Applied research and product management - fueling innovation

Companies often mention "innovation" as a critical business driver. However, Applied Research and scientific discovery can often be overlooked when it comes to sustained business investment.

Innovation typically originates from teams in the midst of shipping products and maintaining platforms. This innovation is often treated as side-of-desk work rather than a genuine strategic investment that creates long-term business value.

According to some studies, the failure to have sustainable innovation is rooted in the lack of innovation strategy, or the alignment between innovation and business goals.

The power of product management in applied research

Some of the world’s most innovative advances in industries ranging from telecommunications to automotive, industrial and manufacturing have come from Applied Research labs. But Applied Research in a corporate environment isn't just about scientific discovery; it's about translating those discoveries into tangible products and solutions that address real-world business and customer challenges. 

This is precisely what Product Management is designed to do. Product Managers strive to bring a unique blend of strategic thinking, customer focus and business acumen to problem identification and solution evaluation. They can bring this thinking to their partners in Applied Research teams and work together to select and prioritize from among the infinite list of potential experiments. 

The best outcomes happen when both teams find ways of generating a chain of short term value creation into the biggest possible, even transformational, long-term results.

Bridging the gap between research and market needs

Product Managers act as a crucial bridge between the research lab and the market. Their role is to understand customer pain points, market trends and business objectives, ensuring that research efforts are aligned with actual user needs and commercial viability. This customer-centric approach helps applied research teams focus on real-world problems that truly matter and develop solutions that will be adopted and valued.

Established products maintain a backlog of feature improvements, anchored by a high-level strategic direction and goals that are driven by user problems. 

In the applied research space, you typically have to maintain two backlogs:

• A business/customer backlog of yet unsolved problems or potential areas for increasing business value
• A research backlog of promising technologies and research avenues to experiment with against said business problems

As Applied Research teams work through these backlogs, it is critical to find the intersection of the two in order to prioritize those problems, technologies and proofs-of-concept to pursue first. 

Explore-exploit balancing

A core concept of decision making is the explore-exploit tradeoff. Exploration involves experimenting with new options that could yield lasting long-term results, even if it means sacrificing a short-term value capture. Conversely, exploitation entails selecting the best available option based on existing (potentially incomplete or inaccurate) knowledge of the system. A key challenge in many decision-making scenarios aiming to maximize long-term gains is to strike the optimal balance between these two approaches.

Traditional product development teams primarily operate in "exploit" mode once their backlog is prioritized and the top customer problem is identified. Exploitation primarily remains as an infrequent, focused activity through hackathons or innovation sprints.

In contrast, Applied Research teams can remain in explore mode for longer periods of time as they progress through their research and business backlogs, shifting to exploitation after a proof-of-concept has been established and proven, and the team has decided to move on to the Minimum Viable Product (MVP) stage. To maintain long-term vision, Applied Research teams deliberately maintain some of their capacity towards exploration, even post-MVP stage. 

Product Management plays a vital role in Applied Research by continuously providing input to exploration and exploitation tradeoffs. Its ability to apply both short- and long-term prioritization, guided by business value and a research agenda, is crucial for success. Without continuous, rigorous prioritization, it's easy to fall into the traps of optimizing for near-term gains or deviating from strategic imperatives of the business.

Elevating both product and applied research

The best outcome of a close collaboration between Applied Research and Product Management is a whole that is greater than the sum of its parts. By having researchers engage with business partners early and understand the strategic direction of the business, it enables them to bring business logic into their experimentation, as well as adjust their targets to align with those that will maximize business value. 

Additionally, bringing Product Managers into the research and experimentation process pushes the Product Managers to elevate their understanding of the technologies that their products will be based on. Having an intuitive understanding of the tech allows Product Managers to have rich exploratory conversations with business teams and provide an initial assessment of where different technologies may be beneficial.

From research to impact at Capital One

A prime example of this powerful collaboration can be found within Capital One's AI teams.

With the advent of Transformer technology in 2017, and its implementation as a consumer product in 2022 and onward, our Frontier ML team understood that this burgeoning research avenue is a critical one to invest in, and its applications may have far-reaching consequences for our business.

While 2023 saw the team exploring various research and development fields including GraphML, Data & Privacy, Anomaly Detection and Sequence Mining, the powerful capabilities of Transformer models quickly became apparent, especially in enhancing existing use cases like fraud detection, which previously relied on sequence mining and Markov processes. This led to a long-term exploration of Transformer applications across diverse prediction tasks, while also securing immediate benefits by implementing a Transformer-based fraud detection solution.

Through extensive internal exploration, our Product Managers identified both short-term gains and substantial long-term potential for enhancing decision-making in key business applications like fraud detection and personalized customer experiences. This realization led to a joint decision across Research and Product to focus on prioritizing further advances in Transformers and sequential modeling for these types of real-world applications.

We’ve seen incredibly successful early results, which is a testament to a very specific approach: strong research focus on performance-enhancing avenues (model enhancements, tokenization techniques, infrastructure optimization), alongside long-term strategic research focus areas such as synthetic data generation. Concurrently, the product team collaborated closely with business partners to identify optimal opportunities for short-term value capture as our scientific work advanced. This collaborative exploration was crucial in prioritizing things like data and model configurations for the initial version amidst numerous possibilities.

Ultimately, the success of this approach and the team behind it relied on robust cross-functional collaboration, a deep understanding of both research priorities and strategic business drivers and a continuous process of balancing exploration with exploitation in prioritization.

Key takeaways

Strategy-backed Applied Research: Connecting business needs and long-term term strategy into Applied Research agenda is critical for long term success. Product management can help bridge the gap between innovative ideas and real-world, impactful products.

Balancing "Explore" and "Exploit": A key challenge is managing the "explore-exploit" tradeoff, which is the balance between exploring new, uncertain ideas and exploiting well-understood, proven concepts.

Dual Investment Strategy: A dual investment strategy can effectively manage this tradeoff by maintaining separate backlogs for business/user problems and research initiatives.

Collaborative Success: The success of this model hinges on the close collaboration between product managers and applied researchers to ensure alignment and prioritization of both backlogs.

What does this mean for you if you’re not in Applied Research? 

While the concept of Applied Research seems specific, some of the principles we discuss here apply broadly to all teams. Here's what you can take away, no matter your role:

• Embrace the “explore-exploit” mindset: The explore-exploit tradeoff isn’t just for researchers. In your work, think about how you balance exploring new, creating ideas with exploiting proven methods. Dedicating small amounts of time to exploration can lead to unexpected improvements.
• Innovation benefits from focus: While many like to think of innovation as “unconstrained whitespace thinking,” in reality the best innovative efforts benefit from business focus and targets. Having a long-term strategy and aligning your innovation efforts towards that strategy will help those efforts bear fruits and result in implementation.
• Collaborate with an Applied Research team in your area: While not everyone has the luxury of being able to employ Applied Research teams, if your broader organization invests in innovation or applied research, seizing the opportunity to explore high-leverage business problems with researchers can help you drive value in your area both short- and long-term.