Har­ness­ing AI for Al­pha

Fidelity is now us­ing ar­ti­fi­cial in­tel­li­gence and data to im­prove in­vest­ment prod­ucts. Will the ef­fort boost re­turns?

Financial Planning - - Contents - BY SEAN ALLOCCA

Fidelity is now us­ing ar­ti­fi­cial in­tel­li­gence and data to im­prove in­vest­ment prod­ucts. Will the ef­fort boost re­turns?

Firms are look­ing to an­a­lyt­ics to cre­ate pre­cise in­vest­ment ve­hi­cles that in­crease their chances of gen­er­at­ing al­pha.

Fintech firms have long sought ways to build out the ad­vi­sor tool kit, but new tech­nolo­gies are also shap­ing the in­vest­ment prod­ucts ad­vi­sors ul­ti­mately rec­om­mend to their clients.

From nat­u­ral lan­guage pro­cess­ing sys­tems that un­cover stock mar­ket trends to an­a­lyz­ing decades of in­vest­ment de­ci­sions to root out in­vestor bias, firms are look­ing to an­a­lyt­ics to cre­ate pre­cise in­vest­ment ve­hi­cles that in­crease their chances of gen­er­at­ing al­pha.

Data to Al­pha

“There is def­i­nitely move­ment to­ward, and real value, in be­ing able to have sci­en­tists look at the data sources,” says Vi­jay Ragha­van, se­nior an­a­lyst at For­rester. “To find sig­nals and draw in­sights to make bet­ter rec­om­men­da­tions and guide in­vest­ment de­ci­sions — that’s the di­rec­tion ev­ery­thing is mov­ing to­ward in fintech.”

The data re­search ul­ti­mately helps quan­ti­ta­tive an­a­lysts un­der­stand macro trends and mar­ket con­di­tions, in­for­ma­tion that is then used to cre­ate the next gen­er­a­tion of mu­tual funds and ETFS of­fered to their cus­tomers.

“There’s an un­be­liev­able amount of data in the walls of this build­ing,” John Avery, head of ar­ti­fi­cial in­tel­li­gence and ad­vanced an­a­lyt­ics at Fidelity In­vest­ments, told Fi­nan­cial Plan­ning at the Bos­ton­based firm’s of­fices in Septem­ber. His depart­ment main­tains a data­base of 400,000 re­search notes on in­vest­ment de­ci­sions dat­ing back to the 1970s, he says.

“That’s where we are look­ing to gen­er­ate value that no one else is go­ing to see,” Avery adds.

By work­ing with as­set al­lo­ca­tion groups, the Fidelity an­a­lyt­ics team is look­ing to cat­a­logue all the firm’s in­vest­ment de­ci­sions over time to pin­point in­vest­ment bi­ases, knowl­edge that could drive fu­ture al­go­rithms, he says.

For ex­am­ple, con­fir­ma­tion bias makes an­a­lysts sub­ject to dis­miss­ing ev­i­dence that is con­trary to their be­liefs, says Darby Niel­son, man­ag­ing di­rec­tor of re­search at Fidelity. The re­search gives

quan­ti­ta­tive an­a­lysts an in­de­pen­dent view on stock se­lec­tion and port­fo­lio con­struc­tion.

“We’re able to go back and study the good de­ci­sions from the bad de­ci­sions,” Avery says. “And find out what par­tic­u­lar bi­ases they’ve had in or­der to push mod­els for­ward.”

In ad­di­tion to the cat­a­logu­ing project, Fidelity an­a­lyzes all phone con­ver­sa­tions on their cus­tomer hot­line to bet­ter un­der­stand client needs, he says. “The things that we can find out about the in­for­ma­tion in that data — it’s mind­bog­gling,” Avery says. “That’s where we’ve been spend­ing a lot of our time.”

Fidelity is cur­rently work­ing on a back­log of 80 projects, many of them fintech star­tups, that look to boost an­a­lyst ef­fec­tive­ness, Avery says. “We haven’t found the sil­ver bul­let yet,” Avery says, adding that a lot of an­a­lysts are us­ing the same data sets as their com­peti­tors. “But by im­prov­ing the pro­duc­tiv­ity and ef­fi­ciency of the re­ally smart an­a­lysts, then by def­i­ni­tion, you in­crease al­pha.”

Fintech In­vestors

Fidelity’s work is all part of $8.8 bil­lion fund­ing ecosys­tem made into U.S. fintech firms in the sec­ond quar­ter, up from $2.8 bil­lion from the year-ago pe­riod, ac­cord­ing to this year’s KPMG’S Pulse of Fintech study. Ven­ture cap­i­tal in­vest­ments soared to over $3 bil­lion.

A smaller fintech is look­ing to nat­u­ral lan­guage pro­cess­ing to un­cover al­pha. The St. Louis-based in­vest­ment re­search firm Prat­tle scans hun­dreds of thou­sands of pieces of com­mu­ni­ca­tion — like earn­ings calls from 3,000 com­pa­nies and 15 cen­tral banks — and then of­fers the pro­pri­etary data sets to fund man­agers look­ing for a pulse on the mar­kets.

“We’re built to in­gest in­for­ma­tion fast,” says Prat­tle CEO Evan Sch­nid­man.

The startup at­tracted $3.3 mil­lion in fund­ing last year. On the bro­ker­age side, quan­ti­ta­tive funds have been early adopters, he says.

TD Amer­i­trade is also us­ing nat­u­ral lan­guage pro­cess­ing to op­ti­mize ef­fi­ciency. The firm re­cently launched a Twit­ter chat bot for han­dling queries,

“We haven’t found the sil­ver bul­let yet,” says John Avery, head of AI at Fidelity In­vest­ments.

which can help cus­tomers find trad­ing in­for­ma­tion quickly and get ac­cess to the lat­est in­vest­ment prod­ucts. Clients can send a di­rect mes­sage to @Tdamer­i­trade to be con­nected with the firm’s AI tech­nol­ogy.

“That’s where things are head­ing,” Ragha­van says. “A re­li­able ap­pli­ca­tion where clients can punch in some ba­sic queries with­out hav­ing to pick up the phone and call a fi­nan­cial ad­vi­sor — and don’t even have to go be­yond their Face­book app.”

Will An­a­lysts For­get the Process?

While AI is open­ing up new doors in as­set man­age­ment, some ex­perts worry the tech­nol­ogy may pre­vent an­a­lysts from fully un­der­stand­ing the pro­cesses go­ing on in­side ma­chine learn­ing tools. In turn, this could limit ad­vi­sor ac­cess to some new prod­ucts.

For ex­am­ple, it may be dif­fi­cult to dis­cern if big data al­go­rithms are re­li­ably an­a­lyz­ing data sets to find the best out­comes.

“Ma­chine learn­ing is only a black box if you don’t want to look in­side,” Sch­nid­man says.

Data ag­gre­ga­tion presents another chal­lenge, Ragha­van says. Wran­gling in­for­ma­tion across com­pany si­los is a siz­able and costly hur­dle. “The biggest chal­lenge es­pe­cially for legacy in­sti­tu­tions is get­ting that data nor­mal­ized and in a spot where an­a­lysts can ac­tu­ally draw in­sights,” Ragha­van says.

By dove­tail­ing the out­side data sets with in­ter­nal data, Avery and oth­ers hope to shape the next gen­er­a­tion of in­vest­ment prod­ucts.

“The say­ing goes, ‘He who turns over the most rocks wins,’” Avery says. “What we’re do­ing is mov­ing more rocks faster — and find­ing rocks with more op­por­tu­nity.”

The Fidelity AI team keeps a data­base of 400,000 re­search notes and in­vest­ment de­ci­sions dat­ing back to the 1970s.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.