Q&A: Simplifying User Reporting Still Key BI Goal
A faster, more powerful system is not the only way to make reporting faster and more efficient. Users need better reporting tools, too.
Discussions about speed and performance in business intelligence applications often emphasize the power of the system itself, but speed and efficiency in reporting can also be achieved by giving users better reporting tools, argues Sanjay Bhatia, CEO of Izenda, a company that offers integrated self-service reporting for the .NET platform. (As an OEM solution, it is integrated into hundreds of .NET applications with over 25,000 end users, including the U.S. Navy, WellPoint, BHP Billiton, JC Penny,the EPA, Volvo, and Gateway.)
"How fast a report runs is only one element of performance," Bhatia points out. "Speed is also about simplicity." With simpler tools, he says, users can do true self-service reporting without format training, making their own simple changes to reports as business conditions alter. IT is often shocked to find out how capable users really are if you just get them started with a good set of reports and some initial training, Bhatia says.
BI This Week: There's considerable interest in the BI community in extending more power to those users who don't have deep technical skills. What does the term "self-service reporting" mean to you, and how do you view this move toward better self-service BI?
Sanjay Bhatia: True self-service should mean that the user can do anything and everything without any IT involvement. All too often, users can view a report but it has to go back to IT or a consultant in order to make even a minor change. Instead, users should be able to make simple changes quickly and without formal training.
Technologists and engineering teams spend a lot of time and energy creating systems that perform well. Unfortunately, they often end up creating systems that are so rigid that quickly adapting them to changing business conditions is nearly impossible. After all, how fast a report runs is only one element of performance. A decade ago, a 5K server came with about 32MB of memory. Today, a 5K server comes with 32GB of memory. Those gigabytes of memory mean that the vast majority of queries can run in seconds if indexing is done properly.
So better hardware has addressed speed and performance issues in many cases?
Yes. With that element of performance taking care of itself in many situations, the focus can turn to making the customization process many times faster. Rather than waiting for IT, true self-service means that users can not only see a report quickly, they can also change it as they like.
Speed is also about simplicity. It means not having to go to a training class or download a 100MB tool and configure your VPN if all you want to do is add a field to a report. That's the kind of thing people often want to do in a hundred different ways over time. True self-service BI means no one has to manually change the model, download software, or write SQL code when all you need most of the time is a simple variation.
The world of blogs and content management has already gone through this evolution. Once upon a time, the only way to fix a spelling error or post new content was to make a request to IT and wait forever. Today, content management systems and blogging software let you add content and make changes instantly using a browser. You can also start up a new blog in minutes. Self-service BI needs to work the same way.
What are some of the problems you see with the role of a BI consultant, and what's an alternate model?
BI consultants deliver remarkable services that can create enormous value for an organization. The problem arises when they want to maintain a long-term billable relationship after the bulk of their work is done. Companies can easily enter into situations where what's best for the customer is not what's best for the consultant. Sometimes, that means consultants are recommending approaches that take excess time without providing much benefit.
Many BI implementations play right into this equation because they generate a lot of professional services over time. In some cases, over 80 percent of the cost of a BI implementation is services and training. The software ends up being a Trojan horse for an army of BI consultants that seem to linger forever.
If consultants are no longer on the scene, IT or engineering departments get stuck with the burden of constantly creating new reports.
The alternative, and the one that we espouse, is to let the BI experts do the initial training and setup for the first few reports, then let the users take over from there. The majority of the time, the user wants a very simple change -- adding a field or changing the way things are sorted. Today's BI technology puts too many barriers in place that make it impossible for the user to make those simple changes themselves. This creates job security for the reporting department and revenue for the consulting firm but wastes the user's time.
For self-service reporting to be possible, it has to work using a browser and it has to be simple. Since most users aren't allowed to install software and don't have VPN access to the database directly, they generally can't use traditional BI tools. Additionally, the skills required should be attainable by watching a few five-minute videos. They shouldn't be anything that requires knowledge of programming or database administration.
There's been much discussion about ad hoc reporting, but few business users seem able to actually build a report themselves from scratch with today's tools. Why is that?
In my experience, actually, most users can build reports from scratch, but it should never be the first step. Initially, users don't have much of a comfort level with the data model or the tool. The best way to get them started is to create the initial five to ten reports for them and then show them how to customize these reports. Again, a five-minute video that they can return to later if needed is a great way to teach. If it takes more than five minutes to explain how to make a simple change to a report, your system is way too hard to use.
Once users are comfortable making the small changes they need on a day-to-day basis, they naturally start creating their own reports from scratch. They now both understand the tool and know where the data is. We've seen many cases where a dozen initial reports turn into hundreds of reports over the course of a year. In fact, IT is often shocked to find out how capable users really are if you just get them started with a good set of reports and some initial training.
We're hearing a lot about "the cloud" these days. Where do you see BI headed in relation to cloud computing?
I firmly believe that the cloud is simply another choice, and that the customer should be allowed to make that choice, and also to change his or her mind later.
By that I mean that a great way to deploy applications would be to let people simply sign up for a free trial, get started, and then move the application back inside the firewall once it gets significant adoption and needs more controls. At that point, you could also decide that you don't want the burden of maintaining an application anymore and move it to the cloud. Either way, it should be transparent to users, and the migration should be fast and virtually risk-free.
The cloud simply extends the Internet's natural strengths into your server environment. The Internet was originally designed to survive nuclear war. Your application should really be able to survive far less drastic things such as everyone logging in on the last day of the quarter. A good cloud platform makes scalability and performance automatic, so you don't have to spend a ton of money on data centers or systems architects. Instead, you can roll out your pilot application in a matter of hours.
Specific to BI, how might applications in the cloud change things?
For BI, the cloud is about speed, scalability, and broad access. You can instantly light up a new BI project without having to requisition servers and wait for them to be shipped and configured. Additionally, if you need an amazingly powerful server just a few times a year, you can rent a multi-million-dollar infrastructure at an hourly rate. Although people will still spend money on powerful hardware and software infrastructure for BI, they'll eventually only do so after they've done a pilot at very low cost and very little effort.
You define Izenda as a market leader in application-level BI on the Microsoft platform. How does what Izenda do fit into what we've talked about here?
Most users really don't know what a key performance indicator is or why you would need a data warehouse, nor do they want to learn. What they do know intimately is the line-of-business application that they use every day to do their jobs. We believe that BI should be a core part of primary applications rather than a separate standalone system. Just as the Internet is now part of nearly every new application at a fundamental level, our technology lets you integrate BI in the same way.
We think the best way to deploy BI is to make it a part of the applications that people use every day, and we've created a platform that lets you do that. Once BI simply becomes an application capability rather than a complex new IT initiative that takes forever, people get comfortable with it very quickly.