challenge for the tax agency is the owner will have changed, and won’t be an exact match to the previous business. A Social Network Analysis can automatically compare the new registration in real-time to previously closed businesses.  The SNA will create a risk score to show the proximity of the new business to ones previously closed by the department.  In the example below, Suzie Smith has a direct relationship with one closed business and a second degree relationship with another closed business.  This could trigger the state to deny the new registration or conduct additional investigations in a cost effective manner. Identify Fraud Rings.  Once an instance of fraud is identified, the SNA can identify related businesses.  This could include businesses with related addresses, bank accounts, phone numbers, email addresses, or other identifying characteristics.  Through the use of an SNA investigative tool, a revenue agent can identify additional individuals and/or businesses that may be related to the same fraud ring, saving the investigator time and effort. Assisting with Locating a Delinquent Debtor.  SNA’s are frequently used during the debt collection process to identify related individuals.  Collectors have utilized manual tools along these lines for years, contacting next-door neighbors (for example) in an attempt to locate a debtor.  This tool can greatly enhance and automate this effort, by finding people who share the same physical address, phone number, email, etc. Finding Successor Businesses.  When a business ceases operation, the business can re-open in a new location or under new ownership.  When the original business owes money, the government can in many cases pursue that debt if there is a successor business.  Through the SNA tool, successor businesses can be quickly identified.  Businesses and individuals can be identified that have relationships with the prior businesses, such as those sharing bank accounts, physical address(es), phone numbers or other identifying criteria.  This can save the collector a significant amount of time for what otherwise would require significant manual resea

Categorically, artificial intelligence (AI) can appear be an odd juxtaposition of order and disorder — we direct the AI with algorithms, yet the system produces new insights seemingly magically. This two-part blog unpacks the mysteries of two very different AI techniques: supervised and unsupervised learning. Supervised Learning: The Workhorse of AI Most of the well-known applications of machine learning and computational AI involve supervised learning. The modeler amasses a vast set of existing data (e.g., financial transactions, internet photographs, or the texts of tweets) and a base-level “ground truth” outcome that is already known, perhaps in retrospect or by expensive human investigation. Equipped with any number of computational algorithms, the scientist becomes the “supervisor” whose code trains the model to reproduce, in the lab, the known outcomes with a low probability of error. The models are then deployed to live a happy life scoring credit risk and fraud likelihood, finding pictures of Chihuahuas and muffins, or flagging insulting tweets. Technically, each model computes a probabilistically weighted predicted outcome that we believe to be like those outcomes from the training examples. The state of the art for supervised learning is now well established; you can choose from dozens of comprehensive predictive analytics and neural network packages. Unsupervised Learning: Inferences in the Absence of Outcomes But what if there is no set of “true outcomes” known, or the ones at hand are restricted in quality or quantity? What can machine learning do for us then? This is the domain of the far trickier unsupervised learning, which draws inferences in the absence of outcomes. Good unsupervised learning requires more care, judgement and experience than supervised, because there is no clear, mathematically representable goal for the computer to blindly optimize to without understanding the underlying domain. The Challenge of Outlier Detection A central task within unsupervised modeling is outlier detection: Which examples are most unlike most of their peers? Outlier detection and transaction fraud scoring provide an easy illustration: Which customers request money transfers with patterns substantially different from most of their peers? Which medical providers bill insurance for sets of claims most unlike their peers? Which transactions on an individual payment card are most different from a customer’s usual behaviors? The solution pattern for these tasks is a problem- and domain-specific transformation of the raw data into a quantitative vector space of features — up to now, exactly in line with supervised predictive modeling. This is followed by a more generic mathematical construction to yield a numerical score of the “degree of outlier-ness,” in the absence of ground-truth training outcomes. Because there are far fewer principles, and less didactic instruction and widely available software compared to classic supervised modeling, there are even more analytic “gotchas” requiring deep analytic scientist experience and judgement. Difficulties and considerations in outlier detection include: The need to define a metric or distance. Many techniques require defining a “metric” or “distance” function between pairs of observations. One problem is that the individual components of this feature vector have qualitatively different meanings – how can one balance adding or subtracting apples and oranges, and kumquats and kangaroos?Often this is done ad-hoc or, unfortunately, without intention as the underlying algorithm method assumes a metric. What should be done in the real-life scenario of a combination of quantitative and categorical features? Supervised modeling can often be blissfully ignorant of this problem, since the quantitative optimization with known targets tends to scale and transform each feature automatically, to the degree that it contributes predictive value. In an unsupervised context, an explicit metric will

Whilst UK consumers are likely to be enthused by the reforms around switching mobile provider and switch requests announced by Ofcom recently, this will give mobile operators an additional regulatory headache when it comes to churn prevention. The reforms are centred on making the switching process easier and enabling consumers to invoke the process by sending a free text message to their current provider. Ofcom research has suggested that one of the biggest hurdles that consumers face is “having to speak to your current provider, and facing unwanted attempts to persuade you to stay.” So, whilst it’s clearly in the operators interests to have a conversation with the customer to incentivise them to stay, customers “will control how much contact they have with their current mobile provider, preventing companies from delaying and frustrating the switch process.” This raises questions that operators will need to consider. Most providers cite their customer satisfaction measures, like Net Promoter Scores, as absolutely key, and consider exceptional service to be a competitive differentiator. In fact, recent research presented by Analysis Mason at the European Telecoms Summit 2017 suggests that for every 1% improvement in customer satisfaction there is a correlating 4% in churn reduction. Yet with around 5 million customers per year choosing to switch provider, it is clear that operators still have work to do in order to maintain loyal relationships with their customers. Removing the “traditional” retention conversation could re-focus providers into ensuring they are managing their customers appropriately before the customer begins to consider their options. Churn Analytics This is where advanced analytics comes into its own. Analytics can proactively segment customers according to their likelihood of leaving, with churn scores based on their profile and network activity data, such as the age of their mobile device, number of negative customer support interactions and dropped call or service incidents. These customers can also be segmented by value and other items, including cost to serve, in order to inform the decision as to what tailored, automated actions will mitigate each individual’s risk of churn. This includes incentives that exceed expectations and delight customers, or indeed whether the investment to retain a customer could be better used elsewhere and allow some low-value customers to churn. Having the ability to interpret the churn signals coming from different customer segments, added to the tools to execute decisions that mitigate the risk and cost-effectively retain customers, is becoming ever more important in managing churn prevention. This is supported through research by Three quoted in the Ofcom report:  “61% of switchers who were offered a reactive save deal said they had made their final decision before starting the switching process, and their previous provider could not reasonably have done anything to change their mind about switching.” There must be scope for operators to understand these switchers at a more sophisticated level and take action that dissuades them from making switch requests. How Fast Can You Respond to a Switch Request? Some savvy customers will of course use the process to their advantage as they do today; many customers will say they want to leave in order to get a better deal and the churn prevention activity is relatively straightforward. However, one of the policy aims of the Ofcom decision is to reduce unwanted attempts to persuade customers to remain. They lay out that expectations of providers are, in short, not to proactively contact the customer about the Port / Switching code in a way that may induce the customer to not proceed with the switch or to call the current provider. However, they do recognise that once a switch request has been made, operators are still free to contact the customer to attempt a save. With a one working day, or less, time frame (in most cases) between