PredictiveEverywhere

Please post your replies here for Q #3

In a word: Balance. Key is to earn the trust of customers without relying on luck.

- transparency is a must in this space IMHO.

Stopping fraud is often in the customer's best interest and right communication to the right customer can help reinforce

@dez_blanchfield I agree that transparency is key. However, I wonder if the masses share that thought. I think not.

AND -- predictive analytics can help with that as well -- right message, right person, right time

- trying to bury an incident usually leaves the doors wide open for repeat offenders.

@jane_hendricks - making it as pain free as possible to actually report and act on an incident is critical, if it's too hard, few will waste their time reporting incidents.

A system that reduces fraud is part of the CX. It's about trust & confidence in the organization. @IBMbigdata

Fraud has costs to customers as well as an organization. Fraud prevention protects customers, and improve their experience

By using data-driven analytical models such as descriptive, predictive or social network analytics in a complimentary way plus keeping trust is the key

@tcrawford - I'm finding that transparency is a generational thing, older generations will bury and hide fraud, but younger generations are quicker to "out" an incident and deal with it..

involving customers in the approach taking and anticipated benefits, drawing on your social audience really supports transparency and engagement. Don't bury it.

@dez_blanchfield It MUST be transparent, again a watched pot never boils.

I've been impressed by companies that are proactive contacting the customer when they suspect fraud

@jane_hendricks @craigbrownphd : Your thoughts are welcome here..

@craigbrownphd

@DrZeydy AND, because someone does, you likely expect it from everyone now. A #newnormal.

Strong multifactor authentication on all transactions, for a starter (e.g, password + SMS + challenge questions). Consumers are used to it now. Not a huge crimp in customer experience.

@jane_hendricks - don't assume fraud is always a customer end thing of course ;-)

@dez_blanchfield Sure, unless that's the mode of distraction from their real fraud. Layers of sophistication.

@jane_hendricks - ah but predictive algo's are still such a rapidly developing space, who's models do you choose ;-)

.@CSurdak Absolutely! There's nothing worse than being "stranded" with a denied CC transaction w/o knowing what's going on

This chat is getting better and better.. and now it is time for Q#4 at the top of your screen.

@tcrawford I think most management disagrees. Denial and hiding seem to be the rule of thumb. I disagree with that.

Machine learning that identifies typical customer transactional patterns, flags anomalous possible fraudulent activity, and triggers customer service rep reachback for customer confirmation. Customers appreciate the safeguard

@jane_hendricks - but with that in mind we've seen the US banking system produce the GFC and with very little internal repocussion - nobody went to jail !!

@CSurdak Oh I agree with you.

Investment in processes and technology that provide the necessary protection. Confidence in the solutions utilized and a successful track record for customer review. Transparency to customers also helps with reputation.

Please post your replies to Q #4 here..

Easy, assume that it isn't and that you'll need to change/update within 6 month.

the threats are not only not static, they're accelerating!

Think of your solution as a product that must always evolve, is never complete and must always stay one step ahead.

Probably no magic timeline for updates. Automation, notification and alerts should be part of the strategy put in place

- collaboration, sharing models, publishing findings, validating and peer review of platforms, and moving at least as fast as the bad agents.

Like the identity theft services that tell you you've been hacked after the fact... too late then.

@CSurdak - there's a risk in cost and fear assuming you're always broken though.. but it's a KPI to map at some scale.

@CSurdak - while the incentives are high, the interest is going to always remain high..

@dez_blanchfield Invest at least as much in blackhats as you do for whitehats. Hack yourself on a regular basis.

It probably also depends on your specific fraud use case. Some may need real-time monitoring some may have a longer tail. #machinelearningneedspeopletoo

Going back to what @dez_blanchfield was saying at the beginning, need a good Strategy from the start

@jane_hendricks And the longer the tail, the worse it gets. Repeat offenders always escalate.

@CSurdak - if you draw a risk model a a pyramid you find that you have to invert it to measure high risk against low value / high volume.

@IBMbig Depending upon the type of model and its purpose (e.g. descriptive or predictive), various backtesting activities can be undertaken to keep solutions up to date like data stability, model stability and model calibration.data

Your data scientists need to explore threat/fraud behavioral patterns on your production systems, feed those updated threat/fraud signatures, plus rules, into your internal remediation and external customer engagement workflows.

@CSurdak on the plus side, repeat offenders often use a similar and catch-able pattern. It's the new patterns I worry about...

- #fraud is never going to be a "solved issue", any more than spam or data breaches are, so you have to establish a high water mark you can maintain and manage.

@jane_hendricks YES! Don't get comfortable with your own approaches. Throw a monkey wrench in every once in a while. That's what the offense is doing to you.

@jane_hendricks - Jane we can't keep every day things like OS X and Windows "updates" working sensibly, multiple the risk model to apply #fraud and the challenge is frighteningly larger ;-)

@jane_hendricks - indeed, it's not the script kiddies you need to worry about, and more so it's probably not the rogue operators any more either, but soon we'll face rogue nations in this space !

On that note, let us jump to question #5 at the top of your screen.

@dez_blanchfield The best hacks are on 15 year old tech that most people have forgotten. Look at threats to the cell phone stack, for instance.

machine learning will be a key part of having fresh technology in the mix

Use analytics to find out where controls are not working or are ineffective and don’t forget to look for controls that cannot be governed by application control settings. #predictiveeverywhere

Proper use of #data #analytics and #predictive #analytics with a teaspoon of good data science and some luck wouldn't hurt.