Solving the Data Science Insights Disappointment

Data scientists are already hard to find and expensive to hire. Yet, many companies with significant investments in data science teams remain insight poor and reactive instead of being forward-looking leaders driven by data-driven insights.

“In some organizations, there seems to be a mismatch of business leaders’ expectations and what a data scientist can realistically deliver. As business leaders learn about data science, many believe that Artificial Intelligence (AI) and data science can instantly transform their businesses,” says Suganthi Shivkumar, vice president of Asia at Alteryx.

“Likened to be an inverted pyramid, the wide top reflects the business leaders’ oversized expectations for data science impact, and the small point represents the data science team’s current capabilities,” she adds.

The business’ view on AI also has to change. Shivkumar believes too many business leaders see AI as a “silver bullet” that can help them find a solution with “a snap of their fingers.” “The reality is far from their expectations. Data science projects typically involve trial and error methods and executing the same process multiple times before reaching the final outcome. This may take months to achieve.”

Data science teams also face other reasons for the slow insights: reluctance to share data across departmental lines and longer times to prepare unstructured data.

“These scenarios leave the data science teams dependent on the timelines of others. By the time the stale data arrives, business stakeholders might have moved on to new questions or devalued the role data insights can play in making business decisions,” observes Shivkumar.

Democratization is not easy

With demand outstripping talent supply in data science, it’s time to frame this growing problem differently. One popular approach is to democratize data analytics and not make it a function of a professional few.

Essentially, democratizing analytics is about employee empowerment. “It makes working with data less daunting, allowing knowledge workers faster access to insights through self-service. This results in increased ROI in analytics and greater innovation for the organization as a whole, as the power of data-driven insights rests in the hands of many, not just the data scientist teams,” says Shivkumar.

But a few things need to be done and put in place before this happens. First, companies need proper data governance, which requires a bit of reorganization.

Traditionally, companies worked with data scientists or analyst teams embedded with departments working on a siloed view of the disparate corporate data set. CDOs need to align data ethics, access, and security policies for democratizing analytics.

“This also helps when the CDO measures company investments in analytics with an overview of transformation priorities, use cases, skillsets, headcount, and business outcomes,” says Shivkumar.

You also need to modernize data and the data estate concurrently. It involves migrating data from legacy infrastructure to a modern database. The most common route to data modernization is migrating the data into a cloud database.

The data silo bane

Modernizing data estate overcomes a crucial hurdle that slows many data science teams in driving insights faster: data silos.

The non-standardized collection, managing, and analyzing of data across departments makes data silos a teething problem. They add complexity and put the data science teams on the back foot. Unclear data ownership and interdepartmental politics and worries can also get in the way.

Here, technology-driven automation can be part of the answer. “To break down data silos and ensure that data is accessible by all parties, advanced analytics automation platforms act as a bridge by aggregating data and connecting disparate databases and data sources,” says Shivkumar.

Apart from simply breaking down data silos, a human-centric, advanced analytics automation platform, such as Alteryx’s Analytics Automation platform, ensures that data is accessible to necessary parties by automating an organization’s analytics and data science processes, Shivkumar describes.

Just ask Pacific Life, a provider of life insurance. Alteryx enabled significant time savings, reduced operating expenses, and helped teams accelerate their analytics and reporting. Manual or spreadsheet-based projects that took more than 60 hours to complete could now be done in 30 seconds. Now, 75% of the teams’ analytics projects are done with the Alteryx platform.

Schlumberger, the world’s leading technology provider for the oil and gas industry, created an analytics and automation center of excellence that started with one member and swelled to sixty in only three years. The company leveraged 80 Alteryx workflows to automate accounting processes and reduce the analytics process time for each of their 400 accountants from four to six hours to two to three minutes. Their automated data flow saved the company more than 7,500 hours annually.

In accelerating time-to-value with automated process workflows, Thailand's National Electronics and Computer Technology Center (NECTEC) used Alteryx to create geospatial data analysis for agriculture land use management to build the Smart Farming system in Thailand.

Faced with the mission to transform Thailand’s agriculture industry, Agri-Map is an initiative to habituate farmers to productive smart farming by analyzing environmental changes that could potentially affect the country’s agricultural products. As the project was nationwide and operated in an open-source manner, the research team was overwhelmed with the amount of data types and sources to cleanse, prepare, verify and analyze during the start of the project.

By automating its end-to-end process of data cleansing, preparation, and verification for the geo process to projecting the refined data onto the Agri-Map Application using Alteryx, the team provided accurate feedback to their partners within a shorter time frame while saving substantial man-hours.

Automation can also unravel the knotty issue of data privacy and security. Governments across Asia have increased their scrutiny and added new data laws to protect personal data.

“On a regulatory front, Asian enterprises are held accountable when creating and using customer data at the operational level when improving business efficiency and enhancing customer relationships. As the workforce utilizes Alteryx APA – a self-service platform – to receive diagnostic, predictive, and prescriptive analytics, every employee is empowered to practice accountability for data security,” says Shivkumar.

The right culture

In democratizing analytics, culture is essential. It fosters collaboration between people from different disciplines to come together and focus on a critical problem using data.

This was one of the prime motivators for creating centers of excellence. But in many companies, such centers remain underutilized and cannot provide the kind of data-driven thinking that companies want and need.

Shivkumar thinks it’s time for companies to look closer at centers of enablement instead. “Centers of excellence focus on bringing together people from different disciplines and providing leadership, best practices, support, and training for a specialized area. On the other hand, a center of enablement is a team that runs on an IT operating model that allows enterprises to create reusable assets, collect APIs and enable leveraging of self-service and efficient delivery of insights-driven decision making.”

Companies also need to explore new ways to get the culture on the data-driven innovation track. For example, hackathons are proving to be effective.

For example, Deutsche Bank Manila's manual, non-value-adding processes hampered their digital transformation and compliance efforts. The Philippines’ regulatory board, Bangko Sentral ng Pilipinas (BSP), also enforced stricter measures for the banking and finance industry. Deutsche Bank Manila was at risk with increased operational costs, which could impact bottom-line revenue and create a loss of control and management.

So, Deutsche Bank Manila wanted to build a culture to inspire innovative thinking, explore new ideas and adopt new technology. So, in line with their goal to be data-driven, the bank launched the Alteryx hackathon in March 2021.

With 17 pods, 68 participants, and numerous pain points to address during the hackathon, Deutsche Bank Manila is looking to achieve approximately 566 hours a month of workforce efficiency. Aside from improved organizational efficiency, the Alteryx hackathon enhanced Deutsche Bank Manila’s workforce collaboration, empowered them to make a difference, and reinforced the experience that Alteryx’s platform is easy-to-use and intuitive.

We still need data science teams

So, what happens to the data science team if citizen data analysts now have the tools and a modern data estate to analyze?

We first need to understand what citizen analysts can and cannot do to answer that.

“A citizen analyst’s culture promotes non-technical employees to conduct insights-driven decision making without the traditional skillsets needed. According to Gartner, a citizen data scientist is a person who creates or generates models that leverage predictive or prescriptive analytics, but whose primary job function is outside of the field of statistics and analytics,” says Shivkumar.

But these same traditional skillsets will be needed when companies answer more complex questions, identify important outliers, e.g., an incoming black swan event, or correlate seemingly unrelated data sets. It also is vital when ensuring there is no model drift when building one with an AI algorithm with company data sets.

“There is a common misunderstanding that citizen data scientists will replace expert data scientists. In reality, they are complementary to the existing analytics roles. While citizen data scientists bring their own special expertise and unique skills that can drive successful analytics-driven tasks, data science teams are responsible for validating the models that citizen data scientists form before these models are moved into collaboration,” Shivkumar explains.

By focusing on higher-value data problems, data science teams can also turn disappointments into expectations by uncovering new exciting data-driven possibilities.

Winston Thomas is the editor-in-chief of CDOTrends and DigitalWorkforceTrends. He’s a singularity believer, a blockchain enthusiast, and believes we already live in a metaverse. You can reach him at [email protected].

Image credit: iStockphoto/RollingCamera