5 INSIGHTS THAT WILL GUIDE YOUR 2022 DATA AND ANALYTICS PLANNING
It’s Q4 2021...the home stretch of this calendar year: as an analytics leader, what should you be doing now to prepare for 2022 and beyond?
2022 is going to be the year of bionic data. Just like Lee Major’s character in The Six Million Dollar Man, today’s analytics are faster, stronger, and more relevant. Given the uncertainties spurred by the pandemic disruption, this past year has seen an escalation of analytic requests. The prognosis is that there will be no letup. We will see an increased need for high-value, on-time, and pertinent analytics. To bulk up their analytics machinery, firms will first have to inject capital into their data efforts, but the barriers to getting and using data assets are getting lower every year.
From our perch as a company that works exclusively to deliver data engineering and data analytic solutions for commercial pharma organizations, here are five areas of focus we expect to see in 2022:
1. There will be an increased reliance on data science teams to deliver value. There is no “automate and walk away” button but companies will get better at running ops on their existing models. The automation that we expect will come in the form of assuring that the existing models maintain their accuracy; similar to how organizations have spent decades refining their data cleansing techniques in their ETL (extract, transform, load) processes, AI (artificial intelligence) nand ML (machine learning) models will get more accurate as companies lean on them more and more. Because of this DS teams will be expected to do more in terms of explaining their approaches and they will be asked to explain any and all assumptions they bake into their models.
2. Accessibility and utility of advanced analytics will improve. After working with data science teams for several years and witnessing first-hand the intrinsic value from their investments, organizations will seek to expand the reach of their analytic efforts. Leaders are asking how their team members who are not data scientists can leverage their ML and AI efforts. Another angle on this involves developing even when the data is sparse and incomplete. Firms will get better at building inferences from the data that is already available from various data pipelines, while simultaneously tapping into proxy sources to fill in gaps. A use case in commercial pharma for this is to benchmark the performance of a competitor’s specialty drug in the absence of real metrics while sizing a sales force ahead of a launch. Also, we are all likely to see the term “data fabric” somewhere in a presentation before spring break (if you haven’t seen it already). Making models more accessible is part of that effort to reduce the friction involved in acquiring data and deploying models.
3. The analytic plan will become a must-have. Our experience has taught us that analytic deliveries always start with a solid analytic plan that is understandable to the business yet accessible to the analytics team. The elements of an analytic plan should include the following:
The exact questions (phrased by the business) that need to be answered. If the business problem cannot be articulated, how are the data scientists supposed to solve it?
An inventory of the inputs that are required to address the problem. Nothing can start until this list is complete. Example: if you are creating a sales forecasting model, you will want to have historical sales data as your input. This seems obvious, but it has to be clearly delineated.
A complete listing of the data assets needed. More detailed than the input information, someone has to go through the exercise of locating the exact database, schema, and table names that will be needed. Spending the time upfront to do this engages the data engineering team, and it holds them accountable for making sure that no modeling begins until the requisite data is cataloged.
The analytic approach section of an analytic plan is a step-by-step delineation of how the data science team will engineer the model. Written in plain English, the business should be able to tie this section back to the inputs and how they will be used.
Finally, the analytic plan needs to define the timing and the outputs.
Planning and foresight enable analytics. More than just a project artifact, the analytic plan is the catalyst for high-value analytics delivery.
4. The team structure makes or breaks the data and analytics value chain. August 2021 in the United States saw a record 4.3 million workers leave their jobs. The disruption that started in March 2020 is seeing a ripple effect of workers exiting jobs, employees working from wherever they want, and the disintegration of in-person collaboration. None of this alters the basic tenet that effective teams are made up of experts who are also team players. For successful analytics projects there are three pillars needed for success:
business context experts (analysts) + data engineers + data scientists
The communication chain for the team is critical to its success; emails are not usable artifacts. As a team, the narrative behind each project has to be consistent and every member of the team should be capable of articulating what it is they are working on and why. Beyond well-constructed analytic plans, we expect analytic leaders to have to double down on their efforts to keep their teams focused on their missions. This means that each of the highly skilled practitioners along the three disciplines must understand the modalities of their peers. 2021 looks to be an opportunity for teams to cross-educate across the different lanes.
5. High-velocity analytics are now a way of life. Questions from the business come early and often when conditions are changing. At Sentier, we have never experienced a period when our clients required as many answers in short timeframes as we have over the last twenty months. The expectations have now been set and turnaround times of a half a day to a week at the maximum are the norm. We are calling this high-velocity analytics and in addition to everything listed above, they demand:
The analytic data assets are up to date and ready to go at a moment’s notice. Data-driven recommendations that truly reflect the “state of things” can only be achieved from the latest information available. It is imperative that analytic ready data be managed in a way that provides an up to date starting point for all analytic requests.
Machine learning is a mandatory discipline. There is no other way to sift through large amounts of data and discover the relevant insights at the levels required (i.e. customer, individual tactic, etc.). For lack of a better analogy, you can pan for gold in the river or use machines to search acres of real estate to discover gold. You will find more gold (insights) using brute force and machines.
TAKEAWAYS:
There is a common thread to the challenges that 2022 will present: how can data and analytics teams get the most value out of what they are building and from the people who build them?
Value comes from aligning the disciplines inherent to successful data science projects. Careful planning, attending to the quality and availability of the business assets (both data and models), and being able to articulate the what, how, when, and why of your efforts should be core elements.
----
Sentier is an advanced analytics company delivering disruptive insights to our commercial pharma and biotech clients through our AI driven platform. The platform allows Sentier to deliver insights faster, at a lower level, and more accurately than traditional solutions.
