Design, deliver, and run enterprise blockchain workloads quickly and easily.
All servers and systems
When it comes to big data—the practice of mining large data sets—small and midsize businesses (SMBs) can feel intimidated or disinterested. Big data refers to extremely large amounts of data, with database sizes that few SMBs ever see. As a result, an SMB can conclude that because the data is big and the business is not, this technology is irrelevant.
However, the payoff to using big data is in analytics: the process of finding trends and patterns. That’s a field on which SMBs can play. Data does not have to be big to provide business insights in areas such as internal operations, customer behavior, campaign effectiveness, and market opportunities.
Big data and data analytics have generated a lot of buzz because they promise—and deliver—results. Consider these two compelling statistics:
There are several ways in which SMBs can capture, store, and use the data around them. Smaller organizations do face unique challenges, however, and must navigate around the gotchas in implementing data analytics. Here’s how.
SMBs need to cut IT costs yet build a storage infrastructure that supports analytics. Here's how.
"The first step in capturing data is to understand and utilize what you already have," says Jim Harder, CEO of data analytics consulting company Visual Data Group. Start with departmental reports, such as sales by region, inventory levels, and accounts payable. Then you can move forward to capture and consolidate additional data.
But which data? It may be tempting to get all the data possible and sort it out afterwards. But doing so can lead to analysis paralysis. "There's just too much data to dig through, especially for an SMB," says Chris Stephenson, CEO of cognitive analytics platform provider Topos Labs.
To avoid getting overwhelmed, use context to focus your data collection efforts. "Don't boil the ocean if you need a cup of coffee,” says Stephenson. “Don't take the Google approach; you don't need it all. Soup stock vs. financial stock vs. gun stock: It's all stock, but if you're a financial services company, you're only interested in the second variation."
The idea is to collect data that helps you answer questions, both the queries you know (“What are our sales trends?”) and those you don’t (“Hey look, our sales are up in cities with a history of good jazz”). So make sure you’re asking the right questions and collecting data that enables you to answer them.
Unfortunately, many SMBs are not sure where to start looking for data to collect. Michael J. Prichard, CEO of data intelligence platform company Metis Machine, encounters this when he visits potential clients. When asked for two years of data to start with, many business owners say, “We don't have that, so see you in two years."
However, SMBs usually have more useful data than they know about. "There are lots of things that folks are still not capturing, like ERP or transactional data," says Harder. "But there is a lot of data in every organization. It might not be easy to get to, but it’s a place to start.”
What to look at? Consider sales reports, website inquiries, customer support requests, email marketing, website metrics, and oodles of internal operational data.
“We have it” doesn't necessarily mean “We can use it.” "Most end users who request analytical tools don’t understand the underlying data," says Jason, who works for a midmarket manufacturing company. For example, more recent models of manufacturing machines can offer more available speed settings and capacity than the older ones, but we're still using those previous models as well. Data analysis has to take that into account. “Because of these differences, the data we get from our equipment, between new and old, is not apples to apples," says Jason.
To get the answers you need, you have to normalize the data. Jason offers a manufacturing example: "Operational equipment effectiveness (OEE) is a common manufacturing metric made of three other metrics: availability, yield, and run to standards. Even if you have two similar machines with different speeds, the OEE will be normalized because it takes out the differences,” he says. “In the end, we can answer the question, 'Are we getting what we expected and paid for from our equipment?'"
Great: You’ve discovered that your company has (or can get its hands on) useful data worth mining. But before you consolidate the data into one place so you can run analytics on top of it, consider the deadliest enemy of good analytics: dirty data. Bad data translates to exceptions or undocumented differences in the business rules that drive the separate ways departments store information.
"Data is stored not only from the data perspective, but from a functional perspective,” says Harder. For example, your marketing department may use five-digit ZIP codes in its customer relationship management (CRM) tool, and across the hall, the shipping department uses nine-digit ZIP codes in its logistics software. “When you try to bring them together into one database for analytics, they won’t match up," Harder points out.
Bad data is an insidious problem. "Most companies I've seen have not transferred their tribal knowledge to business rules," says Jason. "Even working here for 16 years, it's hard to see through all the data because of the exceptions."
Other factors contribute to dirty data as well. According to a survey conducted by 451 Research in 2015, 58 percent of bad data is the result of data entry errors.
"It's not the data science that's the most difficult part," explains Prichard. "It's the actual data pieces: getting the data, knowing what is the right data, cleaning the data, and moving the data. 60 percent of data scientists spend most of their time cleaning and labeling data. That earlier part is hard."
You need to invest time and effort in cleaning up the data if you ever want to reap the data analytics benefits. As Harder puts it, "Having data is one thing. Getting it to the point where someone can use it, and it can have impact, is the grist in the mill."
As an SMB begins to collect and consolidate its data, it quickly runs into the second challenge of data analytics: where to put it all.
"To some degree, the best data storage answer depends on business size," says Harder. Smaller companies often store their data in the cloud because of its affordability. That’s especially so if the SMB’s product is data, such as at publishing company Green Builder Media, which almost exclusively stores everything in the cloud. It makes sense; they're already there.
Medium-size organizations may opt for a hybrid solution, Harder says, “because midmarket businesses typically have systems on-premises already." On-prem solutions enable them to capture the data and run analytics speedily, and the cloud enables them to store historic data affordably.
This is exactly how it played out at Jason's midmarket manufacturing company, where some data is on-site and some is in the cloud. “The local data set contains the last three months of collected data; that’s where we run analytics every day," he says. "The cloud is more for data warehousing, so we can look at historical trends and patterns."
Among the storage decisions is balancing speed and cost. "We store most of our client's data in the cloud, but it's not all in easy-to-access spots,” says Prichard. “The most current data is easily accessible, but many of our clients store their historical data in a less-expensive archival system such as Amazon Glacier. Getting at that data may take more time, but at least they're spending less to keep it."
All this effort is worth it only if it improves the business. And, say SMEs, it does.
"Brands are trying to find unique and interesting ways to sell more products," says Stephenson, "and you need to know what drives people to buy." Data analytics can give SMBs that knowledge. "Not every question needs to be successfully answered," says Harder, "but you need to ask a lot of questions."
Technology selection management platform SelectHub uses its data to make all sorts of decisions. Michael Shearer, SelectHub's VP of marketing, details three concrete examples:
Don’t fool yourself: Implementing and sustaining data analytics is not a simple process.
"The whole process is hard,” says Prichard. That’s true in every phase, from ingest (collection and consolidation) to process (modeling and analysis) and delivery (data visualization and reporting), he says. "If you can't do all three, you're subject to your weakest link. People get enamored with the process (modeling and analysis), but you need the rest."
It's rare for SMBs to have institutional knowledge around data analytics, let alone internal expertise. That’s why many SMBs engage with analytics-as-a-service (AaaS) platforms and external consultants to bootstrap their analytics capabilities.
Are the results worth the hard work? Green Builder Media’s CEO Sara Gutterman believes they are. "Our data gives us a crystal ball into the way the market is going," she says, "It helps us understand our audience of early adopters and innovators."
The benefits of data analytics are becoming harder for SMBs to ignore. Data analytics leads to more accurate decisions and enables SMBs to measure the effects of those decisions. And while the data doesn’t have to be big, the impact it has on your SMB will be.
This article/content was written by the individual writer identified and does not necessarily reflect the view of Hewlett Packard Enterprise Company.