Case Document 193-8 Filed 01/24/19 Page 1 of 15 EXHIBIT Case Document 193-8 Filed 01/24/19 Page 2 of 15 From: tare Stewart ?ent: Friday, December 02, 2011 12:51 AM ?to: Tara Stewart Subjeet: Re: {tasks} Bootcamo Project, Part 2: Transaction Level Data {risk-ops} [riskorojectl [risk credits projects] Bennett Woo at 2:59;:m yesterday changed the subscribers. Added: Bennett Woo. [To see this task, go to: Tare Stewart at er?7pro Already posted this on the MC EE discussion, but postihg the relevant data here too: Regarding the ?rst?) In?ow Resoohse: Juet ran the date for my friendly fraud inflow project This rule targeted users under the age of 1? and above the age or 98 who were trying to Spend greater than or eouef to $75 in a single in Pet Society, Backyard Monsters! or EA. $3th Superstars: Reel football 84 soccerlt The rule respooee was the firstw? CC inflowt included refunds and chargebacks in the same loss bucket because together they constitute total friendly fraud. This rule wot active from 8/2/2011 to 10/31/2011. Users who passed in?ow and made subsequent in any of the 3 test apps: 13.3% of the count of txoa were Charged buck or refunded 27.8% of the sum total of were lost as a result of chargehacks refunds 14.9% of these users (count) charged back or were refunded Users in the control group for this rule: 19.2% or the count were charged back or refunded 39.9% of the sum total of time were lost as resott of chergebacks refunds 13.2% of these users {count} charged back or were refunded Users who passed inflow and didn?t spend in the 3 test. trope or users who failed inflow: From a statistically significant semoie, 6% of these users charged back or were refunded. 32% of users ir: the inflow group fall into this segment The results from my project indicate that first?6 inflow altered the chargebact and refund rates. While a anterior percentage of the control group users; charged back or were refunded, they charged back a greater number of transoctioos and a larger total sum? Fewer transactions were ultimately charged back or refunded from the users who oegsed Enflow. Though the inflow didn't make a huge dentin the loss rate, it blocked a fair amount of users "from ourshasiog irr the test agape. lt' this were run at a larger scale, it could decrease the denominator of the segment we are targeting. Anecdotally, many of the remote who did not spend in the test apps after inflow went on to spend in other appsv Unfortunately, due to time eonstreiots, i didn?t have time to dig deeper into the users who passed inflow hut didn?t spend or all-out failed the inflow. This could be a next step. Case Document 193-8 Filed 01/24/19 Page 3 of 15 uitimateiy, wouid recommend a ?rst-6 in?ow tea the iower threshoid users. Title: Bootcamp Protect, Part 2: Transaction Levei Date Created: 8:4Sam May 27th, 2811 by Tara Stewart Tags: risk?ops, riskproject, risk credits projects Priority: lowupri Assigned to: Tara Stewart Reiated to this orig'nal d'scussion on underage users: http: intern tacebook pitta/?Boot ?Camp Projectw The "Financ ?impact of Permitting_ Minors to Spend_ Facebook Credits/Howe} 1.233961701073333 Now focusing exciusively on users tagged with one of the three minor tags in to determine if we should be refunding after a certain spend threshold during review Next steps: 1. Sort the users in the original query by taglt) 2. Determine the size of each of the Samples (it of Aims} and record the median CB amt per account, med'an refund amount and med'an successful amount 3 Repeat steps 1+2 w: ith the second query to get an idea of the totai popuiation of minor? spending credi t3 break this query out by age113 <2 instead or summing by Ali) run a third query that puiis all transaoti ons to determine the typicai number of oer account and the actuai impact on compliance To See thie task, go to: Reeiy to post a comment. Fuii History Tare Stewart at 8:48am on May 27th created the task. Tara Stewart at 8:48am on May 27th changed the title to "Bootcemp Project, Part 2: Trangection Level Data" Tara Stewart - at 8:48am on May 27th changed the description to ?Reiated to this original discussion on underage userz htte: mntern facebook. ens/index one/Boot mCamo Proieetm The jinenc erneect of _Perm?tting Minors to _Spend Facehook Credits/?fbid: 12339617010733i?39 New focusing exciusively so users tagged with one of the three minor tags in CRT to determine if we shouid be refunding after a certain spend threshold during review Next steps: 1.Sertthe users in the original query by tagit?) 2. Determine the size of each of the gametes (d otAl?s) and record the median C8 amt oer account, median refund amount, and median successful amount 3. Repeat steps; 1+2 with the second query to get an idea of the totei population of miner?s spending this query out by age: 13-17 h} FBniiS~OGi3i 37$ Case Document 193-8 Filed 01/24/19 Page 4 of 15 lastead of summing by All), run a third query that all transactions individually to determine the typical number per account and the actual impact on compliance" "fare Stewart at 8:48am on May 27th Changed the subscriber; Added: Tara Stewart, {David Clune, Rees Worden. Tara Stewart at 8:48am on May 27th claimed the task. Tara Stewart at 8:483m on May Z??th added the attachment Tara Stewart at 8:4821m on May 27th changed the tags. Added: 'riskproject', 'risltcops?, ?credits'. Tera Stewart ., at: 8:48am on May 27th added the attachment Tara Stewa? at 8:48am on May 27th added the attachment "bootcamppartZwerk.xlsx?. Tara Stewart at 12:51pm en May 3151: FYE..?preject has been slight?y delayed as we work an de?buggmg the third query (step it) which will determine the actual impact on compliance Tara Stewart at 1:50pm on June 6th Ran the third query and it returned 8 bi?lion lineet..needless say, some more debugging is in order? Ross Worden at 1:55pm on June 6th TABLE Ad SELECT DISTINCT msourcewecceun?id, vamtgucc, I mamtref, vamtch, verderid asourcemaccount??d as sourcemaccoungid, a.c_retereneewid as drder?d, xemtsucc as amtsumz, Qamtre?f as amtref, eamtcb as amtcb FROM LEFT OUTER JOIN SELECT z.seurce_account_id, LCHreferenceJd, z.amount_usd as amtsucc FROM Case Document 193-8 Filed 01/24/19 Page 5 of 15 WHERE z.product__name 2 ?credits? and AND AND ((z.refundmamt20 AND z.cb?amount (3)01? (z.refund_amt IS NULL AND z.cb%ameunt13 )x ON asource?accountwid x.source?accounthid LEFT OUTERJOEN SELECT beaurceaacceuntw?d, b.1rmreferez1cem?d, bemountmuad as amtref FROM WHERE AND AND Uproductwnamez ?credits' AND (bxefunc?wamb? OR ?3 NOT NULL) (1 ON assurcewacmuntid esourcemaccountmid LEFT OUTER mm SELECT d.sourcew_accemtmid, demreferencejd, demonnt?esd as amtcb FROM r} WHEQE AND QND deroduaiwnamez?credit? AND (dcbuamoun?zm OR cmbwamaum: 15 8 ON agourcewacceunt?id esourcehacceunigd WHERE agroductwname 2 ?Cred?ts' anc? AND ANQ R055 Worden ., at 1:55pm on June 6th CREATE TABLE AS SELECT 33?, gage, CASE WHEN qtagidm 131%2480193253 OR q.tagid =1 1.448223653617990 OR qiagid THEN 3. ELSE 0 END AS tagged FROM 3 LEFT OUTER JOIN pbwusetaccount ON aeourcewaccountmidufeccountwid LEFT dimWaELusergWinfo 5: Euser?widzguseriz? ANS 95135261305433 AND gage?crl? AN gage {5 Ni)? NULL OUTER (SELECT yebjec?d as objectid, y?tagid a3 tagid FRQM obgemmteutag?asso: WHERE )q ON Ross Warden at 1:56pm on June 6% i?m pretty sure the dupping is due to joining an account ids; and mutt ids in the first query 4 7 Case Document 193-8 Filed 01/24/19 Page 6 of 15 Tara Stewart at 7:53am cm June 8th joined on id in the first query, but this; returned a huge amount of R553 Worden - 3E $z1?am on June am what of rows? Ross Worden - at 8:123m on June 8th witheutjeining on txn_id (insteao?, 30in an aid) the count is 173,640,493. i?m almost certain that the 1:many compenent of and the jein structure is responsible for dupping Ross Worden - at 8:1423m on June 8th aiso, undate the second query to: CREATE TABLE AS SELECT 3* gage, CASE WHEN qtagid? 131%2480193255 OR qiagid :2 QR qitagfd THEN 3; ELSE 0 AS tagged, qtagdate as tagdate FROM a - LEFT OUTER JOIN pbuuserjccount GN aseurcegccauntidmf.accountwid LEFT OUTER JOIN dimwaiimusergwinfe (3N ?userwidmguserid AND g?m?ZQil?QS?Ol? AND AND gage iS NOT NULL LEFT GUTEREOW (SELECT yebjectid as; objectic?g yiagici as tagici, minide} as tagdate iiRGfV?: abjeciwtommgmagsoc WHERE 12? i ON q.ohject?dzf.usegid; that: mimds) is geing t0 hair; too Tara Stewart - at 8:21am on June 8th Mining on transaction returned ?6,338,139; {owsmm de?niteiy a ioi: smaiier than your originai puii. Just net sure how make this manageable si?ce i can?t ever?: export that many to Excel, i'ii think on Ross Worden v~ at 8:32am on June 8th ithink their ii makea geese giver: {he Query parametew can you post the query Oi? the code so i can integrate it? We won?t have to expert is excei - you can run queries on the new tabie is get: the ievei info, orjust randomiy sampie QNWSENTEAL Case Document 193-8 Filed 01/24/19 Page 7 of 15 Tara Stewart at ?217am on June 8th Query number: 113.3858. Tabie name: bootcampjoinontxn Tara Stewart at 5:259:71 on June 17th Changed the tags. Removed: ?creditg?r Tara Stewart - at 53:25pm on iune 17th changed the tags? Ad?ed: ?r?sk cred?ts progects" Tara Stewart at 1:09pm on July 8th The increased concerns about ob rater from devs reminded me of the second sort of this project If the devs are reaiiy more concerned about and not refunds it cou?d make sense to start refunding for blatant FF-minor. Ross and wit? have to rev?sit this; data- when he has time? Tara Stewart at 8:45am on iuly 11th changed the subscribers. Added: joseph Ross Worden at or: Juiy 13th Th?a shouid work (running Err hipa??i?wlm) and you can we whatever else you wan? to the top SELECT Ciause from SELECY q.objectic?mx ag uid, as tagid, qtagdatemx as tagdate, gemson?objec?afeatures, ?S.Reprpid?) as appid, uname as appname, a.order_id, as (CASE WHEN aref?undyt?me :w 0 OR arbargebackwt?me 0 THEN 1 ELSE (3 END) as fraud, gethonmobwc?da?eatures, ?S.UserAge') as UserAge FROM 1 SELECT as objectidwx, xiagid?y as as tagdatewx FROM SELECT yobg'ectid ag objectid__y, magid 35 tagid?} wig as; 6331? FRQM objectmtoutagwassor: WHERE magid >2 GROUP 8? xobjectidwymiagidj )cr LEFT QUEER JOEN a GM {gobjec?dmx getmgsonwobiec?,afeawreg JIKND ads LEFT JOEN ob__iaunr:hedwapps uOf?? ge?rjsonmobjec?afeaturea Tara Stewart - at 53:54pm on jury 13th That data jg pretty sweet. i iike that you jug: plugged it in at the app lever. it is de?nite?y different than the data set we were origmaiiy Jeokimg}; at, but We w?th that. i?i? make aome pivot charts tomorrow. Case Document 193-8 Filed 01/24/19 Page 8 of 15 Tara Stewart at 4:0?pm on Jute 14th added an attachment that was later deleted Tara Stewart - at 4:08pm on July 14th Underagedatappievel is fun to play around if you are interested in app-tavei iossea per user? Tara Stewart at 11:30am on juiy 18th 1 added the attachment Tara Stewart at ll?Sam on Juty 18th Fixed the data in underagedataaopievel. You can now research the following: 1. Total 3053 and good spend in FBZ pet Underage Tag oer App 2. Totai loss and good spend F82 pet L?i} oer Age The data set includes transactions that occurred between 10/12 and 1/12 from usets that were tagged with one of the three minor tags. The; anvii'onment has Changed a bit since then (with some different agape arising as moat popular with the under?l8 set). We couid use the numbets in thia report to determine if we should start refunding for specific apoa in CRT at a certain threshold. l?rn open to suggestions on the heat way to thia (tater, Tara Stewart at 2:58pm on Juiy 20th i was taiking with ioseoh and Bate yesterday about how we shoutd age thia data to choose apps that wooid be good recipientg for an automatic underage in?ow. Lossy Fmeinor heavy apog that come to mind: PetVilie, Haooy Aquarium, Wild Ones, Barn Buddy, and any Ninja game. We can check out a rule catered to users under the age ot18 and over the age of 100 playing theae games: to guess how many transactions this would impact. i?rn for putting underage users into this inflow that are trying to purchase more than $275 in a single transaot?on or who are attending at smaller levels with a high frequency. We?ii gee what the data shows.? Tara Stewart at 2:57pm on July 22nd Backtested a rule with the foliowing oonditiona: WHERE AND AND ((getjsonwobiectiiteatures, ?S.t?serAge?} OR (getjson_objecttdieaturea ?S.UserAge?) 93)} AND (gemsongbiectictfeatures, 'SAmountReceiving) 75000} AND ({gethonwobiectid.features, OR (gethonmobjectidteatures, <2 GR {geusomobjeotidfeatures, a GRGUP B?r? getmg?sonmobieotidteatutes, 'S.Apeid?} Case Document 193-8 Filed 01/24/19 Page 9 of 15 Since the date range was only 6/1 to 7/19, the "false positives? aka speedy kids, aren?t necessarily valio, Pm betting most of the spend for this age range will be charged back. We can use these conditions to inflow for specific FF?heavy appS. ltestecl with PetViile, and this rule iooka BOMB. Tara Stewart at 3:31pm on July 2206: 0k, after consulting with Josh, we?ve determined that it makes; the most sense for now to focus on the Buy Credits-CC context. was not designed to work in Spend. The only actions you can use in Spend are too drastic for potential FFminor cases: Block, Verify Compromised, DPO, Disable, Blacklist. The downside is that we have no appropriate action right now to take ifa chilci is using his parent?s PayPai account. lim going to test again for PetVille in the Buy~Credits CC context for the last week only to see the kind of results we could get. This is a good first stepi Tara Stewart - at 1:27pm on iuly 26th it added the attachment Tara Stewart - at 1:40pm on July 26th underbacktestjulylto24 shows the resultg for V1 of the CC inflow rule for the buy context Keep in mind that this is unlike traditional frauo~baseo ruies in that we don't take action on underage Spend until after the fact (we don?t ?know? it is fraud untii a ct: occurs or a refund is issued}, Basec? on theae numbers (we needed a large enough sample size) combined with corework insights, v1 is going to run on the foilowing BQQS: 1. Pet Society (11509831134l: popular app for minors; that already has fraudooecif?c rules pushed for can gee how this underage. inflow the rules already in place to drive down the ob rate 2. Backyard Monsters (342684208824): ?green? app in that we haven?t pushed any speci?c rules for minor population 3. EA SPORTS Superstars: Real footbail 8i sooceri app with a large minor population For v2: i?rn going to look at apps that we know are FF-heavy but didn?t have a high enough sample size Monster Galaxy, Ninja Saga}. l'rn going to pull anti refunds; that were labeled as; FF~minor to check out the character?gtics of those accounts. Some ideas for looking forward; We need to determine an appropriate response for Fiat/Pal users. Maybe auto-ectioning on usera, regardless-2 of app, once they are tagged with ?suspecttooyoung.? Monitoring and modifying the v1 apps. Adopting a lower spenei limit for all underage userg. Tara Stewart - at 1:41pm on Jute 26th assigned the task to Ross Worden. Tara Stewart - at 1:41pm on July 26th changed the priority to "hi-pri". Ross Worden at 8:25pm on Julti 28th. Case Document 193-8 Filed 01/24/19 Page 10 of 15 {User?tge <2 37 i UserAge 90) AmountReceiving >2 75800 {mi i 0 CreditsBaiance (m 0 I i 8182: Remppid :2 11509831235: i ReprpicE :2 342684288824 I I Ref/Amid :2 3.137196253324737? i theughts? Tara Stewart at eh juiy Looks good, Ross! RQSS Worden at Bz?gam 0r?; why 27th Atwa Husain, can you push this in buy/?cc with verify?tst?? Ross Wardeh at: Ez?gam cm 27?th changed the Added: Arwa Husaim I Ross Worden .. at 9:53Bam on Juiy 2?th asgigned the tagk Arwa Husain. Arwa Husair: at 631m": or?: juiy 27th UnderAgein?mel is throwing Verify?rst? at use-rs and cantmi gmup E5 queuing t0 hipri Arwa Husain at {5:31pm an Juiy 27th assigned the tagk Tara Stewart: Jesh Kriveghein at; 9:E3pm QFE Juiy 28th Pm net sure that the respense i5 actuaihyr an effective deterrent t0 FEMEE we don?t see Cb reduction, it may not be the targeting {if the: mie. Note that a haw arder id is generated upon compietioh 0f the in?ow, so ?guring out which ids we threw iefiew at that received abs is tricky. E'ii see if the infiew be: inducted in the nexus metrics that Yihua is; working cm (it may just be the compromised inflow he?s making at). Joseph at ?Pr-Mam on Juiy 253th. isn't the whoie paint of first 6 in?ew it} combat Whither? CONFIDENTIAL Case Document 193-8 Filed 01/24/19 Page 11 of 15 Aprii Wharton - at 7:?i83rn on July 29th changed the subscribers. Added: April Wharton. Tara Stewart at 7:53am on Juiy 29th Josh Krivoshein. obviously isn?t the most ideal deterren?i but could be a good ?rst step. it: forces the minor to prove he is in possession oi the credit card. Often refunds/cos occur because a parent permits his child to spend at a smell denomination arid doesn?t realize that the cc info will be stored. Obviously Some kids will be abie to grab {he CC again or write the info down, but this will hopefuiiy curb the spending of the least savvy mirrors. i hadn?t been piarming on looking at the specific order ids that were inflows/ed, out rather the overall spending of the 3.1le that: are flagged. i?m definitely open to suggestions! Tera Stewart - at 8:12am or: Jilly 29th Also, Arwe Hosain, when we greoped this rule today ii; didn't hit anything. Couid you check the syntax? Should be in the buy creditecc context: and look something like this: (Userege 1? i UserAge be Amountheceiving 75000 8:8; 11609331134 I RelAppld =2 342684208824 Reproid 13.337196253247337 i i think the problem might be because i iirsi: tested this ruie in the speed context: and had the logic to remove aooZuser purchasers. li? we kept thig logic: in (which we probably did), it might be messing things up. Ryan Keily at 8:25am or; Juiy 29th changed the subscribers. Added: Ryan Kelly. Josh Krivoshein at 12:06pm on July 29%} Joseph Filip the whole poinr. of first 6 inflow is to combat FFl?viirior, but i. wouldn?t make the assumption that; it is actually effective at doing this. Hopefully, analysis of the roles created by this task should help confirm this or nor. The response was created about 9 months ago, when FF was a big issue. Since that time, it has become less of an issue, but this could be due to other factors (people became more used to credits after all head devs were onboarded, and we new process a lot more volume which drowns out the iiT in terms of ch compliance) rather than the impact of the ?rst 6 inflow response which is on very few rules. i checked with Yihua. the inilow metrics are in the nexus dev tier and will be pushed to prod soon. They do include metrics regarding however they are currently measuring compromised activity that occurs post inflow. i?ii work with Yihua to cheoge this to measure FF that occurs after this inflow. Josh Krivosheio 3:12:1{3om or: hilt; Z?th ?i added the attachment ?card verification io?fiow statsong?. Joseph Filip - at 1:80pm on July 29th Josh Krivosheir?i their is El primary reason we are roor?iiog this. it lei: os know weather this approach even works, and if first 6 in?ow is the right/effective tooi to get the job donei This won't be great at stopping FFAciult or buyer regret but it should keep kids from running rampant with their parents which is a orobiem in certain apps. ll} Fidd?iv??filli??l Case Document 193-8 Filed 01/24/19 Page 12 of 15 Tara Stewart at 3:21pm on August We?re going to have to recheck the syntax since this ruie still hasn't hit any UiDs. Ran the becktesting query from 7/28 to 7/30 and it should have hit 11 UiDs in EA Sports, 6 in Backyard Monsters, and 5 in Pet Society. Query 467948 "fare Stewart at, 3:21pm on August is?: ?3 assigned the task to Arwa Husain? Ross Worden at 71G73m on August 2nd greo'ing showa 23 tags or: since 23m this morning hours elapsed} Arwa Husaio at 8:31am on Auguet 2nd essigneci the task to "fare Stewart Tara Stewart at 9:03am on August 2nd Just perused the first Sampie of accounts that hit this; rule. Looked at: 18 tags] of which 10 were unique, perhaps showing that minors; are more likeiy to attempt to make raplo, highncost purchases in a row. The inflow did work to stop some lie 1311397878). cher minore passed the inflow. Oniy one user in this group was over 18 (he iisted his age as 106 and was olaying Sackyard Monster?s). We wili continue to monitor the hits to this rule to iearn more about spending patterns. We also iook to see how totai spend is affected in these apps. iara Stewart at 3:52pm on August 2nd Seems iike most of thee-e games with FF~minor problems are defauiting to the highest?coat setting in the purchase flows. This: oniy exacerbates the prooiem since it doesn?t necessariiy iook iike "real" money to a minor. i?m going to puii some data for Ftheavy apps to see how many detaizit doilar transactions; have already been refunded/charged back in the past. week? We can compare this to lower, non~default amounts to see if the chargeback/rei?und to good percentage is higher for the default setting, if so, we car: recommend that they default to a iower doliar amount. Tare Stewart at 9:043m on August 4th changed the subscribers. Added: Doug Fraser. Tare Stewart at 9:87am or: Aogust 4th Action items from meeting on 8/4: Tare continue FF minor aoaiysis; infiow study; minor spend iimit analyais 2. Aprii - FF aduit anafvsis, Chargebaoitw specific 3. Doug User coofusiori analysis, refund~speci?cg root?cause analysis 4. Joseph SMS messaging; FF minor Meeting in a week to talk about resuits. Tara Stewart at 823mm on August 16th 11 Case Document 193-8 Filed 01/24/19 Page 13 of 15 i've been neglecting this task and focusing on iossy app rehab for the past week. However, the refund rates for our test apps have destined to normai ieveiss, though we'll have to puli data to determine how much of this can be attributao?e to the new inflow oroeess. From a quick skim of accounts that were recently hit by our Vi inflow ruie, it appears; that whiie a significant portion can pass in?ow, they typicaliy go on to make iower denomination purchases (again this is just anecdotal). in the queues, the FFrninor l'rn seeing for these test apps typicaily occur on their parents? accounts or accounts where age>20. l?il revisit this when i return from PTO, but any pressing questions can be directed to Joseph Filip. Ayse {Jibeia at 1:453:11 or; August 25th 4? changed the euoscribers. Added1Ayse Dibek. Tara Stewart at 23:41pm on August 30th Joseph Filio, which eoo? do you think wooid be good to look at moving forward? Off the too of my head, i?rn thinking Car Town and PetVille. Joseph at 5:12pm on August 30th Tera Stewart, these are eerie teat: have been identified as having strong/moderate F?Minor probiems contributing significantly to iosses. Sociai Empire-:5: Pocket God: Wiid Ones: MiniPienet: Ninja Wart: Diamond Dash: These wouio be excelient starting ooints as we begin to roii out this rule more broadly. Tare Stewart at 9:13am on September 6th Talking with Pat and Josieoh now about potentialiy provioing suggeetiong to developers about ways to deai with FF. Since you authorize an ago to access your information, and the dew can thus see your birth year, it could be beneficial to provide different price packages to minors. instead of the max package of $100 or 3150, they could offer a max of $50 or something if age 18. We?li discuss this further later on in the week. Tare Stewart - at 4:35pm on October 25th i?ii Arwa, couid you pause UnderAgeinfiole? We should stop and anaiyze the date it this is something we want to continue to work on. Thanksi Tera Stewart at 4:35pm on October 25th changed the priority to "iow-ori?, 12 SQNFEQENTML Case Document 193-8 Filed 01/24/19 Page 14 of 15 Tara Stewart ., at 4:35pm on October 2E3th assigned the task to Arwe Husain. Arwa Hesain at 10:34am on October 31st 3 have deteted Underageln?ewtfl from sigma buywcc Arwa Husain at1023??ram on Octohet 331st assigned the task to Tara Stewart Bennett Wee - at 2:59pm yesterday changed the subscribers. Added: Bennett Woe. Tara Stewart at 4:47pm Already pasted th?s 0n the MC EE discussion, but nesting the relevant data here tee: Regarding the Fest-6 inflow Response; just ran the data far my friendly fraud inflow etoject. This rule targeted users undet the age of and above the age or" 9&0 who were trying te spend greater than or eeuat to $75 in a single in Pet Society} Backyard Monsters, er EA SPORTS FIFA Superstars: Real football soccetl. The rule respense was the ?rstw? CC in?ow. 9 included refunes and chargebacks in the same tags bucket because together they constitute totai frienely frauds This rule was active from 8/2/2011 to 13/31/2311. Users who passed intiew and made subseeuent tans in any at the 3 test apps: 13.3% of the count of tens were charged back or refunded or" the sum tota? m" tans were lost as a result. of chargehacks s? refunds 14.9% of these users (ceunti charged back or were refandee Users in the central group fer this rule: 19.2% of the count were charged back at refunded 393% of the sum tote; of tens were lost as a result at chergebacks refunds 13.2% of these users {want} charged back or were refunded Users who passed inflew and dtdn?t spend in the 3 test apes or users who failed inflow: Frem a statisticai?y significant sampte, 6% at these users charged back er were refunded. 32% of users in the inflow greup taf? inte the, segment. The results my project indicate that ??rst-6 infiaw aitered the chargebatk and refund sates While a smaller percentage of the centre? group users charged hack er were refunded, they charged hack a greater number at transactions and a target total sum. Fewer transactions were ultimateiy charged heat: er refunded from the users who passed inflow. Though the in?ew aidn?t make a huge dentin the less rate, it blocked a fair ameunt et users from purchasing in the test apps. if this were run at a larger State, ?t could decrease the denominatnt of the segment we are targeting. ONFEBENTSM FE?lEl-?Q??m?a? Case Document 193-8 Filed 01/24/19 Page 15 of 15 Anecdotally, many of the miners who did not spend in the test apps after inflow went on to Spend in other apps. Unfortunately, due to time constraints, {didn?t have time to dig deeper into the users wire passed Enfiew but didn?t spend or all~outfailed the ir1flow.?l?his could be a next Step. Ultimately, I would recommend a ?rstn? inflow to the lower threshold users. To see this task, gr.) tn: