By now, every industry has positioned themselves as the next great sector to benefit from big data. Healthcare? Of course. Finance? Early adopters. Agriculture? Very much so. Ommichannel retailing? Absolutely necessary.
Even transportation logistics and trucking, from UPS’s ORION project to Schneider National’s extensive datasets. Data analysis leads to increased revenues as well as cost savings, both running in the millions annually.
Study the big data successes in these industries, and you’ll find a number of common threads, beginning with the creation of a strategy or roadmap. Having a guide to achieving your organization’s vision is a base-level need and can only be developed when a clear vision exists.
Failure looms if your data strategy can’t answer:
What do you want to know?
Which factors are most important for increasing margin on a given product line?
How much does social media reflect recent activity in your business?
From where will the data originate?
What external data is available that might give insight to internal data?
Which outcomes do you want to predict?
6 steps to strategy
Much of the success of a big data strategy is tied to an organization’s culture, its appetite for growth, and executive-level support for the initiative. These six steps have been implemented in multiple projects to deliver a winning big data strategy and will ensure your vision is properly focused.
1. Convene the perfect multidisciplinary team
Big data is not an IT project. It is a business initiative, and the initial team should have more representatives from business departments than from the corporate technology group.
Members typically include knowledgeable staff or managers from finance, business development, operations, manufacturing, distribution, marketing, and IT. The team members should be familiar with current reports from operational and business intelligence systems.
A common thread? Each has ideas about performance indicators, trend analysis, and data elements that are currently not readily available to them. More importantly, they know why having that information readily available would add value.
2. Define the scope of a given problem
What problem should be analyzed? Take three problems you’d like to have solved and formulate them into questions. Write them as the subject line on three emails. Send them to all members of the multidisciplinary team. The replies will guide your efforts in narrowing (or expanding) the initial scope of study.
A 360-degree view of all customers in an enterprise may be too ambitious for an initial project. But finding the characteristics of commercial customers who have bought products from multiple lines of business in five key geographic markets might be a more manageable scope right out of the gate.
With this approach, iterations in development provide expansion to all lines of business or to all markets in cadence with a company’s business pace.
3. Assess internal data sources and silos
Know what data exists internally from a functional viewpoint before getting into the technical details. Gap analysis will uncover incomplete data, and profiling will expose data quality issues.
If customers for one line of business are housed in an aging CRM, and customers for a newer line of business are found in a modern system, a cross-selling opportunity analysis will point up the need to integrate those data sources.
Do you have an inventory of data sources written in business language? In forming a strategy, a team will want to have references, such as vendor contracts, customer list, prospect list, vehicle inventory, AR/AP/GL, locations, and other terms that describe the purpose or system from which the data is derived. The list can be expanded for technologists later.
4. Select external data sources from government, social media and industry
External data sources augment the understanding of data from internal transaction systems. Data.gov has over 100,000 datasets, some containing millions of rows covering years and decades. The range of options can easily lead to data hoarding.
Plan to download only five datasets for each of the three questions that you are trying to answer. For instance, the Consumer Price Index – Average Price Data from the Department of Labor Statistics includes monthly data on changes in the prices paid by urban consumers for a representative basket of goods and services. City and state governments are now providing downloadable data, too.
Twitter, Facebook, and Pinterest posts may have a greater impact on your operation than you realize. Be sure that a couple of members of the team pursue data from social media sources to include in the initial study.
5. Determine output and analytic measures
Continuing with the questions you posed to limit scope, determine the proper output you expect. Is an overview valuable, showing just a few key indicators, with an option to drill deeper into details? Who needs to see output? Should they see it only on a large monitor at their desk or is mobile viewing and response expected? How should the output be validated?
There are many points to address, including the availability of data through existing dashboards and reporting systems.
6. Get experienced guidance
Jump-start your big data implementation by engaging an experienced team who has led others through data strategy and implementation. Seek guidance especially if your organization doesn’t have a data glossary, network administration, or knowledge of the intricacies of implementing new technologies.
The lessons learned and templates built that your guide has developed over many engagements can provide your company with a smooth progression through the many tasks at hand. Expect the creation of online documentation and knowledge base in a form more usable than text documents in a shared folder.
A data glossary is a result of governance and provides a common set of terminology that will be used when analyzing output of data processes
Business data elements, or BDEs, are key to creating understandable and cross-company analytical outputs, including reports, charts, graphs, indicators, and other visualizations. Proper setup of a multi-server starting point, even if only 4 or 5 nodes, is necessary to avoid immense headaches when your success necessitates adding 10 or 20 more nodes.
That big data strategy edge
Planning a big data strategy will require you to rethink the way you manage, operate, and analyze your business. And getting guidance while developing your strategy prior to your implementation can help you avoid problems that might waste valuable time to resolve on your own. It’s important to recognize that big data ecosystems are still relatively new.
But these six steps for planning a strategic and informed response to the tidal wave of data that is now available will give your company a considerable edge in your market.