Even though cloud-based business intelligence has been around for nearly a decade, a recent trend is driving renewed interest: Companies are generating and storing more data in the cloud.
"What I think will happen is people will move the analytics app closer to the data," says Joao Tapadinhas, a Gartner analyst. "As more data sources move to the cloud, it makes more sense to also adopt cloud BI solutions because that's where the data is. It's easier to connect to cloud data using a cloud solution."
In each of the last four years, around 30% of respondents to a Gartner survey said they'd run their mission-critical BI in the cloud. This year, however, nearly half -- 45% -- said they would adopt cloud BI.
Researchers at Gartner say that 2014 may be the tipping point for cloud BI. In each of the last four years, around 30% of respondents to a Gartner survey said they'd run their mission-critical BI in the cloud. This year, however, nearly half -- 45% -- said they would adopt cloud BI.
Historically, cloud BI products have been most appealing to smaller businesses, in part because those are less likely to have an IT department that can manage an on-premises product. However, analysts are starting to see larger companies adopting cloud BI, typically starting with individual groups or departments.
Shifting data analytics to the cloud doesn't come without its challenges, though. For example, it's unlikely that all corporate data will move to the cloud, particularly in larger enterprises. That means many businesses will have to map data from both cloud and on-premises sources to the BI software, whether that software itself is on-premises or in the cloud. Also, bandwidth constraints may slow down data transfers and can lead to increased costs, if a business must upgrade its connectivity to improve data transfer.
Nevertheless, some businesses have already adopted cloud BI services, analysts report anecdotally, though specific figures aren't available. Many companies that have made the move say that the benefits -- including fast time to market, no need to maintain on-premises software and simplicity of use -- outweigh any downsides.
Mixing up data sources
Take Millennial Media, which sells a mobile advertising platform. It needed to pull together data from disparate sources, both on site and in the cloud.
Around two and a half years ago, Bob Hammond, CTO for Millennial, began looking into BI as a way to marry data from Salesforce with transactional and financial information from in-house systems and then let decision makers at the company visualize it.
"No human I know of can . . . make business decisions based on data that hasn't been brought together into a single source," he says. The company needed BI, he says, because "we weren't able to take data from multiple systems and connect that data logically and view that data in a UI so that we could understand what was going on."
Making the move
Most traditional BI vendors will start shifting toward the cloud, if they haven't already, says Carsten Bange, founder and CEO of Business Application Research Center, an analyst firm that specializes in enterprise software. "How these will look is the big question," he says.
Some of the traditional vendors are likely to offer essentially hosted versions of their software, rather than full multi-tenant SaaS, he says. A multi-tenant SaaS app serves multiple customers from the same server. In a hosted scenario, one instance of the software serves only one customer.
A hosted environment isn't necessarily bad for end users but it's not cost effective for the vendor. The vendors are likely to go through the pain of rewriting their apps in order to deliver them as true SaaS, meaning at some point there will have to be a transition to a new service, Bange says. That could present challenges for users.
-- Nancy Gohring
He also wanted to let more people in the organization, like data analysts, assemble reports, rather than limiting report-making to technologists who know how to code and interact with back-end databases. Plus, he needed a system that was flexible so the software would be easy to maintain and it would be easy to create new use cases.
Hammond eliminated on-premises BI software options in part because he didn't want to incur the costs associated with managing and maintaining it. Time to market was also important.
Millennial ended up choosing Good Data's cloud BI offering and had its initial project in place in about three months. Subsequent projects have taken closer to a month to get up and running, Hammond says.
Sending on-premises data to Good didn't turn out to be much of a problem for Millennial. Each day the company generates around 10TB of raw data but transfers only around 18MB of compressed data to Good. "We do all the transformation of raw data into only the specific data we want in our systems before we transfer it into the cloud," he says.
Not all businesses do such a great job of managing that data transfer, though. "What we tend to see is it's rather difficult to keep the amount of data moving between the database and the analytics tool small," says Gartner's Tapadinhas. In other words, keeping data transfers small is important in cloud BI to manage both costs and upload/download bandwidth issues.
At Millennial, engineers handle the job of extracting data from the various sources and uploading it to Good Data. In addition, two data analysts have now created 500 reports. Around 40 additional people at Millennial have access to those reports and can combine them, drill down into them and create portfolios of reports to share.
Building tiers of users, each with different permissions, allows more people in the organization to work with the data -- but safely, Hammond says. That means business executives, who aren't necessarily trained to be data scientists, have some latitude to combine and rework reports but are less likely to make mistakes because they don't have the permission to, for instance, pull in new data from a back-end database, he says.
Speed and flexibility drive cloud adoption
Athenahealth, a provider of Web-based software and services to medical practices, had most of the data it wanted to analyze in one place internally. About a year ago, the company set out to find a better way to track the hundreds of customer implementations it might be working on at any given time, says Adam Weinstein, director of core analytics at Athenahealth.
Because we have a cloud-based platform, we have real-time access to see what's going on. Adam Weinstein, director of core analytics, Athenahealth
"Because we have a cloud-based platform, we have real-time access to see what's going on," he says. The biggest challenge: "Taking the data we have about what our clients are doing and how they're progressing in the implementation process and turning that into what we call a nerve center, or a way we can actively monitor exceptions to the process."
Athenahealth wanted a system that would collect information about every point in the implementation life cycle in order to easily find problem areas. For instance, clients route their fax machines to the Athenahealth system. If no faxes are coming in for a given customer, it could mean the customer hasn't yet rerouted the fax number. Or, for a long-time customer, if the percentage of fax information coming in increases relative to electronic information, that could mean someone mistakenly changed a setting.
When Athenahealth started looking for a BI product that could meet its needs, it had a few additional requirements. The vendor "had to be able to move quickly because we had a fairly strict timeline, in the two- to three-month time frame, to deliver on this project," Weinstein says.
Also, the company wanted a product that would meet analytics needs going forward, too. "We wanted to invest in more of a platform, not just a one-time solution," he says.
Distribution Market Advantage (DMA) first implemented cloud-based PivotLink BI in 2005. However, the setup had a few shortcomings so the DMA recently switched providers, going with a service from Manthan.
DMA is owned by nine regional food service distributors. It offers access to data as a service to the restaurant customers of its distributor owners. Those restaurants can run reports about how much product -- say, French fries -- they have purchased and how much is stored in which warehouse. Users can customize the reports to show the data by month, for example, or by individual restaurant.
Over the years, PivotLink began to pivot its mission a bit, and now focuses primarily on retail analytics. Since DMA is a business-to-business operation, Jim Szatkowski, vice president of technology and data services for DMA, began to wonder if PivotLink would continue to invest in business-to-business tools.
Plus, PivotLink lacked a couple of capabilities that he was looking for. Users couldn't set up workflow rules that might alert them about inventory shortages or other problems that need to be addressed quickly. "We don't want to have people sifting through data to find things that are actionable," he says. "We want the system to find actionable things and bring them to users."
In addition, PivotLink didn't offer data visualization. DMA users were creating what look like Excel spreadsheets.
DMA began to look for a new vendor and initially researched 12 companies, including Salient, Birst, QlikView, Good Data, Tableau, MicroStrategy and Pentaho. It looked at how they handled security, whether they handled the volume of data that DMA required and how easy their software was to use, how they handled data transfer and what their dashboards looked like.
DMA's new system is now live, with 55 medium-size food chain restaurants using it.
DMA trained 175 end users on the system, primarily via GoToMeeting with a few in-person trainings, says Szatkowski.
Similar to its setup with PivotLink, DMA continues to funnel data to Manthan via iTradeNetwork, a company that offers a number of services including an application that DMA restaurants use to place orders from the distributors. ITradeNetwork also manages a data warehouse for the DMA, normalizing data that comes from distributors as well as from the ordering app. From there, the data is sent to Manthan.
With so many people across different companies needing access to data, cloud BI offers a big improvement over the way DMA used to handle analytics. Before the PivotLink implementation, a restaurant would call its distributor and ask for a report. The distributor would call the one person at the DMA responsible for creating reports, who used Access to build reports. "It was horribly inefficient," Szatkowski says.
-- Nancy Gohring
Weinstein quickly found that some of the large, traditional BI vendors were not going to be able to roll out Athenahealth's initial project quickly enough. In addition, some were too complicated to use, potentially limiting future projects. Athenahealth considered products from both IBM and Oracle, and then moved on to the cloud BI offerings, ultimately choosing Birst.
Athenahealth didn't run into problems with having most of its data stored on-premises and not in a cloud environment. The company has over 50,000 provider clients and tracks more than 100 metrics about each one every day, Weinstein says. That data is pulled from an internal data center into a separate internal data warehouse. From there, the relevant data is uploaded to Birst.
The data uploads happen automatically, several times each day, as part of a process that the company built using tools and scripts, some of which were provided by Birst, he says. "It doesn't keep me up at night," Weinstein says of the process. He has to intervene only if there's an error. "But that is part of our standard monitoring and would be expected as part of a complex data warehouse environment."
Millennial Media, Athenahealth and DMA (see "Early adopter") all say that using a cloud BI service meets their needs. But there are a few roadblocks that companies should look out for when considering cloud BI.
One is "cloud washing." Some vendors say they offer a cloud BI product but in fact may still require software that runs on users' computers or may offer only cloud storage, says Gartner's Tapadinhas. In that case, users may not get all the benefits of a true cloud offering, like offloading software maintenance.
A cloud BI service might also not be as flexible as an on-premises offering. "Although they are quick to deploy, in some cases cloud BI solutions don't offer enough customizations or at least not as much as we have now on-premises," Tapadinhas says.
On-premises products might also offer more possibilities for integration with third party-products, he says. Good Data, for one, has made some strides to allow third-party tools to access data repositories stored with Good, but even its openness is limited, he says.
Plus, traditional BI tools typically have a broader feature set and may make a better option depending on what a company is trying to achieve, says Carsten Bange, founder and CEO of Business Application Research Center, an analyst firm that specializes in enterprise software.
There's also the chance that, like any cloud offering, a particular cloud BI service might be slow. "There are other issues, like performance and latency of cloud solutions," Tapadinhas says.