What data can tell you about customers

Big data is a con­cept that arrived with an appro­pri­ate­ly enor­mous amount of hype. We all know the dig­i­tal age has gen­er­at­ed huge amounts of infor­ma­tion and that this moun­tain of data is espe­cial­ly promi­nent in the finan­cial ser­vices indus­try. But the ways in which this avalanche of detail is being exploit­ed are many and var­ied.

Carmine Gioia is a vis­it­ing pro­fes­sor at the Mass­a­chu­setts Insti­tute of Tech­nol­o­gy (MIT). Although he is a res­i­dent of Copen­hagen, the Ital­ian aca­d­e­m­ic spends a lot of time in the Unit­ed States work­ing with MIT’s Big Data Ini­tia­tive. This is not abstract research as his work bears fruit in the world of invest­ment bank­ing where Pro­fes­sor Gioia also oper­ates as chief data sci­en­tist at Denmark’s Saxo Bank.

Saxo Bank has 27 offices world­wide trad­ing cur­ren­cies, stocks and bonds. Pro­fes­sor Gioia is con­cerned with learn­ing how to give Saxo’s clients the best pos­si­ble expe­ri­ence when they deal with the bank. Every time they log on to the bank’s web­site their clicks are analysed to try and work out just what it is that would most meet their needs.

“We want to make the expe­ri­ence of deal­ing with the bank as rel­e­vant to them as pos­si­ble, so we can offer them infor­ma­tion based on their pre­vi­ous search options,” says Pro­fes­sor Gioia. He is deal­ing with what he terms “rich data”, infor­ma­tion that flows from mul­ti­ple sources includ­ing social media. The algo­rithms Pro­fes­sor Gioia and his col­leagues devise are intend­ed to pull out trad­ing options that dove­tail exact­ly with the per­son­al pro­file of each client.

The key to the boom in data analy­sis is that analysing infor­ma­tion has become cheap­er. Cloud com­put­ing allows banks to store and assess data at infi­nite­ly low­er rates than was pre­vi­ous­ly pos­si­ble. But this brings its own risks. If an insti­tu­tion doesn’t give enough thought to how it goes about this exer­cise, it can still get over­whelmed with infor­ma­tion that sti­fles deci­sion-mak­ing. “Big data can mean paral­y­sis,” Pro­fes­sor Gioia cau­tions.

ANALYTIC PROGRAMS

US soft­ware busi­ness Birst writes ana­lyt­ic pro­grams that take data and pull out valu­able insights for clients such as the Roy­al Bank of Cana­da in Toron­to. Southard Jones is in charge of prod­uct strat­e­gy at Birst and points out that tasks which once required scarce skills are now accom­plished by the soft­ware his busi­ness rents to clients over the inter­net using the cloud mod­el. “In the past you need­ed data sci­en­tists to run the algo­rithms for data analy­sis. Hir­ing them was very expen­sive,” he says.

Today Birst can analyse the char­ac­ter­is­tics of account trans­ac­tions and queries using 40 dif­fer­ent aspects of a customer’s behav­iour, and iden­ti­fy the most prof­itable cus­tomers its bank­ing clients should con­cen­trate on.

The eco­nom­ics of this data analy­sis oper­a­tion are best under­stood in the con­text of data sci­ence costs. Birst can run a data analy­sis project for a bank for between £100,000 and £500,000. The ded­i­cat­ed soft­ware pack­ages that banks once employed for data analy­sis could cost well over £1 mil­lion and the accom­pa­ny­ing team of data sci­en­tists could expect to earn more than £100,000 a year each.

If these fig­ures illus­trate how data analy­sis costs have plum­met­ed, the oth­er side of the coin is the explo­sion in the scale of the data that needs analysing.

SHARP CURVE

Kx was found­ed in 1993 to address the sin­gle prob­lem of how to explore and exploit mount­ing quan­ti­ties of data held in com­put­er sys­tems. The tech­nol­o­gy avail­able to the finan­cial ser­vices sec­tor then is anti­quat­ed by the stan­dards of an online world where mil­lions of peo­ple check their bank bal­ances using mobile devices. Simon Gar­land, a British math­e­mati­cian who is chief strate­gist with the US com­pa­ny, talks of how data vol­umes have risen in a very sharp curve.

The key to the boom in data analy­sis is that analysing infor­ma­tion has become cheap­er

Cit­ing records, each of which rep­re­sents one trade or price quote, he describes the kind of infor­ma­tion the New York Stock Exchange (NYSE) gen­er­ates. “Since 1993, the NYSE has cre­at­ed one tril­lion records. But that vol­ume start­ed rel­a­tive­ly low and then rose sharply from the late-1990s onwards. Today the NYSE can cre­ate a bil­lion records in just one day,” he says.

Kx col­lects mas­sive amounts of data and then presents it in a way that oth­er sys­tems, such as data analy­sis pro­grams, can work on. “We pro­vide a very, very fast data base so peo­ple can act on inter­est­ing move­ments in the mar­ket,” says Mr Gar­land. Speed is of the essence in the world of trad­ing sys­tems and Kx is the high-per­for­mance engine that keeps a whole raft of oth­er clever soft­ware prod­ucts on the road. The veloc­i­ty at which Kx works reflects the fact that traders need to react to events inside sec­onds.

The reach of big data work extends from the gid­dy activ­i­ty on Wall Street to high street banks in the UK. Har­vey Lewis, research direc­tor at busi­ness advis­ers Deloitte, explains that finan­cial insti­tu­tions have an appetite for analysing cus­tomers how­ev­er large or small they may be. “The point about big data is that you can use it to under­stand your cus­tomer, and tar­get them based on what prod­ucts and ser­vices they real­ly want,” he says.

The tra­di­tion­al bricks-and-mor­tar bank branch can be just as impor­tant to sophis­ti­cat­ed data analy­sis exer­cis­es as any amount of online inter­ac­tion, accord­ing to Mr Lewis. “Phys­i­cal and online pres­ences feed off each oth­er. The branch can dri­ve more dig­i­tal busi­ness,” he says.

Com­bin­ing data gath­ered on the inter­net with details of cus­tomer trans­ac­tions in the bank branch is a ris­ing trend. For finan­cial ser­vices, there is no end to the data that can be prof­itably analysed.