Okay -- so I've been harping on the pattern of executive-selling Mattersight's common stock, without the company ever having generated any GAAP Earnings Per Share -- for 13 years. As it is wont to do, when in such straights, Mattersight today put out a pufferey press release. Mattersight proposes to sell a phone-call/live-voice fraud analytics tool, using the same data-sets it uses for all customer interactions. While I might be inclined to agree that live voice attempts at impersonating people for the purpose of committing fraud are on the rise, I am not sure that the business world is ready to pay high margins for the rather ordinary insights Mr. Conway's data might offer. Especially suspect is the claim that he can match a caller against a database of known fraudsters, by "voiceprint" -- in any significantly close to real time way. It seems unlikely that his systems could do so in real time -- in other words, quickly enough to actually prevent a fraud, while it was underway.
I could be wrong, but I don't think even Langley (the CIA) or Quantico (the FBI) would make such a claim -- and if the CIA/FBI doesn't have the raw computing power to do so -- how can Mr. Conway claim to possess enough processing power to analyze, with instant accuracy, and meaningful insight -- all those tera-bytes of data, in real time?
Okay -- enough of my skepticism -- let's hear from Mr. Conway, shall we?
Yes, let's. From Mr. Conway's latest press release, then:
. . . .Voice-facilitated fraud is a rapidly growing problem in a number of industries. Financial Services companies and Property Casualty Insurers have realized their call centers are a vulnerable and lucrative point of attack for fraudsters. Sophisticated fraudsters and fraud rings are able to gather enough customer information from publicly available sources to exploit the authentication procedures in call centers and gain access to customer accounts. The complexity of capturing and analyzing call center interactions to identify fraudsters has heretofore made it nearly impossible for companies to measure the frequency of these types of attacks, let alone proactively identify this activity and take preventive action.
Mattersight's Fraud Analytics solution solves these problems by capturing customer interactions and automatically analyzing every second of every captured interaction in the cloud, using millions of proprietary algorithms and unique behavioral models. The output of this analysis is hundreds of contextually accurate data attributes on every captured interaction. . . . Mattersight also leverages a consortium database of known fraudster's voiceprints that every caller is compared against to identify repeat fraudster activity. . . .
This looks to me to be more of a product "by accident," than "by design." That is, it seems someone at Mattersight just realized that this was another (albeit non-real-time) use for all that stored voice data: create some metrics to try to identify fraudsters. Great. But that is only truly useful, if the filter(s) can be made to work in near real time, not some post-hoc analysis. But maybe I'm wrong, and the Fortune 500 will pay high prices for this analytical tool. I doubt it, but it could happen.
No comments:
Post a Comment