Executive report
Data miningfor
lotterysystem lotterysystemss
®
D a t a m i n in g f o r lo t t e r y s y s t e m s
Abstract The challenges of creating and maintaining a lottery system t hat appeals to customers a nd provides revenue for your state’s vital programs can be eased when data mining using SPSS products. When data mining with SPSS, you use h istorical data and become e mpowered to analyze an d improve performance, survey and predict customer needs a nd re port results to a broad audience.
Introduction: history of the modern American lottery and the challenges lottery systems face today Lotteries in t he United States d ate t o colonial times. In fact, King James I of England used proceeds from a lottery to help fund Jamestown, Virginia, the first British colony in America. Since the n, lotter ies in the United State s have played a numbe r of roles. The modern American lottery has its roots in a lotte ry created by New Hampshire in t he mid-60s (a fter a 60-year Supreme Court ban on lotteries). Other states soon followed an d crea ted th eir own lotter y programs. By 1971, lottery ticket sales in the United States passed $100 million. According to the National Association of Stat e an d Provincial Lotte ries ( NASPL), 38 U.S. sta tes, the District of Columbia, Puerto Rico and the U.S. Virgin Islands all offer government-operated lotteries. Year ly revenue s are in excess of $38 billion nat iona lly. Massac huse tts a nd New York lead t he way with sales of over $3.7 billion an d $3.6 billion, re spec tively ( fisca l 2000). Today, with lotter y drawings ann ounced nightly on local newscasts and stories about avera ge Ame ricans becoming instant millionaires, the lotter y is a b ackdrop to American life in the 21st Cen tury. And, becaus e modern Ame rican lotteries fund vital programs — for example, 14 states allocate 100 percen t of the ir lotte ry profits for educat ion — crea ting profitable lotter ies that int erest a variety of customers across your stat e is a must. How does your depart men t evaluate customer preferenc es to determ ine what’s neede d to increase revenues? Would it help if your state lottery could predict the key features, color scheme s and other ch aracte ristics tha t appea l to the most people? How does your lottery departm ent in clude new players and m aintain th e inter est of curren t players? What if your departm ent c ould use historical data to determine t he region in which games have the most success and what time pe riod games experience the m ost profit? How does your depart men t communicate lottery data and results to the appropriate a udiences? What if you had a m eth od in which you could easily and c learly report results to th e right pe ople? Data mining with SPSS empowers you to combine your business knowledge about your lottery system with a nalytical techn ologies t hat tap into your historical data so you can predict the future. When da ta m ining with SPSS, you become empowered t o analyze and improve performan ce, survey and predict customer n eeds and report results to a broad audience .
Data miningwithSPSS Data mining enables you to discover unse en pat terns a nd relationships in large amounts of data. Powerful analytical technologies quickly and thoroughly explore mountains of data, isolating valuable, usable informa tion — the bu siness intelligence — tha t you nee d to improve your lott ery syste m. Data mining with SP SS gives you a reliable guide to t he future of your lottery system, so you ha ve the power to ma ke the right decisions today. Data mining empowers you to ch ange t he futur e of your lottery system, by delivering accur ate p redictions. For example,
®
Executive report
2
D a t a m i n i n g f o r lo t t e r y s y s t e m s
data mining tells you which game at tributes are most likely to inte rest your customers. With this view of the future, you increase your return on investmen t ( ROI) by marketing th e right type of game s to the right types of custome rs. Your decisions are base d on sound bu siness intelligence , not on instinct or gut re actions. And those dec isions de liver c onsistent results that can h elp keep your lottery system p rofitable for years to come. SPSS is the data mining leade r, with over 30 years experien ce de veloping field-test ed a nalytical software. And, when data mining with SPSS, you can be assured your project is conducte d using proven da ta m ining met hodology. SPSS follows the CRoss-Indu str y Stan dar d Proc ess for Dat a Mining ( CRISP-DM) . Creat ed by industry e xperts, CRISP-DM provides ste p-by-ste p guidelines, tasks and objectives for every stage of the data mining process. When data mining your lottery system data with SPSS, you can :
Increase lottery revenues Make be tter use of your ma rketing dollars Identify the best , most profita ble new games Better understand your customers Share information with decision ma kers, the pu blic an d your customers
This white pape r de scribes how you can perform data mining with SPSS products t o solve the se business problems. This white paper a lso feature s a product/ap plicat ion guide to show which data mining products your lottery system can use t o solve business problem s and a sample application story, describing how data mining can help you find a nswers to your tough business que stions.
Increase lotteryrevenues Over t ime, once p rofitable lotte ry system s may experien ce lagging sales. Man y factors can contribut e to this. For example, do economic conditions — wheth er good or bad — have any relationship to lotte ry sales? Does your state lottery have higher pr ofits at cer tain t imes of the year? Perhaps your lottery system generates more revenue during the holiday season when people buy lotte ry tickets t o give a s gifts. Declining revenue s after an initial period of growth is a common experien ce for lottery systems. States t hat newly adopt a lottery system ca n expect t o see a trem endous am ount of revenue growth during the fir st few weeks, months or even years. The a mount of publicity and med ia att ention initially given to a n ew lottery system in a stat e genera lly is quite high. Accordin g to David Gale, executive director of the North American Association of State and Provincial Lotteries (NASPL), lottery sales will eventually experience a downward trend in revenues. Factors tha t cont ribute to downward sales include initial exciteme nt wearing off, jackpot “fatigue” or simply boredom playing the s ame games. Most commonly, the novelty of the first large jackpot for a state lotto drawing wanes an d must be surpassed by a bigger jackpot in a future drawing to generate the same level of interest. Eventually, people become unresponsive to drawings that once gener ated a great deal of atten tion. Lottery statistics show tha t approximat ely 75 perce nt of customers play only when jackpots are la rger tha n $15 million. And even if your state lottery has strong returns, what can you do to ensure the system rem ains profitable? When you mine your da ta and a pply your business knowledge, you can predict what factors increase and decrease revenue.
®
Executive report
3
D a t a m i n i n g f o r lo t t e r y s y s t e m s
Performing analysis in SPSS Trends or DecisionTime/WhatIf?, as part of your data mining solution, can give you valuable insights for more strat egic marketing plans th at le ad to incre ased re ven ues. SPSS Tren ds gives you an easy-to-use gra phical int er face to an alyze hist orical informat ion a nd predict future events using time-series an alysis. DecisionTime is the best way to quickly creat e powerful foreca sts using SPSS’ proven, hea vy-dut y analytics — the kind spreadsheets don’t offer. While WhatIf? empowers you find the answers to tough questions, such a s “How will chan ging the odds of the lottery affect our cu stome r’s purch asing pat tern s?” Both SPSS Tren ds a nd DecisionTime/WhatIf? include “inte rvent ions” an d “events” in m odeling. Interventions and events are outside occurren ces th at may affect t he series you’re forecasting. Examples of interventions and events include an increase in the n umber of numbers drawn, unexpect ed events t hat p reviously affected sales, or even a cha nge in consumer confi dence. Clementine, AnswerTree and SPSS also can help you predict factors that may increase sales or alter cu stomers’ buying patte rns. For example, you could use Capri for Clemen tine to det erm ine the sequence of events in a person’s decision to purcha se a lottery ticket or the sequen ce of events your system n eeds t o take for market ing profitable games. Perha ps you would like to examine t he pr edicted seque nce of events tha t occur when you change your lottery’s odds and how this chan ge is likely to affect profits.
Make better use of your marketingdollars Man y lottery systems face the challenge of reaching the most poten tial players with as little mar keting dollars as possible. According to th e NASPL, in 1996, North Ame rican lotterie s spen t $400 million (USD) on advert ising and re ceived $34 billion in sa les. Advertising expen diture s accounte d for 1.17 perce nt of total revenue. By contra st, restauran t owners spent 3.2 percent of their re venue s on advertising, beverage m anufacturers 7.5 percent, c osmet ics companies 8.8 perce nt, an d can dy ma kers 12.7 per cent. Advert ising account s for less of the cost of a lotte ry ticket th an virtually any other consumer product. With a large percen tage of revenues usually man dated for various causes, the percent age of revenue s allocated toward advertising is unlikely to increa se. You can achieve bett er ret urns on your market ing investm ent by mining historical sales d ata, survey data from the public and survey data from reta ilers. Data m ining gives you the insight ne ede d to ma ke bet ter use of your limite d mark eting dollars so you can more accurately reach people likely to buy tickets, ensure c ustomers play the lotter y on a regular basis, iden tify regions most likely to embrace a pa rticular game and so forth.
Identify the best, most profitable new games What if you could know the key feat ures th at a re m ost likely to int erest customers in p laying new lotter y game s? What if you could predict t he customer group most likely to purch ase tickets? What if you knew what kind of game th eme s appea l to the most customer s? Gen erally, lotter y systems create new games based on results from surveys taken in focus groups, games solicited or reque sted by the public directly to the lottery depart men t or games solicited by ticket vendors. Take your new game research further to confidently develop plans based on both your business knowledge and historical data — to market games that are m ore likely to result in higher profits — when you data mine with SPSS. SPSS can give you the t ools nee ded to conduc t your own survey research an d apply data mining to the responses (see t he “Better understand your custom ers” section in this pap er for more informat ion about survey rese arc h). Using Clement ine, AnswerTree, SPSS or any combinat ion of these p roducts, you can a nalyze your survey data or h istorical data for bet ter game development that leads to higher revenue. ®
Executive report
4
D a t a m i n in g f o r lo t t e r y s y s t e m s
Lottery systems a re increa singly challenged to cre ate gam es tha t appea l across many regions in your sta te. Your sta te m ay have a crossword-the med s cra tch -off game t hat ’s ver y succe ssful in certa in area s of your state but is an average seller in other area s. Perhaps cha nging the color combinat ions — such as silver/blue instea d of red/purple — for this scra tch-off game in the average-selling areas will increase your revenue in those p laces. Data mining can help you predict the color combinations t hat will be m ore successful in specific regions of your state. It ca n also help you an swer a num ber of other que stions when developing new games or redesigning existing games, including:
Are certain shapes more appealing than others? Do certain font styles produce more profit able scrat ch-off games than others? Are certain reta ilers, such as convenience stores or large grocery stores, more likely to successfully market a nd sell tickets?
Better understandyourcustomers Like man y lottery systems, your state may outsource demographic studies to an indepe ndent research firm. Surveys may seek answers to th ese types of questions:
“How often do you play lottery games?” “Do you wait for lotte ry jackpots t o reach a cer ta in am ount b efore playing?” “Would you be more likely to t he play the ‘Word J umb le’ scrat ch-off gam e or the ‘Basket ball Cham ps’ scra tch-off gam e?” “Do you pre fer highe r pa yoffs or be tte r odds of winning?”
These studies are crucial in determining your lottery system’s marketing practices, new rules and games. However, independent research firms lack th e expert knowledge your organization has about how you run your state lottery (such as, state lottery rules) , your customers an d your retailers. Using a combination of SPSS products to collect survey research data, analyze them and deploy them to the proper audiences, your lottery system could bring research inhouse and use a combination of data c ollection software and d ata mining software to bette r under stand your customers. Use products in the SPSS Data Entry family to bring the power of surveys to your state lottery dep art me nt without t he ne ed for program ming or Web publishin g knowledge. Your staff ca n create surveys that you can deploy to focus groups and other survey respondents (such as your retailers) on t he Web, pap er, phone or t hrough pe rsonal inte rviews. You can also deploy surveys on kiosks in supermarket s or convenience stores without needing an Internet connection. This mean s your depa rtmen t can easily reach your potent ial respondents in a variety of formats. Feat ure-rich design tools and a library of proven s amp le questions, which you can a dapt to mee t your lotte ry system’s need s, help you write qu estionnaires th at e ncourage complete a nswers and minimize data ent ry errors. When you use t he SPSS Data Entry Enterpr ise Server, survey responses are saved in SPSS file format for instant ana lysis in SPSS. These data can a lso be read directly by Clemen tine a nd Answer Tree. Using SPSS Data Entry products, your st aff can de sign a form t hat allows a sur vey responde nt t o ent er th e ZIP code in which they live. SPSS Data En try Enterprise Server can au tomatically calculate th e geograph ical “Area” for the r espondent and then display questions you want to ask people living in th at par ticular area .
®
Executive report
5
D a t a m i n in g f o r lo t t e r y s y s t e m s
If your lottery system has difficulty finding focus groups that reach residents from across your state, deploy your surveys on the Web to allow for more complete coverage of every region. With Web-deployed surveys you have acce ss to a greate r an d more diverse sa mple of the popula tion. Also, Web -dep loyed su rveys ar e also more cost e ffect ive in t he long term beca use your organization doesn’t ha ve th e adde d expense of paper and p ostage that apply to mailed surveys or long-distance cha rges associated with phone surveys. Using data collection software in-house and t hen using data mining to analyze the da ta from surveys mea ns th e pe ople who know the most about your lottery system are t he people analyzing survey data. This gives your department more intimate knowledge about your customers — and re tailers, if you choose to survey them as well — so you can m ake bet ter decisions that lead to increased revenues.
Share informationwithdecisionmakers, the public and your customers Once you have data mining results, it’s important to communicate the m with the people who can or ne ed to use the m. Your audience m ay vary a great deal — in term s of what th ey need to know an d how th ey’ll use results. SPSS product s give you a variety of report an d dep loymen t options so you can communicate results to a wide audien ce. The people who will ultimately make decisions based on r esults want information to be clear and described in non-tec hnical te rms. They’ll likely want t o explore results on t heir own an d see how chan ging variables, such as decre asing ticket prices by three, five percent or seven percent for scratch-off games, affect results. If you use DecisionTime to create forecasts, analysts can use WhatIf? to deploy forecasts on t he In tern et or intran et. Decision maker s can th en a sk “what if” questions about the da ta in the forecasts to find an swers to poten tial scena rios and ha ve the information ne eded to ma ke th e best, most-informed de cisions. If you’re dat a mining with Clemen tine, your depa rtme nt ca n use Clemen tine Solution Pub lisher for deployment . Clementin e Solution Publisher en ables data m iners to crea te dep loymen t applications quickly and cost-effectively, and because re-deployment is effortless, your depar tment can ke ep up-to-date with changes in your lottery system. Since your lotter y system likely funds vital program s for your stat e, it’s possible you n eed to dep loy informat ion to more th an just your decision mak ers. You may be requ ired to re port results to your legislature, a state agency or a stat e a ccounting office or local school boards, which rely on lott ery revenue for fund ing. You might ne ed t o keep your pu blic updat ed on t he payoff amounts and odds of winning. SmartViewer Web Server, the analytical content portal for SPSS products, enables you to easily share information with a variety of people. Using SmartViewer Web Server, analysts publish reports from SPSS into a secure database. Then, your audience can view the reports, including interactive reports, using a Web browser. Smart Viewer We b Server en ables you to quickly sha re import ant r eports a nd informat ion with anyone you choose. Make a ppropriate information available t o customers, the public and other groups in real-time. With SmartViewer Web Server, you decide who receives reports and how much they see. Confidential, sensitive or fee-based reports sta y out of the wrong han ds. And importan t informa tion gets to the right people — on time.
®
Executive report
6
D a t a m i n in g f o r lo t t e r y s y s t e m s
State lotteryneeds andSPSS capabilities Use th is table as a guide t o selecting products that can h elp your lottery system increa se revenues, make bett er use of your ma rketing dollars, identify profitable new game s, better understand your customers and share information. Visit www.spssbi.com for more inform ation on the se products or e-mail SPSS Sales at sa
[email protected] to discuss h ow your organiza tion can start data mining. Situational questions
Impact
SPSS capabilities
How does your department poll your citizen’s preferences for lottery-sponsored products?
Does your department find it difficult to obtain current information from a full crosssection of your population?
SPSS Data EntryEnterprise Server (part of the SPSS Data Entry family): What if you could use an intuitive graphic interface to quickly create surveys and forms for Internet, workstation, paper, phone and personal interview surveys? Would it also help if you could instantly analyze those results?
What type of analysis does your department conduct to determine which new products to introduce to your citizens?
Does your department find it difficult to accurately predict the combination of odds, cost, shapes or colors of your games that lead to higher revenues?
Clementine, AnswerTree and SPSS: What if, using your own data, you could more accurately predict the combination of game characteristics that lead to higher sales and higher profits?
How does your department forecast sales and revenue for the next week, month, year?
Does your department find it difficult to precisely forecast revenues given the irregular sales cycles for lottery products?
SPSS Trends, DecisionTime/ WhatIf?: What if you could include special, one-time events, legislation changes and rule changes in your sales forecasting analysis? Would it help if you had a product that’s intuitive for management to use but powerful enough for analysts?
How are you sharing the results of your analysis with your department or with outside entities, such as your state’s legislature accounting offices, local school boards, retailers and other agents, journalists and citizens?
Is your department’s decisionmaking ability limited because it is unable to access or use critical analytical results? Is it difficult to provide information you are required to report to outside entities in a way that’s meaningful to an audience that may not have as much knowledge about your state lottery?
SmartViewer WebServer: What if you could share the results of your analysis with a larger audience, including those who aren’t as knowledgeable about your state lottery, with differing information needs?
Does your staff need help getting started with data collection, analysis or reporting? Would you staff like guidance so it can deploy SPSS products more quickly and more effectively?
SPSS Business Intelligence Consulting: What if your organization could get the help it needs from data mining leaders, who can help your organization realize the full benefits of the SPSS products you choose so your staff can solve problems independently?
How will your organization implement its data mining solutions?
Clementine Solution Publisher andWhatIf?: What if you could quickly and easily deploy data mining results (Clementine Solution Publisher) and forecasts (WhatIf?) to decision makers.
®
Executive report
7
D a t a m i n in g f o r lo t t e r y s y s t e m s
Sample application: mininglotterydata todiscover factors leading to increased revenue Situation A state provided SPSS with historical data of instant ticket sale s to ana lyze. The state wanted to know how it could use historical dat a to find the m ost profitable and widely accepte d games. In addition, the stat e wanted to know in what region are t hose games most successful and for what period of time they are m ost profitable. The stat e p rovided a dataset containing information for a lmost 400,000 individual ticket sa les. These dat a re present ticket sales from J uly 1999 through J uly 2001.
DataminingwithSPSS to discover factors that leadto highnet-profitgames SPSS Business Inte lligenc e division considered a num ber of data mining tech niques a nd then tested several techn iques on the available data. SPSS generated relevant results using tech niques in AnswerTree a nd Clement ine, looking at t he da ta in both th e conte xt of the ent ire state and also on a regional basis. Without t aking into account any geographical region, SPSS discovered the most importan t factor in pred icting game profita bility is the n umber of months tha t the game is maintained in the public. SPSS used data mining technique s that e xamined aggregated n et profits and game the mes. Using this state’s data , SPSS discovered tha t crossword game t hem es have the highest overall profit, if no othe r factors are t aken into account . From this information, SPSS concluded that a combination of a crossword them e and t he optimum nu mber of months the game is available on the market can generat e larger net profits.
How can lottery programs predict the optimum amount of time that it should offer an instant ticket game ? Data mining empowers you to take a ll of a game’s attributes and historical sa les data and use t hem to predict the ideal number of days, weeks or months that a game should be in the market. Here, SPSS used a rule induction algorithm, Classifi cation and Re gression Tree ( C&RT) in AnswerTree, to create this prediction. The C&RT algorithm uses a tree-based classifi cation a nd prediction method. C&RT models tend to be e asier to understand than some other model types beca use the rules derived from this model ha ve a very straightforward interpretation. ®
Executive report
8
D a t a m i n i n g f o r lo t t e r y s y s t e m s
Using the Classification and Regression Tree (C&RT) model in AnswerTree, SPSS discovered tha t for net profits of greater th an $1 million, a game should be in th e mar ket for less than 5.85 months. SPSS confirme d this result by runn ing a C5.0 model in Cleme ntine . SPSS selected the C&RT model because it achieved a higher confiden ce level, or the probability th at a predicted event will occur. However, because it’s best to run multiple models on a data set in order to compare the a ccuracy of results, SPSS also used the C5.0 model. A variety of algorithms and techniques
for better, more reliable models Different types of data work bette r with different algorithms, and your state lottery system probably collects ma ny types of data tha t ar e useful when data mining. It’s also likely your d at a ch ange over t ime a s you create n ew programs and experience a cha nge in player d emographics, among othe r factors. Therefore, you nee d t he a bility to try different algorithms a nd techn iques with your data t o find the best fi t. That’s why SPSS gives you a wide variety of algorithms and techn iques from which to choose. Using a range of algorithms a nd t echniques e mpowers you to build better models and compare results against results obtained from an other mode l to dete rmine which one works best for your da taset . For example, Clemen tine gives you a variety of tech nique s so you can classify, cluster and segment groups, dete ct associations, do data reduction and create custom algorithms. AnswerTree gives you four t ypes of decision tre e algorithms from which t o choose to h elp you r eveal segment s and predict h ow groups will respond. And, SPSS gives you a variety of flexible algorithms for describing your data, predicting nume rical outcomes an d iden tifying groups. Using SPSS add-on m odules with SPSS Base gives you even more options for prediction, classification, segmentat ion and time-series an alysis. Your departm ent can u se some of these products or use the ent ire ran ge of software available for data mining.
More specifically, the grea test profit levels occurred with games th at ended anywhere between 4.3 months to 5.05 months a fter t heir start dates. This prediction has a confidence level of almost 100 percen t. Some game s also generat ed high profits after more t han 16 months in the market, but a s expected, the average mont hly profit level is consider ably lower. For game s that th e state kept on th e market for less tha n 5.85 months, SPSS examine d secondary char acter istics that d rove th ese games to highe r profit levels. SPSS discovered t he following:
For games in the m arket for the ideal time p eriod, the second most important variable in producing high profits is the game’s key feat ure. The following key featur es generat ed t he highest profits in the shorte st am ount of time: Myster y Numbe r, 15 Ways, Multiplier, Bonus Word, Blue Line , Extra Number s an d 30 Ways. Games t hat cost $3 or $5 to play have a greate r likelihood of ach ieving net profits of great er th an $1 million. With a confidence level of 70 percent, SPSS discovered that a game that costs $10 to play generates net profits of less than $500,000. June generate d the h ighest total profits or net re venu es. However, games started in March an d August had the highest n umbe r of high-profit game s. Focusing on March an d August led t o no stan dout variable in defin ing why those two months h ave the m ost highprofit games. It’s possible t hat demographic information attr ibutes to t his discovery.
To bette r understa nd how demographics influence lotter y profits, SPSS examin ed data describing the gaming character istics from one area of the state . In Area X, SPSS discovered that to genera te ne t profit s of great er th an $1 million, key featur es should be: Myster y Number, 15 Ways, Multiplier, Bonus Word, Bonus Blue Line, Extra Numbers, Lucky Numbers or 30 Ways. Net profits in Area X dropped by more th an 50 percen t on t wo gam es, Genera l Wild Joker a nd Crossword Doubler. In Area X, certa in ticket colors helped t o generate higher revenue. For ne t pr ofits greate r th an $1 million in Area X, SPSS discovered the following color schem es should be used: pink/gree n, red/pu rple, paste l yellow/orange, blue/yellow an d tea l/yellow. Certa in colors result ed in low revenu e: silver/gree n, black/blue, yellow/multi, blue/white , purple/green an d red/black.
®
Executive report
9
D a t a m i n i n g f o r lo t t e r y s y s t e m s
What if you could predict the fin ancia l success of a n instant ticket game before its release? In this case, SPSS used a neural network algorithm in Clementine to predict how successful a group of instant ticket games was likely to be base d on their attributes. A neural network is a s implifie d model of the way the human brain processes information. The network learns by examining individual records, generating a prediction for each record and making adjustments to the weights whenever it makes a n incorrect prediction. This process is repeated many times, and the ne twork continues to improve its predictions until one or more of the stopping criteria have been met.
Dataminingisanon-goingprocess The state lottery program is also inter ested in kn owing the idea l date to sell a sports-them ed card. For example, is it more profitable to sell a sports-them ed car d when a sports season star ts, at the h alf-way point of a season or close to the date of the play-off and cha mpionship games. However, the st ate curr ently doesn’t have data that would predict this date . In th e future, the state could collect data th at h elps it predict t he ideal date t o star t a sports-the med lottery game . The state might also want to know if retail locations play a role in profitability for sports-themed or other t ypes of games. For instanc e, do conven ience stores ha ve differen t sales patt erns th an large grocery stores? As the state lotte ry program continues t o collect dat a, it can contin ue data mine to improve its game s and market ing approach es to gene rate h ighe r net pr ofits.
Data mining empoweredthe state lottery department to discover:
For net profits of greater th an $1 million, a game sh ould be in the ma rket for less than 5.85 months. More specifically, the greate st profit levels occurred with games tha t en ded anywhere between 4.3 months to 5.05 month s after their start da tes. Secondary character istics tha t help drive these games t o higher profit le vels include the game’s key featur e an d price to play Specific key features pr ovide higher ne t p rofits in different regions of the state
Using the knowledge this state gained through data mining, it can cha nge its market ing plans to plan for more games t o last between 4.3 and 5.05 months a nd concen trate certa in key feature s based on re gional preferenc e.
®
Executive report
10
D a t a m i n in g f o r lo t t e r y s y s t e m s
Clementine ’s visual interface enables you to use extremely powerful algorithms without programming. The models generated in Clementine can then be browsed or deployed as code or as executable files a nd used to score new data that relates to the model.
Conclusion When data mining with SPSS, lotter y systems like yours can get a ccurate predictions using historical data, custom er surveys and your business knowledge. Results obtained from da ta mining can em power you t o increa se revenues, make bette r use of limited m arketing dollars, create new, more profitable games, survey and predict customer ne eds and rep ort results to a broad aud ience — and m ore. Data mining gives you the a nswers you need to overcome the ever-chan ging challenges your lottery system faces — and ensures your organization can make better decisions.
About the SPSS Business Intelligence division The SPSS Business Inte lligence division helps pe ople solve busine ss problems using sta tistics and data mining. This pred ictive te chnology enables our customers in th e commercial, higher education and public sectors to make bett er decisions and improve results. The SPSS Business Intelligence division software and services are used successfully in a wide range of applications, including customer attraction and retention, cross-selling, survey research, fraud detection, enrollmen t m anagement , Web site p erforman ce, forecasting and scientific resea rch. The SPSS Business Int elligence division’s ma rket-leading product s include SPSS,® Clementine,® AnswerTree,® DecisionTime ® and SigmaPlot.® For m ore information, visit our Web site at www.spssbi.com.
®
D M LTW P -0 4 0 2
Ex e c u t iv e r ep o r t
11