The explosion of content prompted by the dawn of the digital age has given rise to new issues for technology companies. A major challenge for businesses is calculating the impact of past and current performance using the data they have acquired so as to define new business models and implement optimization strategies in accordance with areas needing improvement. Hotspot analysis (HSA) is an emerging technique being used in an increasing number of different analytical fields to address this very concern.

HSA analyzes data systematically to identify significant relationships and correlations amongst many variables accurately, as well as generates profiles of the areas or sectors that require prioritization. Other statistical and online analytical processing tools cannot match the analytical capability this method offers. Using AI&ML technology, this multidimensional analytics technique executes all of the following simultaneously:

• Segmenting collected data
• Creating profiles for said segments
• Analyzing data in accordance with requirements
• Automatically choosing variables as per relevancy and reported correlations
• Prioritizing hotspots based on severity and impact
• Identifying drivers for hotspots to provide actionable recommendations

A technique such as this has the potential to be used in many different business operations management settings. Examples include conducting investment analyses to assess company share performance and performing loan fraud investigations based on past customer data.

Our client, a multinational tech corporation, was looking for an agile, automated, and scalable solution to help identify hotspots based on customer behavior and data for improving customer support. The business operations managers (BOMs) of the client company found it a challenge to track multiple customer support metrics due to the following reasons:

• The large enterprise supports multiple commercial strategic business units (SBUs) spanning hundreds of products and services.
• All these products and services are delivered to over 70 countries, involving thousands of customers who speak different languages and have differing requirements for each product/service.
• The client’s customer support is handled by a combination of full-time equivalent (FTE) as well as affiliated and third-party vendors.
• Additionally, billions of combinations are possible for all tracked customer performance metrics, such as customer satisfaction score (CSAT), wait time, and time to resolve (TTR).
• Further, performance can be defined and measured using multiple metrics that vary across SBUs. Hence, there was no consensus on which areas of concern are to be focused on.

Therefore, analyzing related data and gleaning actionable insights that could be used to bring improvement to customer support is no small feat. The client wished to know which areas needed attention on priority and requested an automated tool that would increase visibility into performance metrics and customer support health.

From standardizing the definition of a hotspot to developing strong data visualization capabilities for the software, at every stage of the project, the end user was taken into account in order to guarantee that the design and display are simple, clear, and intuitive. This approach meant that TheMathCompany accomplished what the giant corporation has been attempting to do (internally as well as with several different vendors) for years, resulting in highly positive feedback from all levels of the organization. The scope of the solution was also iteratively expanded to cover more SBUs, products, and KPIs, in line with the client’s requests.

“In my last 20 years at this company, I have never seen a solution this visually impactful—especially one where I can just copy–paste simplified data insights on to a slide and share it with colleagues. We have tried to accomplish this a few times in the past but have failed. I am excited to see the outcome of this solution after complete development!”

– Worldwide Support Leader, Client Organization, after pilot project launch

Figure 1: Stages of the solutioning journey

The first step in the solutioning journey involved reaching a consensus with the BOMs on what qualifies as a hotspot. This was accomplished by conducting multiple working sessions with these managers. These sessions helped understand the nuances of tracking different KPIs and the varied levels at which the different stakeholders viewed hotspots.

As the next step, TheMathCompany’s analytics and customer support teams examined the client’s requirements for automation, scalability, and customizability; identified all areas of concern; and defined KPI benchmarks based on historical data; to ensure that the hotspot analytics solution is tailor-made to address all issues of the client corporation at a granular level.

The next step in the solutioning journey was the experimentation stage, with the aim of selecting an ML model that would identify the factors that cause drive hotspot formation, or "hotspot drivers." TheMathCompany’s expert programming and analytics teams experimented with multiple ML models, both linear and non-linear—an exercise that required the teams to go through thousands of iterations. Through multiple regressions and analyses to understand hotspot drivers, the team also built multiple driver models, altered based on interpretability and fit. This was done to simplify the analysis outputs so that it would be easy for the users of the solution to consume the insights. It took our teams over 10 iterations to reach the finalized version. Upon completion of the ML models, the entire solution was migrated to the cloud seamlessly.

Having a dedicated internal consumption team was integral to driving adoption and easing consumption among the client company’s users. As part of the final stage, this consumption team iterated through more than 10 versions of the UI/UX design for the screen. This was done to ensure that the analytics insights were displayed in a simplified, clear, and attractive manner to assist BOMs and business leaders in their decision-making. The data visualization software helped not only to make powerful visuals for the end user but also improve data discoverability and visibility, thus enhancing data processing capabilities. Throughout the feedback process, TheMathCompany’s teams built strong, organic relationships with key stakeholders in business, analytics, and provisioning teams, which ensured that all stakeholders trusted our processes and the final result.

Figure 2: Hotspot summary view

Tool implementation and evangelization were phased, giving enough time for optimal adoption of the tool by the client company and its employees. Indeed, as with any new technology, analytics presents potential challenges — not only with the implementation and management of analytics platforms but also with security, governance, and even organizational culture. The deployment and management of analytics platforms can be difficult, as they can with any new technology. Any and all challenges that arose were dealt with in a suitable manner, as described below, to ensure resolution was complete.

1. As mentioned before, there was no predetermined, concrete definition of a hotspot because of the scale of the organization and the complexity of the problem at hand. As a result, we attempted to standardize the definition of hotspots across all products, SBUs, providers, and regions. This was accomplished by holding working sessions, where the solution approach was iteratively altered in light of historical data on customer behaviors as well as the perspectives and degrees of understanding of stakeholders.
2. Potential users were hesitant to start using a solution due to the complexity and magnitude of impact of the solution, especially those without a background in analytics. TheMathCompany outlined a 4-step process to address this issue rather than simply developing the solution and letting the client company and its management handle the rest:
   • We arranged several workshops, training sessions, and open office hours to go over the solution in detail and answering any concerns or questions that employees of the company had about adopting it.
   • We gamified the launch of the software using tools like Kahoot quizzes. This ensured that the learning process was engaging, and thus more efficient, helping the employees better acquire information about the new tool, retain the knowledge, and strengthen existing skills that would improve tool usage in the future.
   • We also offered simulation training to the staff of the client organization by laying out different hypothetical scenarios that allowed them to practice tasks that mimicked actual situations they were likely to encounter when utilizing the product. This was done to ensure that knowledge about the tool was retained and would be used effectively, even under pressure.
   • Additionally, we created detailed weekly solution usage analytics reports and examined them to uncover any UI problems, prioritize feature enhancements, and assess user adoption.
3. Our analysts and programmers solved all technology-based issues, such as those relating to data availability, collection, and sanitization—all of which are inevitable when developing such complex solutions—by using appropriate coding, selecting the best software and technology, and fully automating the solution, end to end.