At the center of a successful customer relation, Predictive Objects provides intelligence on every phase of the customer lifecycle, from acquisition to upsell opportunities to churn prevention
Scroll to discover
In contract-driven organizations, customer acquisition is expensive and every lost customer directly impacts the bottom line. The vast majority of unhappy customers will never tell you about their issues - until they cancel their contract. But you can know.
Leverage all the data at your disposal, internal or external to the organization, to detect weak signals that would be otherwise missed. Use Brute-Force algorithm selection to automate the search for the best combination of algorithms and data sets.
Don't just identify which clients are the most likely to cancel/leave: obtain actionable prescriptions to preserve the relationship. Written in natural language, these actions are sent to the account manager or contact center representative and are directly usable.
// Call Center Agent
Adam Smith, contract AB5678XY, has recently checked the international roaming options on our site but did not sign up for any add-on. Last year in August this customer incurred heavy roaming charges in Norway. With summer vacation approaching, it is likely that this customer is planning another trip abroad and looking for another mobile contract. Call to offer the Europe-wide roaming option at no cost for 2 months, as a thank-you for his loyalty.
A new job, a first child, a home purchase, retirement, etc. trigger significant changes. Detecting and predicting these changes gives you a unique opportunity to adapt your relationship with your customers.
Use all the data at your disposal to identify what is the best offer to present to a customer - including exogenous data which can be as varied as unemployment rates, school ratings or weather patterns.
Identify opportunities and provide recommendations to the front-line operator and/or generate automated marketing actions that will promote a new service, offer a contract change, etc. Written in natural language, these actions are directly usable.
// Branch Manager
William and Mary Gomes got married 2 years ago and opened a joint checking account, number 135246-90. Credit card charges show them shopping for pregnant women fashion and they are browsing for baby items online. With a baby on the way, their one-bedroom condo (paid in full) will be too small. Send them a few simulations of mortgages they can afford (based on their paychecks, maximum $1200/month) and our local realtor partner will call to suggest some properties within matching price ranges.
Most effective fraud techniques use a myriad of small, disparate transactions, that are difficult to match for the human brain. But because these transactions are computer-generated, there is a pattern, one that takes all the power of another computer to identify.
Use all the data that’s available - the transactions of course, but also all the knowledge about the customers, their usual behavior, as well as the patterns of recently identified frauds of recently identified frauds. This enables it to detect weak signals that would be otherwise missed.
Identify risks and provide a recommendation to the front-line operator about the best course of action: preventively lock a credit card, contact the customer to ascertain the problem, file a complaint with authorities, etc.
// Fraud Hotline Operator
Mike Debarry’s Visa card ending in XXX-9002 was used 3 times yesterday to buy gas near his home address. It was also used for one online purchase of $1.45 from a webstore registered to a Venezuelan address, that posted last week 23,000 transactions between $1 and $2 to credit cards we issued but never more than once per card. Email this customer asking to recognize these 4 charges. If no response is received in 24 hours call them. If unable to get through, or if the customer does not recognize the charges, block the card and issue a new one.
Interested in our Predictive Objects features, set-up, model, etc?
Fill in the following form to request a meeting.