5 pitfalls of self-service BI
- 20 October, 2017 21:00
Over the past several years, businesses have increasingly adopted self-service business intelligence (BI). In its November 2016 “Self-Service BI Market – Global Forecast to 2021” report, Research and Markets forecast the global self-service BI market would double to $7.31 billion by 2021.
The benefits seem clear. Whereas traditional BI is frequently seen as slow and rigid, self-service BI promises ease of use and agility. With self-service BI, business users can get access to the data and insights they need, when they need them, without having to rely on IT, which can often be a bottleneck with traditional business intelligence. By bypassing IT, the business can better capitalize on opportunities and quickly react to problems.
Dave Mariani, founder and CEO of startup AtScale, provider of a universal semantic platform for BI on big data, believes self-service BI offers many advantages, but he also sees several unintended consequences that organizations adopting the self-service model need to address. In his years as vice president of engineering for advertiser analytics and then vice president of development, user data and analytics at Yahoo, Mariani experienced these pitfalls first-hand.
[ Deliver deep insights with the 7 keys to a successful business intelligence strategy and learn why machine learning is the new BI. | Get the latest business intelligence and IT strategy analysis by signing up for our newsletter. ]
Following are five side effects of a self-service BI strategy that you’ll want to avoid.
1. Business metrics chaos
To get value out of BI tools, business units need to feed them data. In general, this means business units stand up and manage their own data marts — subsets of data warehouses that contain data specific to a business line. Because individual business units are typically responsible for all the hardware, software, and data that comprise their data marts in a self-service BI environment, those business units will inevitably create their own definitions and metrics. That's not such a big problem if that business unit is the only user of the data, but it becomes a large problem when trying to compare reports from different business units.
"You've gone from a central model where there was tight control of the business metrics, and you've put that in the hands of the masses and it creates conflicting definitions," says Mariani.
Mariani notes that during his tenure at Yahoo, the company's business units had myriad definitions of ad impressions and visits.
"Everyone calculated them on their own," Mariani says. "It's a problem because now the business is not aligned. They're all telling different stories. That causes a lot of consternation. They don't trust the numbers anymore because people have come up with their own definitions to satisfy their own goals."
"The average organization cannot get anything done just relying on one business unit," adds Bruno Aziza, AtScale's chief marketing officer. "For example, sales and marketing need to work together. If they can't name the problem, they can't fix it. If you can't drive alignment, it's really hard to solve a problem."
2. Business users forced to become data engineers
Self-service BI and data marts can lead to some quick wins for business units, but as data marts grow and the data changes, business users must spend increasing amounts of time managing them. Eventually, if the data grows big enough, the business users that once sought a way to bypass IT may need to go to IT for help, and dump a gnarly mess in IT's lap.
"Basically, we're asking our business users to become data engineers," Mariani explains. "That's less efficient; it's definitely not the best way of accomplishing the goal. The business units may have some early success with standing up new visualizations or dashboards, but when the data changes, they have to fill that role that IT has been filling historically, and they're not able to do it. They then have to call in IT to bail them out, and IT is not prepared or happy to bail them out."
Some organizations have tried to address this problem by embedding business users in IT, or deploying IT pros to individual business units, where they act as brokers between the lines of business and IT.
"That was the model we had at Yahoo," Mariani says. "Each business unit had their own IT and data teams. That's where we saw the proliferation of different tools, platforms, ways of storing the data. It was too costly and created an enormous amount of complexity."
In response, Mariani says, Yahoo's analytics team created a single data service that the business units could access via their BI tools.
"The single data service created tremendous value for the business as opposed to the distributed model," he says.
3. Data security suffers
With business units owning and managing their own data marts, the overall business loses control of the security of its data. The data tends to proliferate throughout the organization. Sensitive business data may end up sitting in spreadsheets and other BI tools on laptops and other devices, and it can become impossible for the organization to keep track of where copies of the data exist. Data origination and data lineage can become highly problematic.
"A loss of control of the data means also a loss of security," Mariani says. "You've got TDEs [Transparent Data Encryption files] sitting on laptops, core data assets are now being distributed through the organization with no way to secure it."
4. Self-service doesn’t scale
Self-service BI has arisen, in part, to achieve greater agility for lines of business. It has traditionally taken IT a long time to provide the data and insights that lines of business need to capitalize on business opportunities. Self-service BI offers the promise of giving business units access to the data and insights they need, when they need them. And in many ways, it has delivered on that promise. But as soon as you need to scale your BI efforts, the challenges become apparent.
"Self-service is faster than the queue for IT," Aziza says. "But it's small thinking for big data. It works for you and it works for your business unit. But it's just a matter of time before it's not able to scale to the need you have."
"IT is thinking long-term and thinking about the future more than a lot of business units that are just trying to solve a problem," Mariani adds. "But the business hates that centralized model because they don't want to queue up and they don't trust that IT is going to be able to deliver them what they need in a timeframe that is acceptable. At Yahoo, it took me a month to add one new dimension to look at display ads. The business can't wait a month."
Consider an apparel retailer. It can use self-service BI to perform analysis for a single brand. That may provide useful insights quickly, but the retailer knows its customers shop across multiple properties and brands. And if each business unit is working with its own definitions and metrics, scaling analytics across multiple brands becomes a Herculean task.
Once you reach this point, IT typically needs to step in to apply order to the mess, and the agility you've gained through self-service BI often takes a severe hit.
5. License costs add up
At the end of the day, self-service BI often costs more than the centralized technology it replaced. First, there's the technology cost: Each of your business units will need to buy licenses for their preferred BI tools, and you'll lose out on the economics of scale you benefit from when buying for the entire organization. Then again, allowing business users to work with their preferred tools may be worth that cost. Beyond the technology cost, though, there's also the cost in human labor.
"It definitely costs more," Mariani says. "You can't get economies of scale if you let the business roll their own. And human labor is an issue. You basically have to distribute that knowledge. You need people out in the business to understand that whole stack to do it right. That's a tough ask, especially if it's not their primary job."
