There’s no question that data analytics can be a significant competitive differentiator for companies, delivering insights that can help boost sales and market share. But how much enterprises gain through analytics can depend a great deal on how well they’re taking advantage of the latest technologies, and how prepared they are for future developments.
Here are some suggested best practices for getting the most out of data analytics endeavours.
Take advantage of self-service analytics
Self-service analytics enables business users to perform queries and generate reports on their own, with minimal or no support from IT and without the need for advanced analytics skills. They can leverage easy-to-use business intelligence (BI) tools that have basic analytics capabilities.
A self-service analytics approach can help fill the gap created by the shortage of trained data analysts, and can get data directly to the users who need it the most in order to do their jobs.
Business users can make decisions based on their analysis of data, without waiting for data scientists or other analytics experts to generate reports. This can be a huge benefit for companies that need to move quickly to adapt to market changes or to shifting customer demands.
The first step in deploying self-service analytics should be to fully understand the user community, including what information requirements they have and what tools they will need, says John Walton, senior solutions architect at IT consulting company Computer Task Group.
“Information consumers and executive stakeholders require a very different analytic tool suite than data scientists, and it’s important to align tools with business requirements,” Walton says. “Also, self-service analytics is highly dependent on clean data. If an information stakeholder loses trust in the dashboard they’re using, it’s really hard to get their trust back. They’re going to say, ‘I don’t believe what I’m seeing,’ and it goes south from there.”
It’s also a good idea to establish information consistency through a data governance initiative, Walton says. “Once this is in place, you can use a dimensional data architecture as the ‘plumbing’ for self-service analytics,” he says.
In such an architecture, the key performance indicators and measures displayed on a dashboard have been pre-computed based upon approved business rules, associated with the appropriate business filters or dimensions of analysis, and stored in the database. The analytics tool user doesn’t have to do all of this heavy lifting, Walton says.
Deploy machine learning capabilities
Machine learning can play a significant role in enhancing the data analytics process, particularly for organisations that handle massive amounts of information.
Machine learning will require a different architecture than analytics, Walton says. “Here you don’t want to apply pre-computed metrics that will skew the data and obscure potentially valuable insights,” he says. “ML wants to crawl through a vast amount of very granular data, most likely within a relational database, to most effectively apply its capabilities.”
For example, in the health insurance sector, a company might be dealing with massive data sets of claims data, patient encounter data, and both structured and unstructured notes.
A best practice for machine learning is to use the right layer of data for the right purposes, Walton says. “The bottom ‘ingestion’ layer is all the data coming in from your different sources, the rawest data that’s ideal for ML,” he says.
The middle, or “conformance” layer is where data has been taken from various sources and conformed to standards according to established data governance rules, Walton says. The top layer, composed of a series of focused data marts, is ideal for analytics, he says.
Manage data end to end
Many organisations are struggling to manage enormous and growing volumes of data from a variety of sources, and this can hinder analytics efforts. Deploying technologies to help manage data across the enterprise can provide a solution.
Healthcare supply company Paul Hartmann AG is using a central management platform from SAP, called Data Hub, to unify, access, and analyse data across multiple internal and external sources. The goal is to maximise the potential of data and gain the necessary insights needed to optimise manufacturing and supply chains, says Sinanudin Omerhodzic, CIO and chief data officer.
“With access to these findings, we can and keep our customers stocked with the products they need at any given time, ultimately saving patient lives,” Omerhodzic says.
By leveraging the Data Hub technology, Hartman was able to establish a “single source of truth” for customer, supplier, and operational data, helping it to better understand customer challenges.
The company is now in a position to better leverage technologies such as artificial intelligence (AI), the Internet of Things (IoT), and predictive analytics. And it can potentially use new data sources on factors such as weather and epidemics to better predict demand at hospitals and pharmacies and ensure that they have the supplies they need at the right time and in the right amounts.
Educate business users about overall data strategy
The business users who will be leveraging data insights need to understand the company’s strategy for data science, AI, machine learning, and data analytics overall. That way they’re more likely to make sense of what they’re seeing.
“Conduct discovery sessions so that business and operational leaders understand the benefits of AI and ML,” says Venu Gooty, global practice head of data sciences and analytics at HGS Digital, a digital transformation consultancy that helps organisations use data to elevate their customer experience.
“This is particularly important for organisations embarking on the data science journey for the first time,” Gooty says. “The biggest hurdle [HGS Digital] faced when implementing [AI and ML] was to educate the business users about the outcomes attained after delivering data science projects, and to explain our approach to delivering data science projects,” he says.
Organisations need to have a data strategy in place that explains how different departments work together, Gooty says. “This is required because ML initiatives require working with multiple departments,” such as marketing, IT, operations, and others, he says.
Machine learning involves working with large volumes of data, Gooty says. For example, in order for a retailer to predict customer churn, it needs many data sets such as customer demographics, purchase history, products purchased by the customer, etc.
“These data sets typically come from disparate data sources and there may not be a consolidated source to pull the data,” Gooty says. “So the team will have to work with different departments to get the data into a consolidated platform. In organisations where data strategy and data governance is defined, this is a much more seamless process than in organisations with no clear data strategy.”
Leverage analytics in the cloud
As with just about anything else in IT, the cloud offers cost-effective and efficient options for data analytics. It’s especially beneficial for organisations that need to analyse massive volumes of data and don’t have the internal capacity to handle the demands.
Any company that’s planning to perform analytics in the cloud should first define a clear migration strategy, Gooty says. “For most organisations, this will be the first time data is moving to cloud,” he says. It’s best to start small, learn from the experience, and make changes as needed, he says.
Also, define a clear governance framework with security policies. “Moving to cloud means moving internal and external data and users to cloud,” Gooty says. “The security and privacy policies must be clearly defined, and the owners of each section must be clearly defined. The right level of access needs to be provided for each user.”
Another good practice is to automate as much as possible, Gooty says. “The power of cloud is agility and automation,” he says. “There will be a lot of requests to do manual or one-time loads, and it’s better to push back as these one-off requests adds up.”
Establish an analytics centre or excellence
Organisations form centres of excellence (CoE) to provide leadership, share best practices, develop research, and offer training in a particular area of focus. Given the important strategic role of data analytics today, a CoE focused on these efforts makes a lot of sense.
A 2019 survey of CIOs and other senior IT executives in the U.S. by research firm International Data Corp. (IDC) showed that 93 per cent said their organisation is using some form of CoE to drive AI and data science initiatives. “The centre of excellence is the primary hub for all things AI, BI, and analytics,” says Serge Findling, vice president of IDC’s IT Executive Programs. “As an organisation with both central and distributed resources, it focuses on enterprise-wide coordination.”
Global consulting firm Keyrus notes that to get the best return on investment and the most value from its data, an organisation should establish an analytics CoE. The CoE streamlines all of the analytics efforts at the organisation.
“Imagine a highly capable team of experts that knows your organisation from within and is well-acquainted with your data sources,” the firm says. “This team possesses the skills and capabilities to leverage the data at your disposal to steer all of your efforts in the right direction.”
Keyrus says an analytics CoE should provide functions such as defining the organisation’s analytics vision, including selecting tools to use and determining which key performance indicators (KPIs) are needed; building a technology blueprint; establishing standards for areas such as how to share data sources; managing programs and controlling funding; developing user skills; and organising methodology leadership.
The firm notes that no two CoEs are the same, and how the group is structured might depend on the size of the company, its industry, its goals, and other factors. The CoE should be tailored to an organisation’s specific business objectives and organisational structure.