
Businesses understand the benefit of investing in artificial intelligence to keep up with accelerating trends. Even business leaders who for years prided in their ability to create successful ventures by trusting their " gut feeling" alone, have conceded to the power of data driven decisions. Whether organizations want to build their own full stack Data Science team or decide to partner with an AI consultant, business leaders are well aware of the high costs and time involved in implementing machine learning algorithms. So most companies work with management consultants to identify critical business areas that could be transformed by AI and work towards bringing data science recommendations in those impactful processes. This approach is more fail safe and cost effective than introducing machine learning models across the entire organization.
Discovery sessions and innovation days with business units help bring the spotlight on key challenges in the business and then data science product managers are tasked with building requirements and collaborating with data science teams to automate and optimize the process while trying to achieve precise outputs. Most data science product managers come from an analytics or engineering background and they do possess the wisdom on how to deliver the right solution with the data science team partnership. But this also ends up being a vulnerability, because the product managers are keen on solutioning and more often assume that the business team has done its groundwork on assessing the processes in its entirety. With most product managers who come from an engineering background, there is always a voice whispering through their brains, relating all the ways to deduce the right logic, data needs to solve the problem even early in the requirement phase. Leaving behind the promptings of technical wisdom, data science product managers should instead open their minds to getting an in depth understanding of the business process. Data science product managers should start by building an end to end process flow while factoring in interconnected systems and processes before honing in on the solution. Just focusing on setting things up and sticking the landing without understanding all the threads in the middle will stifle the benefit the organization would have reaped from completely adopting the data science output.
For instance, every retailer agrees that to stay ahead of the competition, one has to use advanced analytical models that syncretize historical data with all the external variables impacting consumer behavior to build innovative and personalized solutions. There is extensive research on the part of the data science teams to analyze market insights and transaction data to recommend the right product for the right time and location. But to get the full potential of the data science models, product mangers shouldn’t just look at this requirement from a merchandizing perspective but also spend time in understanding downstream supply chain constraints. Sometimes the logistics costs to meet hyper localized demand outweigh the revenue increase from personalized offerings. If product managers start with building a Process map, it would help bubble up impact from all interconnected steps in the process. At this stage, the product managers should have an honest conversation with business on rethinking and redesigning some of their processes, if they hope to make data driven decisions. Before having these conversations, the product managers should work with data analysts to create estimated ROI from analytics to drive their point. Sometimes this might involve taking a step back on delivering a hyper tailored product mix for each region and instead opting to build optimal customer profiles clusters which can help create an equipoise between localization and meeting supply chain constraints. But by doing end-to-end process analysis and having a discussion early in the requirements phase on the execution constraints, the data science product manager can ensure that the final product seamlessly is adopted and utilized.
As a product manager, the initial question is never "How can I solve this?", but rather "What are we trying to solve" and more importantly " Why do you need this?". Sometimes business teams are so used to performing quotidian tasks like second nature that they don’t look to enhance the routine, the requests are mostly around new challenges. To put things in context, there might be a request for omnichannel pack optimization; an algorithm to effectively build optimal case pack mixes to ship to distribution centers or stores. With COVID, most retailers might be looking for an effective solution to addresses omni demand at a geographic level while also factoring in logistics costs. But there could be other core activities downstream on this process that employ outdated tools that cannot handle the output from a complex algorithm. In cases like the above example where multiple domains are involved, it’s hard to stay in sync on project goals. The downstream systems and processes need to be updated first before the analytics output is implemented. Building an advanced analytics solution before any process redesign will in no way significantly save costs or increase revenue and the data science investment will be rendered useless. Before investing in the solution, clarify the efficiency of the current process and the existing systems in place.
Data science Product managers’ strength in engineering or analytics and obsessing over the details shines through after the process mapping phase. Often business teams tends to overlook the data or system dependencies between the tasks, product managers could overlay a data flow diagram over the process map to unveil technical challenges that could crop up later in the project.
In conclusion, data science product managers should focus on having a thorough understanding of the business process. The process map should not stop with the analytics output being ingested but also capture all the systems/processes that lead up to the end customer. After this phase, data science product managers can leverage their engineering background to add those finishing touches to the requirements by narrowing in on the technical details and addressing complexities that might arise in execution. Data science product managers should don the hat of a business client during the requirement phase while also bringing an analytical view afterward to bring to light process/ system redesigns that need to happen if a domain were to successfully implement data science solutions.