»» insightsCan Fraud Alerts RaiseCustomer Loyalty?Stop fraud faster while building trust by bettermanaging the customer experienceFor most people, being contacted by a bank about potential fraud isn’t a wonderful experience—but it can be if it’s managed in a timely, personalized manner. In fact, customers of a UK bankconsistently point to fraud intervention as a“golden moment”of customer service in theorganization’s quarterly surveys.Mobile devices and electronic channels are enabling banks to engage customers, using the mediaof their choice, in intelligent automated dialogs that stop fraud in its tracks. With no increase in fraudmanagement personnel expense, banks are stopping fraud faster, while building customer trust andloyalty in ways that drive account usage and spending. And by constantly feeding back to analyticsthe results of these customer interactions, they’re improving the accuracy and adaptability of frauddetection systems.This white paper discusses how leading banks are driving gains by managing individual customerexperiences as part of fraud intervention. We’ll share some of the results they’re seeing in terms ofbetter fraud detection, increased efficiency and higher customer satisfaction.The paper also reveals how industry leaders are:• Stopping fraud earlier before losses mount.• Finding more fraud without annoying customers.• Overcoming channel/product silos for customer-level intervention.• Making detection and action smarter by closing the loop between them.• Adapting faster to changing fraud patterns.Number 64—December 2012 www.fico.com Make every decision countTM Learn how one bank resolved 250% morefraud cases with 30% fewer declines, 95%customer satisfaction and zero FTE increase
www.fico.com page 2Can Fraud Alerts Raise Customer Loyalty?»» insights91%happy withauto-resolution 94%would recommendit to family & friendsBank 187%willing to receivecontacts in future78%agree strongly thetechnology has improvedthe overall level of servicethey’re experiencingBank 284%ﬁnd auto-resolutioneasy to use 89%said auto-resolution hasincreased their conﬁdencein using cardBank 3It’s been called a“moment of truth.”When a bank contacts a customer to check on a possiblefraudulent transaction, the experience can be a negative or positive one for the relationship.In fact, customers regularly characterize such moments as“golden”in one UK bank’s quarterlycustomer service surveys. This demonstrates how much opportunity there is to turn what couldbe an annoyance into a valued service.Timeliness and personalization are critical to better managing the customer experience. Whena suspicious transaction is detected, banks must be able to reach customers where they are,through the best means for that individual given the circumstances. That means being ableto make contact not only through calls to landline phones, but through calls, emails and textmessages to mobile phones, tablets, laptops and other devices.It also means engaging customers in an efficient automated dialog. For card transactions, insteadof customers receiving a one-way communication, hours or days later, requesting they call thefraud line, most cases can be resolved within minutes, some in seconds. Instead of a legitimatetransaction being declined, customers immediately confirm they are the one using the card,and can continue spending. And for online banking, instead of banks having to choose betweenletting a risky funds transfer go through and meeting regulatory requirements for processingtimes, they can quickly check with the customer before releasing the funds.When banks leverage such auto-resolution capabilities, customers can resolve suspected fraudcases on their account themselves, often simply by pressing a key on their phone or computer(see Figure 1). While there’s always the option of talking with a fraud analyst, they don’t have tobecause they can interact with a virtual analyst—able to perform the same functions, includingaccessing records and updating the account—in a direct, immediate and discrete manner.Customer feedback has been extremely positive for banks that have implemented this approach.At the UK bank mentioned above, 76% were highly satisfied with its auto-resolution service, 84%found the service easy to use and 89% said they had increased confidence in using their cardagain. Other banks report similar or even higher customer approval rates, as shown in Figure 2.»» Turning “Moments ofTruth” into Gold43% of consumerssurveyed want their bankto engage with themmore on account updates,potential fraud and othernon-marketing matters.—FICO research, 2012Figure 1: Speedy fraud caseresolution by SMSFigure 2: Strong customer approval of auto-resolution—survey results from threeFICO clients in the UK
www.fico.com page 3Can Fraud Alerts Raise Customer Loyalty?»» insights»» insightsThis heightened customer satisfaction and loyalty aren’t earned by punching out messages tocustomers with cookie-cutter contact strategies. They come from assessing multiple factorsabout the situation and the customer to form and execute an individualized strategy forcontacting that customer in the right way with the right timing.As depicted in Figure 3, banks today need to couple intelligent, automated fraud detection withintelligent, automated decision making for customer experience management. And they needto be able to perform these functions, when appropriate, at high volumes in real-time.Closing the loop between detection and action makes both smarter.Action is smarter: When assessing whether action is required, an auto-resolution solution can consider an overall fraud score from a multichannel/multiproduct detection system, or look at multiple scores and data from differentchannel- or product-specific systems. It may perform additional checks, such aslooking to see if the telephone numbers and email addresses on the accounthave ever been associated with suspicious activity in the past or match thoseon a fraud watch list. Banks can also, of course, create any number and variety ofcase generation/management rules to guide auto-resolution.To decide what action to take, the solution considers the customer’s contactpreferences and contact history (e.g., has she responded better to SMS ora phone call?). It looks at the time zone of where it appears the customer iscurrently located. It selects the right channel or combination/sequence ofchannels, avoiding any that may have been compromised. A good system, forinstance, will check to see if there’s been a change in a mobile phone SIM cardsince the last transaction or if an inbound caller ID has been altered.Detection is smarter: The resulting interventions generally resolve fraud casesswiftly, with results fed back to detection in a closed loop. That’s a powerful»» Managing IndividualFraud ExperiencesWhat else do we knowabout this customer andthis situation? Whichcommunication channelsare available and secure?Identify the besttreatment for this customer,in this situation, at thispoint in time.Contact strategy executedacross one or morechannels. Results monitoredand strategy adjusted asneeded.Legitimate or fraudulenttransaction conﬁrmed.Customer assisted. Resultsfed back to analytics toimprove detection.Assess Decide Intervene ResolveCustomertransactionsHow risky is thistransaction? (Single scoreacross channels/productsor multiple scores fromdiﬀerent detectionsystems)DetectionScoreAction! ✖LOW HIGH✓✓Real-time feedbackFigure 3: A closed loop between detection and action makes both smarterwithin30 secsMost customercontacts made15 secs(from 20+mins)Improvement inaverage time-to-case50%Reduction in averagecontact time67%Cases resolvedwithin 20 minsFigure 4: Auto-resolution speed metrics—typicalresults from FICO clients worldwide
www.fico.com page 4Can Fraud Alerts Raise Customer Loyalty?»» insights»» insightsadvantage if the detection system includes an adaptive model able to use this feedback toadjust in real time to changing fraud patterns and customer behavior.This closed-loop approach also enables banks to deliver each customer a planned, managedexperience that’s dynamic and fully responsive to what is happening with the customer andin the production environment. As the possible fraud scenarios depicted in Figure 5 show, thedelivery of this experience is a communication-enabled business process that extends acrosschannels and products. Customers get a consistent experience because the virtual analystworks round-the-clock and covers all touchpoints. If a customer receives a contact through onechannel and chooses to respond through another, the virtual analyst still knows who they areand everything that has transpired on the case.Other recent transactions in thiscountry on other accounts?Changes to SIM card since lasttransaction? What time is it attransaction location andcustomer home location?If other transactions and no SIMchange, unblock immediately.If no transactions, contactcustomer via SMS (customer-preferred method). If SIMchange, send email and leavevoicemail on landline. Adjuststrategy based on time zones.SMS inquiry to mobile phonesent within minutes oftransaction denial.Customer conﬁrms legitimate.Retried transaction approved.Uses card conﬁdently duringrest of trip.Assess Decide Intervene ResolveHigh-risk transaction for largeamount with merchant outside ofcountry, not purchased frombefore. No notation about travelon ﬁle. Transaction declined.DetectionScoreAction! ✖LOW HIGH✓✓Faster Payments regulationsapply, but transaction is highrisk. Are payer/receiver accountowners the same, or are thereknown relationships betweenthem? Have there been any SIMcard changes since lasttransaction?Put transfer on hold while usingall available and secure channelsto contact customer.Customer contactedsimultaneously via smartphoneapp and voice calls to mobileand landline.Customer conﬁrms fraud viasmartphone app, so payment isreleased for settlement.Customer is provided withreassurance and next steps forre-establishing account security.Wire transfer attempt looks likecoming from customer (correct IPaddress, HTTPS certiﬁcate,one-time passcode, etc.), butbehavior pattern is unusual.Malware may have interceptedand receiver account valueschanged.What time is it at customerlocation (odd hour for banking)?Have there been any recent SIMcard changes or a diﬀerentinbound caller ID from previousinteractions?Monitor account whilecontacting customer.Customer contacted via email. Ifcontact not made within 24 hrs,also attempt via voice call tolandline and mobile phone.Block any subsequent high-riskpurchases or fund transferattempts.Customer conﬁrms new user isnot authorized before fraudsterhas opportunity to break outand empty funds.New authorized user has justbeen added to debit account.Could be ﬁrst sign of accounttakeover. Several unusual smallonline purchases raise risk level.Credit/debit card transactionOnline bankingNonmonetary transactionFigure 5: Auto-resolution examples, possible fraud scenarios based on best practicesReal-time feedback
www.fico.com page 5Can Fraud Alerts Raise Customer Loyalty?»» insightsResource constraints and concern about annoying customers with false positives havetraditionally discouraged banks from generating cases for transactions in midrangefraud scorebands.Auto-resolution eliminates both resource and false positiveconstraints. While an expert analyst is often able to resolve fewerthan 20 cases per hour, a virtual analyst can resolve an unlimitednumber with no staffing increase and while keeping customers happy.By reaching deeper into scorebands, banks catch more fraud. Asshown in Figure 6, the incidence of fraud is lower at a score of 450than it is at a score of 850. Nevertheless, fraud is still present at thislevel, even with a lower detection rate, so the benefit to detectingfraud in this range is significant.FICO clients across the globe are reaping the benefits of diggingdeeper: One UK bank that implemented auto-resolution is resolving250% more fraud cases—achieving 30% fewer declines and 95%customer satisfaction—with zero FTE increase. Another UK bank hasreduced the average number of transactions per fraud found from 5to 3. A US bank reduced net fraud losses by 75%.Banks vary in how they’re assigning this expanded range offraud cases to human and virtual fraud analysts. Some write casemanagement rules that continue to direct the highest-risk casesto their most experienced analysts, while directing midrange riskcases to auto-resolution. Some prefer to put auto-resolution on thehigh-risk cases first, since at this level, speed-to-contact and rapidresolution are most crucial for keeping losses down.Some banks are even opting to deploy auto-resolution acrossthe entire score range, freeing their expert analysts to investigateconfirmed fraud and interact with customers on very complex andhigh-value cases. Best-in-class auto-resolution solutions include easy-to-use business rules management and workflow management toolsthat enable fraud department managers to experiment with differentapproaches, and quickly capture and compare results metrics.»» Stopping MoreFraud FasterNormalcall centerfocusAdditionalautomationcapacityFraudriskscoresTypicalfalsepositive:fraudratios850 10 : 155 : 1450Figure 6: Reaching deeper into scorebands tocatch more fraud and minimize customer impact50%Fraud loss per case70%Unrecovered fraud losses35%Overall fraud lossesFigure 7: UK banking group reduces losses withauto-resolution60% (–30% FTE)Cases worked300% (same FTE)Cases workedFigure 8: US and UK banks use different auto-resolutionstrategies—both achieve major efficiency gains
www.fico.com page 6Can Fraud Alerts Raise Customer Loyalty?»» insightsAnother way auto-resolution helps banks stop fraud faster and reduce losses is through thefeedback to detection analytics. This feedback comes in quicker with auto-resolution. If the bankis using auto-resolution to probe down into midrange scorebands, there’s also more feedbackto work with. And if the bank has online and mobile apps that customers can use to quicklyconfirm multiple transactions, as shown in Figure 9, the volume of data coming back toanalytics is even greater.Fraud detection systems can leverage this incoming data in several ways.Fraudulent online authentication attempts, for instance, are highly predictive ofaccount takeovers. But when fraudsters fail authentication, they never get to thetransaction step, so fraud detection systems don’t generally receive any data. Auto-resolution systems, however, can immediately pass data on confirmed fraudulentauthentication attempts to detection systems.Incoming auto-resolution data can also be used by detection analytics to self-adjust inreal time to changing fraud patterns and customer behaviors. This is a unique capabilityof FICO® Falcon® Fraud Manager.As shown in Figure 10, the Falcon system incorporates an adaptive model. This analyticlayer contributes a valuable additional perspective because it’s focused on andhighly responsive to what is happening in the current production environment. Alltransactions the base model scores above a bank-selected“cascade threshold”go tothe adaptive model for an additional score, which is then blended with the first one.»» Leveraging RapidFeedback for HigherPerformanceBase analytics Adaptive analytic layerUpdating ofvariable weights(continuous)Fraud/no-fraud tablesRapid feedback of fraud/no fraud conﬁrmationsProﬁle Scaled proﬁle ModelReviewthresholdConﬁrmed fraud casesfor ﬁnal resolutionCascadethresholdExpert analystVirtual analystModelVariable evaluation replacementbased on relevancyto production data(frequent intervals)VariablecandidatepoolTransactional dataBasescoreBlendedscoreAdaptivescore êézii ii, p, px qclRapid feedback from auto-resolution accelerates theunique adaptive capabilities ofFICO® Falcon® Fraud Manager,enabling banks to detectemerging fraud schemes andadjust to changing customerbehavior fasterFigure 10: Rapid feedback increases adaptability of detection modelsFigure 9: Gathering more fraud informationwith a smartphone app
www.fico.com page 7Can Fraud Alerts Raise Customer Loyalty?»» insightsThe adaptive model uses fraud case resolutions and data to evaluate which data variables areproving most predictive of fraud now and in the recent past (past day, past week, etc.). For example,are the values for the dollar amount variable significantly different between electronics purchasetransactions confirmed as fraudulent vs. similar transactions confirmed as legitimate? The adaptivemodel swaps these currently high-performing variables into its fraud/no fraud tables whileswapping out those currently less predictive. It also uses the case dispositions to continuously adjustvariable weighting to the changing fraud patterns it’s seeing.These adaptive processes improve detection of emerging fraud schemes, so banks can stop themsooner. When combined with other dynamic analytic techniques also built into the Falcon system—such as self-calibrating profiles that tune base model weights to changing customer behavior oreconomic conditions—the result is more accurate fraud detection and reduced false positive ratesat every scoreband.There is evidence, both from what customers say and what they do, that the experience of a timely,dialog with a virtual fraud analyst builds loyalty. It also increases the value customers feel they’rereceiving from the bank, and influences their behavior in ways that drive revenue and profit.The most clear cause-and-effect impact on revenue is, of course, atthe point of sale. When checkout problems arise, immediate, helpfulintervention greatly lessens the embarrassment and inconvenience thatcustomers may experience. It also enables more legitimate high-risktransactions to go through. With the ability to contact customers within 30seconds to confirm legitimacy, one UK bank has reduced declines by 32%.And when a very high fraud score causes a bank to block or restrict acard, faster resolution reduces the impact on subsequent transactions.Another UK bank is able to contact 79% of its customers within 10minutes of taking such actions. If the customer confirms the transactionis legitimate, blocks and restrictions are removed instantly.For online purchases, auto-resolution helps banks avoid losingtransaction fees because of failed authentication attempts by legitimatecardholders. For example, if a customer mis-keys his 3D Secure passcode(e.g., Verified by Visa, MasterCard Secure Code), the virtual analyst canlet him know immediately that this was the reason for the decline andprompt him to try again.These impacts are increasingly important to banks that are changing the way they measure fraudmanagement performance. A growing number of UK and European banks have expanded their KPIs(key performance indicators) to include not only reducing fraud losses, but increasing spend percard also.Quick resolution can also help boost spend by increasing the customer’s confidence about usingthe card again. Instead of switching to another card or curtailing shopping altogether, the consumeris more likely to make subsequent purchases with the same card. In cases of confirmed fraud, itreduces the odds—1 in 5, according to a 2010 Datamonitor global survey1—that the consumer willclose the account altogether.1Consumer-Facing Fraud Prevention Strategies in Payments study, Datamonitor (Ovum), 2010»» Increasing CustomerLoyalty and AccountUsage10%Active debit (spend) population+1%Number of transactions+23%Average ticket value per spend+Figure 11: Higher revenue—auto-resolution results fromone bank’s debit portfolio
www.fico.com page 8Can Fraud Alerts Raise Customer Loyalty?»» insightsMoreover, because auto-resolution is managing a customer-level experience, a single dialog canconfirm status and transactions on multiple cards/accounts. If fraud is confirmed on a debit card, forinstance, the customer doesn’t have to wonder if a credit card attached to her deposit account hasalso been compromised; auto-resolution can ease her fears on the spot. As a result, the increasedconfidence she feels, and its impact on subsequent usage, may also extend beyond a singleaccount, to the full relationship.Though the numbers vary by country, the Datamonitor survey mentioned above found that half ofvictims and nearly as many non-victims agreed they can“do more to protect themselves from fraud.”While detecting and preventing fraud will continue to be the banks’responsibility, two-way mobilecommunications that enable consumers to increase their awareness and play a more active role inthe process can reduce fraud losses and help bring down the costs of using convenient paymentvehicles like credit cards.That feeling—of being part of a valuable relationship, based on trust and characterized by goodcommunication—is precisely what banks are now working so hard to create with their customers.The beauty of adding auto-resolution to fraud management is that it helps banks rise above obstaclesto this goal, such as siloed data stores that have made it difficult to manage fraud interventionsat a customer level. At the same time, it provides an ideal opportunity for initiating a dialog thatconsumers perceive as aimed at their welfare, rather than“just trying to sell them something.”Combining auto-resolution with advanced adaptive analytic detectionFICO offers best-in-class fraud solutions for auto-resolution and real-time detection. When used together, the solutions enable banks toimprove fraud management and increase customer loyalty.FICO® Adeptra® Fraud Resolution Manager reduces the time required for fraud resolution from hours or days to seconds or minutes. Basedon fraud output (i.e., scores, rule results) from a transactional fraud detection system, it performs high-volume, real-time intelligent assessmentof fraud cases, identifying the best contact strategy for each customer in each situation. It executes the strategy across one or more channels,managing a customer-level experience, and feeding back data to analytic detection.Combined with the fraud detection capabilities of FICO® Falcon® Fraud Manager, the feedback loop from auto-resolution delivers powerfulbenefits. The unique adaptive analytics in the Falcon system can use information from the abundant stream of timely auto-resolution datato adapt fraud risk scoring to changing fraud patterns and customer behaviors. The result is faster detection of emerging fraud schemes,enabling banks to stop them sooner.