Credit: Igor Kisselev / Shutterstock
Ideas Made to Matter
10 best practices for analytics success (including 3 you can’t ignore)
Companies are head over heels for analytics, convinced that data is their lifeblood and data-driven insights the key to magically unlocking future success. Yet despite all the enthusiasm, budgets, and resources directed to analytics, the vast majority of projects simply aren’t able to scale, with many failing to meet expectations entirely.
According to Gartner research, only 20% of analytics insights will deliver business outcomes through 2022. That’s because most companies aren’t following a set of established best practices, operating instead from a mostly haphazard and unproven playbook.
Prashanth Southekal, a business analytics author and professor and head of DBP-Institute (Data for Business Performance), is determined to change that dynamic.
“In my experience, most companies have a lot of resources, they have the technology and very smart people, and they have tons and tons of data,” he said in a presentation at MIT’s 14th annual CDOIQ Virtual Symposium in August. “But [success] isn’t about data collection, it’s about data management and insight.”
Southekal outlined 10 analytics best practices, zeroing in on a trio of gold-star tenets he said are absolutely crucial to analytics success.
Three bedrock practices
TAKE AN ANALYTICS VIEW OF DATA. In simple terms, this means reconciling the questions being asked by the business with the kinds of data needed to deliver answers. That answer will in turn dictate what model to use to gain insights.
For example, an organization might be steeped in documents and data helpful for compliance initiatives, but if the business goal is to better understand the customer and offer products and services tailored to their needs, the stores of PDF documents and spreadsheets might not be relevant.
Classifying business data by type ensures it can be more easily pulled in to analytics efforts when and where it makes sense. Southekal identified three major data types: Reference data, covering business categories like plants, currencies, and line of business; master data about entities such as suppliers, products, and customers; and transactional data, which details events like purchase orders, invoices, and payroll runs.
Southekal also suggested companies develop a further set of parameters and conversion rules to transform their data to a state that lends itself to analytics. “Analytics model selection is based on two major things — the questions I ask and the data type [I have],” he explained.
SOURCE DATA STRATEGICALLY. Plenty of companies put the brakes on analytics because they don’t have enough data or the right data, or maybe they believe the quality of their data is bad. But waiting for the perfect state of data is a mistake, Southekal said. “The unicorn doesn’t exist when it comes to analytics in business,” he says. “It’s not about perfection — analytics is all about progress. You need to keep moving.”
Southekal said there are ways to compensate if organizations are lacking in data volume or quality. Data can be acquired, either purchased through providers or from free open source resources, and organizations should balance the cost of acquisition with the value the data brings to the analytics effort. In addition, sampling can make data more useable and reduce cycle time. Another option is feature engineering, which employs machine learning tactics to parlay an existing data set, domain expertise, and intuition into smarter data tuned for analytics.
MOVE FROM ANALYTICS PROJECTS TO ANALYTICS PRODUCTS. Instead of channeling efforts to analytics projects, which are finite and tactical, organizations should set their sights on analytics products, which generate measurable financial benefit from data insights while improving business performance. Data products are typically scalable, teams stick around for continuous improvement, and there is inherently more collaboration, Southekal said.
7 general rules of the road
Beyond the trio of must-do best practices, Southekal recommended the following as guidance for a successful analytics journey:
Tie stakeholder goals to questions and key performance indicators. Everyone knows you have to enlist stakeholders early on in a program to build engagement and support, but it’s less clear how to pull that off in a way that makes sense. The key is asking the right questions, not just about what stakeholders want or specific requirements. It’s also important to clarify assumptions as part of that exercise to provide additional context. Stakeholders should also be enlisted early to establish mutually-agreed-upon KPIs to ensure business goals are being met.
Build high-performance analytics teams. Successful analytics requires more than highly specialized data scientists who work in silos, Southekal said — it demands that companies cultivate a different organizational mindset, one that embraces hypothesis-based methodologies and where analytics expertise transcends financial or regulatory reporting to involve operations and revenue growth.
Only 20% of analytics insights will deliver business outcomes through 2022, according to Gartner research.
Build data literacy by focusing on descriptive analytics and key performance indicators. Experts estimate that 80% of the current work done in analytics encompasses descriptive analytics — that is, an historical look back to determine why something happened — for example, why sales dipped during a certain period or why specific forecasts were off. Broadening these efforts helps get stakeholders used to the new mindset and gives them first-hand experience with the benefits of analytics, Southekal said.
Make compliance an integral part of analytics. While data can be an asset, it can also be a liability. To counterbalance the risks, organizations should put an emphasis on compliance, including government regulations, internal business rules, and industry standards.
Refine analytics models continuously. Building an analytics model is not a one-and-done exercise — companies need to adjust modeling efforts to keep pace with business changes, whether due to mergers and acquisitions or entering new markets. “When things are changing, your data also changes, and when your data changes, your models change,” Southekal explained. “Analytics models are not a constant entity.”
Support analytics with governance. As data collection efforts ramp up, governance becomes a critical factor. Establishing formalized processes ensures data is captured and managed consistently, quality remains high, and there is a common definition and understanding of data across the organization.
Use data storytelling to promote insights. Cryptic or confusing data points won’t be enough to spark new business patterns or change behaviors. Leverage visuals, context, and the financial benefits of data-driven insights to weave a narrative that educates stakeholders and associates insights with financial benefits.
“Insight generation is great,” Southekal said — provided it’s followed up with action. “When it comes to analytics, what you do with that insight is equally important. That’s part of the business efficiency.”