Lessons Learned from Deploying Software Applications: Design Thinking and Portfolio Management

In today's fast-paced technological landscape, deploying software applications effectively requires a combination of thoughtful design principles and strategic decision-making. Over the years, I have encountered numerous challenges and gained valuable insights while deploying software applications. In this article, I plan to share two interesting lessons learned - one related to design thinking and another about portfolio management.

Design Thinking: Active Listening, Iteration, and Ego

When it comes to developing successful software applications, design thinking is crucial. It involves considering the end-user's needs, actively listening to their feedback, and iterating on the product to ensure it meets their expectations. Most importantly, design thinking requires us to put our egos aside and embrace the idea that the initial version and even the next few iterations of the product might not meet the end-users requirements or expectations.

Recently, I was involved in developing an anomaly detection AI feature for a software application that our company embedded into Slack. During the interaction with end-users, we actively listened to their pain points and challenges. This enabled us to truly understand their needs and design a solution that met their expectations effectively.

One of my extremely talented data science teams created a current state-of-the-art anomaly detection system using a machine learning trend detection algorithm, heavily influenced by current peer-reviewed journal entries and other recent developments in the field.? The technology was also oblique and not obvious to human perceptions. It had a magical feel and was very accurate. From an AI/ML design perspective, the application was literally art, like a cherished Picasso or a Chagall piece. Perhaps it was too modernly abstract, just like those two renowned artists’ works. Our primary end-user was expecting something more in line with a financial forecast instrument with a quality control warning boundary. Something that grabbed the other end users attention and made the anomaly obvious and easy to understand.

It was tough feedback considering how the team had designed a state-of-the-art algorithm worthy of peer reviewed publishing. Instead of assuming that we had all the answers, we approached the development process with an iterative mindset. We went back to the basics of control theory and cobbled it to financial forecasting. This wasn’t easy after delivering the current state-of-the-art solution in our mind. We designed a few more prototypes, gathered feedback, and adjusted in subsequent iterations. This iterative approach significantly improved the final product and enhanced user satisfaction. Even though tested and fielded, we continue to iterate improvements based on feedback and new knowledge discovery.

The most crucial aspect of the iteration was keeping our egos in check and out of the way. Our initial design was complex and accurate. Perhaps from a selfish point of view, we could have scaled the technology on different, public data to publish in a peer reviewed journal, like the many IEEE offers.

It's easy to get attached to our own design choices, but embracing others' perspectives and suggestions can lead to a much better result. The final technology combined forecasting and control theory. It was more simple, more transparent, and not sexy. We parked our initial design (our Porsche) at our ego’s garage and designed and adopted a more transparent and simpler model (our VW Beatle) for the users.

By prioritizing active listening, iteration, and letting go of ego, we were able to create an anomaly detection AI feature that surpassed our initial objectives and delighted our users. Plus, it spurred conversations with clients and was financially successful.

Portfolio Management: Timing and Letting Go

In another client interaction, a key lesson learned from deploying software applications is the significance of portfolio management. In an ever-evolving market, timing plays a vital role in achieving success. This includes knowing when to release specific AI features and projects and when to terminate them.

During our software development journey, we incorporated a half dozen AI/ML propensity models into a larger marketing and real time telemetry data usage dashboard. The AI/ML provided smart capability that further enhanced the platform’s ability to increase sales. Each release is a program, and the collection of releases were an exercise in portfolio management.

Our initial plan was to deploy some high ROI programs concurrently or back-to-back. Our thinking was to push the higher ROI technologies to the marketing and sales forces sooner so that they would have greater impact on our corporate profit margins earlier in our fiscal year. However, we soon realized the importance of releasing these projects at the right time or just terminating them ultimately met our clients' needs, which we knew would create value later. These termination choices were based on doing too much at once or reading the changing market differently than before.

While it was difficult to let go of potentially high ROI projects, we made the decision to prioritize releasing the right AI/ML at the right time when they would add maximum value to our clients' operations. Terminating projects was an honest acknowledgement that market conditions changed. This required careful analysis of market demand and feedback from our team and other partners. By adopting a client centric approach to portfolio management, we were able to identify optimal release timings and ensure our clients derived the best possible outcome from our AI/ML applications.

One of the principle drivers for strategic flexibility is collaborative teamwork. VMware’s EPIC2 values of execution, passion, integrity, customers, and community infuses collaboration. At the intersection of these five value principles are the qualities of care, humility, and compassion. These qualities drive a desire to do right in terms of people and product. Interestingly, they also seem to reduce personal biases, which, facilitated the team reading the emerging market more compassionately and accurately. ?

In conclusion, deploying software features successfully involves incorporating design thinking principles and employing effective portfolio management strategies. Through design thinking, we actively listened to end-users, iterated on the product, and, in this use case, put aside our egos, resulting in a superior anomaly detection AI feature. Likewise, our client centric portfolio management approach enabled us to release right AI/ML enhancements at the right time, despite the initial allure of potential higher AI/ML ROI projects.

By emphasizing design thinking and portfolio management in software application and feature deployment, we can enhance user experience, anticipate market needs, and maximize the value delivered to clients. These valuable lessons have become the guiding principles in our software development processes and continue to shape our approach going forward.