Use AI to enhance customer experience and drive digital sales

The Big Picture

For most legacy brick-and-clicks that sell mid to high complexity products, conversions on digital continue to remain sluggish, and businesses face a high degree of customer drop-off, cart abandonment, as well as confused and irrelevant customer conversations. The impediments to digital sales can be very unique for each visitor, such as insufficient product information, in-store pick-up issues, improper search results, price shock, poor product recommendations, and more.

Since most large businesses cannot recreate the digital experience from scratch, they need to adopt a rapid mode of continuous improvement. That means identifying what went wrong in the sales experience, what were the causes, and how the various improvement options can be tested.

Transformative Solution

For a leading retailer, Fractal deployed an advanced AI-based framework to create features from every digital interaction down to minute click-events and identified high-impact causative issues that negatively impacted the customer sales journey.

The Change

As a result of the engagement, the retailer gained more than a 25% increase in digital sales, by removing the impediments in the customer purchase journey.

Match individuals across data sources without a unique key

The Big Picture

A leading credit bureau provided analytics and intelligence support to local credit rating agencies. The company had access to multiple data sources, such as bank data, voter IDs, and tax returns, from where it pulled information for individuals and created their credit score.

However, it was not possible to use information from all these data sources as there was no single unique key. For example, in a bank data set, an individual would have a driver’s license number, but in a voter ID data set, the same person would be identified with a voter ID number. The company couldn’t know that the individual was the same in both data sets. Therefore, there was a need to create a logic to match these different data sets without having a unique key.

This meant that the company needed an improved framework of working with unstructured addresses data so that it could provide a single view of customers across different data sources. It needed to drive measurable performance improvements by improving match accuracy and reducing false positives.

Transformative Solution

To solve the company’s challenges, a solution was deployed to match datasets using names and addresses, since this information was present in all data sources. Since the format of names and addresses were different everywhere, the solution needed to create intelligent and fuzzy logics to standardize names and addresses for mapping purposes.

The approach took raw data and deployed a name and address matching algorithm that was configurable at different levels. The solution incorporated a search capability along with optimization and improvement of matching. Three key steps were:

  • Data standardization: Data was cleaned and normalized to remove components not adding value to addresses. Addresses were segregated into logical components: house number, locality information, and pin code.
  • Address search: The approach searched the request address into candidate data using pin code (and derivatives) as a key.
  • Name and address matching: This step used Fractal’s dCrypt to match all the addresses in a key value pair with request address and selected top 100. For top 100 addresses, the corresponding names were also matched and the best output was generated on the basis of name and address matching scores.

The final output provided a list of names and addresses from the candidate data which match the name and addresses from the reference data. Using the algorithm, a matching score was generated between two strings which could be compared with a base matching score already present in the client’s sample file. Randomly selected samples were manually checked and a confusion matrix was created for both algorithms.

The Change

As a result of the engagement, the company achieved several benefits:

  • Improvement in accuracy by 10% on 11 million household addresses.
  • Incorporation of a search capability in the matching algorithm.
  • Three different algorithms were used for matching as opposed to a single algorithm for name and address matching, which led to better coverage and efficiency.
Reduce high costs of care associated with avoidable ER visits

The Big Picture

The high cost of maintenance and limited availability of Emergency Rooms (ER) facilities are under intense scrutiny by payers, the government, providers and employers. According to the Centers for Disease Control and Prevention (CDC), Americans made 136 million ER visits in 2014, which is likely to increase further. Yet a study in the American Journal of Managed Care cites more than 30% of ER visits could have been avoided.

Avoidable ER visits stem from a lack of coordinated medical attention that drives higher costs of care, longer wait times and sub-standard health outcomes. Redirecting only 20% of ER visits to lower-cost alternatives, such as urgent care or Primary Care Physicians (PCP), could save $4.4 billion, according to HealthAffairs.org.

A multi-billion dollar healthcare payer wanted to identify members likely to make avoidable ER visits, and steer them to more cost effective alternatives.

Transformative Solution

Members may be visiting an ER unnecessarily for convenience, desire for a more effective PCP, insufficient co-pay funds, or an unmanaged condition. To address these challenges, clinical rules were used to identify low intensity conditions where an ER visit could have been avoided. The approach offered more than 50 hypotheses for factors which could be predictive of avoidable ER visits.

To test these hypotheses, we identified different structured and unstructured data sources such as call center notes, geographic details for members and providers, and the availability of providers.

For unstructured data, we applied multiple feature selection algorithms such as InfoGain1 and BNS2. For structured data, we tested hypotheses such as distance of the Primary Care Physician or urgent care facilities, ease of access to an ER, and difficulty finding quality providers. An ensemble of classifier models was developed to predict the likelihood of visiting an ER for low intensity conditions, using advanced analytics such as machine-learning, text mining, and traditional modeling techniques.

The solution identified 65% of all avoidable visits among 30% of the population. This yielded an opportunity to save more than $10M annually by targeting a small group of members for alternative care management and provider interventions.

The Change

The payer was able to gather from this project that

  • Members with past ER visits were 8 times more likely to visit the ER unnecessarily.
  • Members visiting multiple PCPs were twice as likely to make an avoidable ER visit.
  • Each avoided ER visit could reduce costs by $1,500, leading to $10M in potential cost savings.
  • Optimized ER utilization could substantially improve member health outcomes.
  • Creating a framework of text-mining and machine-learning methods could improve accuracy in rare event scenarios.
Deliver next best product recommendations during customer interactions

The Big Picture:

A leading retail bank was facing low customer engagement and satisfaction with its customers. The existing analytical models on product propensities generated lower accuracy, and missed critical data elements, such as offline and online interactions and transactions, and prevented an objective arbitration of offers among multiple competing product offers. This resulted in sub-optimal customer experience and lower response rates.

Transformative Solution:

To address the company’s challenges, a new next best product and service recommender was built using deep learning. It was designed to predict the top three recommendations from among a wide suite of products, and for services.

A single customer view was prepared with 4,000+ attributes such as customer product holdings, transactions, in-bank transactions, and online interactions.

The models were tested on a select population within the lead scoring platform and deployed centrally.

The Change:

As a result of the engagement, customer product-offtake rates jumped by 60%, resulting in significantly higher marketing ROI.