CLIENT STORIES

Optimizing decision supply chain of the repair business for a leading consumer electronics giant

We work with one of the most recognizable brands in the world: A leader in consumer devices, smart phones, and digital entertainment and services.

Here’s a story of our partnership.

Our client faced some difficulties in an important, but often overlooked, facet of the consumer experience: Product repairs.

For some reason, the high volatility in the time it took to repair certain products impacted the accuracy of the repair forecasts that they shared with customers – it was low. This was a big problem, not only because of the impact on the customer experience, but because repair volumes can have major accounting implications. So we were approached by their customer care group to understand the variations in repair times and forecasts.

The Prototype

The customer care group works on many problems – all related to customer experience.

But as we do with every client relationship, we began with a pilot – this one focused on repair forecast variations in just two product segments.

The client already had forecasting models in place. But we found them to be heuristic and manual, making the forecast process both time consuming and hard to scale. Using our platform and processes, we quickly realized that manufacturing batch was used as a variable in the forecasts, when in actuality, the products sold in the market would be a much better cohort for forecasting expected repairs.

In just 6 weeks, we developed a more algorithmic and automated model for repair forecasting, requiring very little manual intervention.

The models were not only more accurate, but could also quickly scale. In fact, towards the end of the pilot, our repair forecast model was scaled to all product lines and products (60 in total) and allowed our client to drill deeper into repair times at different levels – such as region and country.

The early ROI was clear. Our client had a more accurate, automated, and robust forecast model – one that they could scale to multiple products and reduce the turnaround time from one month to just one day in terms of generating forecasts for all products.

How did our problem solving work evolve?

We soon found that the customer care group was not the consumer of our analysis, but rather it was the finance function, who needed the forecasts to set aside financial reserves.

In partnership with the customer care group, we worked to align our models more precisely with the finance function’s needs. In the process, we uncovered the unstated problem of improving accuracy. We built a scalable and repeatable framework which not only generates forecasts on a weekly basis, but also explains the root cause of variation in forecast at different levels.

This drove more alignment across the multiple stakeholders who were involved in the decision supply chain for the repairs business.

But our problem solving journey didn’t stop there. As new products were being launched, we found it difficult to forecast repairs when there was such a limited history of performance.

This is when we decided that our models needed to be adjusted to support a more variable and dynamic business. We updated the muPDNA^TM and had discussions with the client on additional hypothesis. This process made us and the client learn together, in the course helping us revamp the forecasting design as per current needs.

What did the client appreciate about our approach?

In a relatively short period of time (of 4 months) our client saw clear benefits from our relationship.

But the benefit was not just in terms of revenue impact or analytics turnaround time, but we were staging an experience for the business by exposing them to a new way of approaching problem solving.

Here are few key points appreciated by our client:

One. Using our platform and muPDNA^TM approach, we took a design-first approach to problem solving and helped break through a lot of previously held assertions – like manufacturing batch being incorrectly used as a model variable.

Two. Cross-industry learnings. We leveraged the Mu Sigma ecosystem, and applied learnings from claims forecasting in the insurance industry, and reseller businesses in retail, to ideate and innovate in our model designs.

Three. Extreme experimentation – at a low cost. The solution framework that we developed had a 1-day turnaround to deploy a new model, as opposed to one month. The quick turnaround gave us the time to learn from more than 25 different techniques.

Four. A focus on analytics consumption. Our focus on consumption (and not only creation) helped us to better understand the end customer – in this case, the finance function – and better align our solutions to their needs.

Five. A longer view, an analytical Roadmap. Using our muOBI artifact, we nudged the client to plan around the business and behavioral outcomes they seek in the future. This helped us develop an annual roadmap, including 7 new problem areas around customer satisfaction, supply chain and business process management.

Building an automated store planning tool for product placement for one of the largest retailers using a design first approach.

We work with one of the largest retailers in the world: A multinational corporation that operates a chain of discount department and warehouse stores.

The client was facing performance issues with their super centers, across US. Declining sales apart, the shopping experience were also a concern.

Some of the causes for these issues were planogram design, poor inventory management, promotion planning etc.

The Store Layout Planning Team approached us to improve the performance of super centers across US. The other side of the problem was scaling analytics across the organization.

The Prototype phase

The client already had a manual store planning framework in place which relied on conventional theories.

But, the framework could not scale across all its stores in the US.

We took a design first approach and leveraged our muPDNA^TM software to represent the problem. Sales affinity factor between departments was considered in the product placement. The solution design had to be one that could scale as well as capture sales variation at multiple levels.

Multiple frameworks such as genetic algorithms and simulated annealing were explored using experimental approaches.

The final solution took the shape of linear optimization. This framework had maximum scalability and low computational cost. In just 6 months, we provided the client with a platform for store planning that could be implemented across any store format, resulting in 18% cost savings for the client.

How did we evolve?

Post the successful execution of store planning framework project, we worked with multiple subgroups in the organization ranging from IT & Compliance, to Global customer insights and analytics.

We evolved from a 3rd party vendor to a recognized strategic partner supporting numerous business functions.

The client saw value in our vision of harmonizing the solving of specific problem (SSP) with adopting a new art of problem solving (AoPS). We leveraged various aspects of our ecosystem to drive impactful analytics for the client.

Taking a broader perspective

At the onset of our engagement, we charted out the roadmap along with the relevant stakeholders from the client end using muOBI. We clearly outlined the Outcomes to be driven and the behavioral changes (shaped by insights) required to drive these outcomes.

We then identified their broad problems as well as the interactions to multiple smaller business problems using muUniverse^TM. For instance, we showed them how a bad shopper experience could be linked with issues of store operations, demand planning, inventory management, assortment, promotional campaign etc.

This helped the client prioritize problems within the subgroup, allowing shorter decision cycle times. We further built on the problem space with new connections as and when we identified them.

What Resulted?

One major factor that the client appreciated in our style of working was “a culture of low cost experimentation”.

Every problem that the client posed to us, we approached it from umpteen directions, and then kept narrowing it down based on the outcome. This resulted in faster turnaround and a 3 percent increase in overall profitability.

Collaborative problem solving journey with one of the largest drug retailers

This success story is about our engagement with a leading drug retailer in US.

The organization was undergoing a merger and was required to keep the employee costs low.

Clearly, the business had a specific problem in mind – payroll cost cutting or rather lean staffing.

This was a complicated situation; while financial goals of the company had to be met, employee sentiments during the merger was also to be taken into account.

The initial phase

The problem of payroll cost cutting required thorough thinking. We leveraged the muPDNA^TM software enabled by a rigorous muSearch to define the problem

This ensured that we understood the problem as much as the stakeholder did.

Our recommendation was two-fold: better ROI of the existing staff and optimal staffing number. We closely analyzed the greeter-guard arrangement in over 50 stores to identify unmanned locations within the store. This data was analyzed for its effect on shrinkage cost and identifying locations for greeters where the impact was minimal in terms of cost.

This resulted in 18% reduction in the shrinkage cost.

We also tracked criminal rates across store locations and suggested optimum number of guards required. As a result, guards were cut down from the stores where the criminal rate was comparatively low. This ensured that the client has a lean staffing model that could keep the costs low.

Scaling to other avenues

We didn’t stop only at solving the problem at hand. We wanted to leave a bit of Mu Sigma in our client and convert this into a long term engagement.

We leveraged muUniverse^TM software to identify interconnected problems within the store operations problem space.

For efficient store operations, we proposed the idea of a centralized customer care team. The team was to be responsible for customer satisfaction as well as store performance. The call centers were now equipped to handle customer queries as well as enable relationship building with them.

Out of 8000 stores, we started with 1000 stores; eventually implementing the project across the remaining in a matter of 4 months.

Call centers were then exhaustively used for inbound marketing campaigns. This helped us extend our partnership with the marketing team as well – helping the client gauge campaign effectiveness.

Other than campaign analytics, text mining was also performed on the call center transcripts to gauge and analyze customer feedback. This led to a 5% increase in customer retention.

We then analyzed the IVR system for customer behavior through inbound calls and defined a suitable flow.

Over a period of time, our engagement was extended to specialized services as well within stores. We supported their sales force with goal setting, incentive compensation and other aspects of sales force effectiveness.

Our focus on analytics consumption (and not only creation) helped us to better understand the end consumer and align our solutions to the business needs.

Benefits beyond quantifiable

The client was not only benefitted in terms of revenue and cost reduction, but by being exposed to the new art of problem solving, they were more agile in terms of proactively tackling business problems.

Here are few aspects the client largely appreciated us for:

• Our constant drive towards “Learning over knowing” provided the client with an exhaustive search repository to always bank upon and refer to. They are rigorously using our search material for study and research on their own.

• With the use of the Mu Sigma artefact, muDSC, they were able to focus on decision supply chain and drive analytics consumption across the organization for better and faster decision making.

Exploring interrelated problems for a Fortune 100 financial services organization

We work with a leading provider of retirement and financial services in the academic, medical, and research fields.

The client has a strategic imperative – to digitize operations.

Where did we start?

We were approached to solve one specific problem: Understanding the reasons behind the 45 percent participant drop-off rate experienced on the company’s online portal for pension enrollment.

This was a multi-faceted problem, negatively affecting customer experience and hence satisfaction. There was a 22 percent revenue loss in the business due to this drop-off. And the issue was not rooted in large volumes of data.

What did we do?

Using our proprietary muPDNA^TM software to encode the problem’s DNA early on, we helped uncover primary drivers of drop-offs like page errors, lack of clarity on the pension plans, and an unclear fund selection format.

We tapped into our rich ecosystem and connected with the digital analytics team of a leading e-retailer to learn optimal placement of banner ads, design of promotional content and page layout based on customer interests, preferences and requirements.

We carefully analyzed the entire pension enrollment process, and in less than 15 weeks, came up with valuable insights and recommendations for the client. We redesigned the entire pension enrollment process for the client, thereby increasing potential enrollments, capable of generating a projected revenue of $5 million in the next 5 years.

A wider view

With the strategic push to go digital, the analytics team felt the pressure to support many initiatives. They needed help in both prioritizing and understanding digital initiatives such as BI automation, digital customer behavior/ feedback, online campaign management.

Using muUniverse^TM, we were able showcase to the client the various areas we could help them apply analytics – moving beyond just digital to also encompassing marketing, operations, competitor analysis etc.

This focus on problem interactions led to new areas where the digital analytics group could orient the channel marketing as well as brand teams towards understanding the various performance drivers of their marketing campaigns.

What resulted?

We reduced online portal drop-offs from 45 percent to 20 percent. We were able to enhance the overall customer experience and satisfaction for new as well as already enrolled participants. The client implemented our recommendations leading to $3.6 million in additional revenue and 12,000 new enrolments through the year.

Further, an additional ad hoc project around website participant relationships led to a 5 percent increase in customer base for the client.

Evolving from a vendor to an analytics advisor for an insurance major

This is about our engagement with one of the largest insurers in the world: A leading publicly held personal lines property and casualty insurer in the US.

We were approached by the Claim Technology Services (CTS) group of the client.

This team was responsible for paying damages to home insurers, whose property was affected due to any natural calamity. They were facing difficulties with accuracy of claims classification, leading to longer claim settlement cycle, losses, customer dissatisfaction and churn..

The Inception

Typically, claim applications are classified into buckets to guarantee appropriate investigations before the claims are paid off.

We realized, on connecting with the CTS team, that the earlier classification of these claims were not very precise.

The client already had an algorithm in place for claims classification. But we noticed that the algorithm had shortcomings, feedback incorporation being a major one. The existing algorithm lacked the ability to re-learn and re-construct based on any previous misclassifications.

We used our muPDNA^TM software to define this problem and encompass all the factors affecting claims classification such as at-fault accidents, natural accidents, territory of accidents, age of the insured etc.

We used machine learning to read the nodes collected by the customer care executives and created a new algorithm. While doing so we kept in mind the different factors that help predict the likelihood and cost of insurance claims. The new algorithm was intelligent enough to learn from the past in an automated fashion. It was also capable of coming up with a model that can be applied to a new data, based on past instances.

The ROI was very clear at this stage – the average duration of the entire claim settlement process was brought down by 34 percent. The highly accurate claims classification resulted in increased customer satisfaction and anticipated savings of $15 million in a quarter.

How did we evolve?

We were able to build the trust of our clients, to the extent of being referred to other subgroups within the client ecosystem./p>

For the marketing subgroup, we looked into their campaign effectiveness. We measured the Net Promoter Score (NPS) and looked at the customer satisfaction levels etc.

We also partnered with the client’s Operations Strategy Group to check the efficiency of their claims processes. In our regular meetings with the client, we took part in their long term goal planning and used our muOBI framework to understand the desired outcome, behavior and insights and have a roadmap created. We were slowly traversing the path towards becoming an analytics partner/advisor and not just a vendor to the organization.

What did the client appreciate about our approach?

• Cross Industry Learning: While working with the client, we applied learnings from various banking, insurance and finance world. This was hugely appreciated by the client and it helped us engage with multiple subgroups within the organization.

• muOBI: Using our muOBI artifact, we nudged the client to plan around the business and behavioral outcomes they seek in the future. This helped us develop an annual roadmap to tackle existing as well as expected problems for the client.

Developing analytics roadmap for a leading financial services and insurance company

We have been closely working with a leading provider of general insurance, banking, life insurance and wealth management.

Business leaders across this organization are bought into our vision.

They wanted us to systematize analytics across their organizations. They wanted us to make their analytics initiatives more robust and scalable.

But as is always the case, we started in a focused manner. Our partnership began when, after attending a Decision Sciences workshop, the VP of Life Insurance Marketing approached us to solve a specific problem – of building an attrition model to predict which customers will churn from the insurance plans they are currently enrolled into.

The Prototype

We began with two artifacts that are part of the Art of Problem Solving System.

For one, we used muPDNA^TM software to formulate the right questions and come up with all possible factors affecting customer attrition.

Then we proactively mapped out all problems in the life insurance marketing space using our muUniverse^TM software and utilizing the problem ontology defined in muPDNA^TM.

This helped us uncover critical new interactions between attrition model and other problem areas like product pricing, product offer customization, promotion effectiveness and customer segmentation.

When we built the attrition model, the existing customer segmentation fed into it as an input. This input enabled the prediction of customers who are soon to attrite. Based on the recommendations from the attrition model, these customers were targeted with customized product offers and marketing campaigns.

Still further, we were able to show a bigger picture and purpose – bringing more data intelligence into the overall problem landscape.

A holistic view

Even though, our above approach resulted in impressive results, we revisited all the factors influencing attrition, especially the customer segmentation part. The reason being the incorporation of multiple heuristic untested hypothesis within the segmentation framework.

Conventional wisdom suggests that the elderly are not ideal candidates for a life insurance policy. But by factoring in the opportunity to sell additional products more relevant to their demographic, the picture changes – and so does the customer segmentation variable in the attrition model.

In the absence of a robust customer segmentation model, we brought in our cross industry learnings from other “at-risk” customer models in the retail and pharma sectors, and created a new segmentation based on multiple attributes such as customer value, demographics, location etc.

This exercise led to an increase in (marketing) campaign ROI by 14 percent over the previous year.

How did our problem solving work evolve?

That was just the beginning of our relationship with this client. We never lost sight of their primary goal of institutionalizing analytics. We leaned heavily on muOBI to chart out a roadmap for each problem space.

Doing this required that we begin with the goal/ objective in mind – associating each business problem or network of business problems with a set of desired outcomes, behaviors, insights (and findings, and data).

With this long term view, we were able to develop annual analytics roadmap for the client and maintain it as well.