CIOIN

Business intelligence (BI) means many things to many people but there is one thing that it is most definitely not: an oxymoron. While it isn’t a tool that will come up with new ideas on your behalf, BI, when used correctly, offers insights that can be turned into what is called actionable intelligence and used to advance the business.

That promise is driving the Indian market for BI software to grow a snappy 15 percent during 2010, according to Gartner. Despite that, India will account for less than a percent of the $10.8 billion (about Rs 49,500 crore) expected global revenues for BI in 2011. The reasons for this anomaly are aplenty.

It is an open secret that most business decisions in India are based on instinct. While this method cannot be dissed entirely, there is an important role that BI can play. Another reason for its low adoption is that the technology is only expected to reach its tipping point once the ERP, CRM and SCM systems have been stabilized. And Indian enterprises are still in the process of maturing these.

But in the meanwhile, business is getting more complex. According to estimates, Indian organizations create 2.5 quintillion bytes of data everyday. An increasing number of enterprises believe that this staggering number masks hidden opportunities. And more are willing to do what it takes—no mean feat—to create BI systems that will help them produce what they couldn’t have previously imagined.

If you are one of them, this is a great place to start. A number of Indian CIOs leading successful BI projects and experts, offer advice on the challenges you will face before, during and post a BI project. Time to parry, gear up.

Getting Ready for BI

BI is a thoroughly misunderstood term. Get to what it means and why you need it.

A true understanding of business intelligence (BI) starts with an appreciation of what it is not.

BI is not about Excel-based reporting presented snappily on a Web-based platform. In fact, it is not about reporting at all. At HyperCITY Retail, a leading player in India’s burgeoning retail segment, business leaders realized this early on. They figured that their Excel-based reports continuously fell short of their wish to empower every decision maker in the organization. This made them shift to multi-dimensional information cubes provided by BI platforms. “We realized that it’s not about the number of reports you deliver. If that were the benchmark, then all you needed was an OLTP (online transaction processing) or an MIS (management information system),” says Veneeth Purushotaman, head-technology, HyperCITY Retail.

The move would prove to be crucial, especially in the highly competitive sector HyperCITY operates in. The visibility that the BI platform lent the business enabled HyperCITY to draw insights from customers’ buying patterns and introduce targeted customer promotions that were different from all others and were something customers looked forward to. This double whammy lowered costs (that they would have otherwise incurred on mass media advertisements) and also helped the retailer substantially increase revenues from promotions.

If you’re wondering how much an impact a BI system can have on revenues, then figure this. Through insight s delivered to them by BI, HyperCITY realized that there were a lot of customers who enrolled into their loyalty programs but visit the store for several months. That idea was the seed behind a promotional offer, which saw a 21 percent response rate and translated into Rs 2 million worth of sales. “BI helps us get insights by providing the capability to look into multiple aspects of data,” says Purushotaman.

There is no end to the directions a business can take with a piece of projectdata. “But in an MIS system all you will ever know is the numbers against each department,” says Purushotaman. Moreover, reporting can only provide a business with static information and, as a result, gives the user only a descriptive ability. But with BI, users also get analytical, prescriptive, and predictive abilities.

“For example, a sales report may well describe that sales for certain products are dropping suddenly. However, when a sales director tries to explore the causes to rectify the situation, these reports might not be of much help,” says Dr. Darshan Desai, professor of management, Berkeley College, New York.

India Infoline, a large brokerage firm that deals in a range of financial services and caters to almost a million customers, also realized why MIS reports just wouldn’t cut it. CIO Sankarson Banerjee, says they were unable to figure out the reason behind poor sales during a certain period or the factors that influenced the successful performance of specific funds from the reports that they had.

It was true that they didn’t find it difficult to deal with one-dimensional questions of department-wise sales numbers and the likes with an MIS but sooner than later the queries that were being sent to the IT department began to get increasingly complex. “The reality is that you don’t always know the questions to which you are looking answers for. When it came to the harder questions we realized that we needed a more intelligent way of digging for answers,” says Banerjee. And the BI system was the answer. It provided them with much needed visibility, which, in turn, helped boost revenues.

And that anchor in the business is also why BI can’t be treated as a traditional app development project. “BI requires a greater understanding of the business and its drivers,” explains Cindi Howson, founder, BI Scorecard and author of Successful Business Intelligence: Secrets to Making BI a Killer App.

Also Read: Interview with Olivia Rud, Founder and President, OliviaGroup

Senior management tends to think of BI as a magic wand

There are generally two ways business leaders are exposed to BI: Through a vendor demonstration that’s organized by the CIO or via seminars. In either case, at the time of their presentations, most BI vendors showcase multiple features, which business leaders generally fall for. What they don’t realize, however, is the shallowness of the data that the demo is conducted on.

“If vendor presentations are being moderated by you then you have some amount of control on what is being showcased in front of senior management,” says Purushotaman. “But if they have come across the idea at a seminar then you need to inform them of the ground realities and how an implementation at your organization would have different results from what they’ve seen. In our case, though the vendor’s presentation was moderated by us, we had to do multiple sessions with the business to come to a consensus regarding the scope of the project.”

Rationalizing the scope of a BI project and presenting a realistic picture to senior management at the very outset is important. According to Purushotaman, the best way of doing this is by producing a proof of concept. “It was a challenging task and takes a lot of effort. The whole procedure has to repeated for the actual implementation, but it helps bring expectations closer to reality,” he says.

While implementing HyperCITY’s BI solution, Purushotaman also realized that few vendors initially talk of the components that go into a BI stack, which is, in fact, very high. A BI tool bought off the market is just an application layer that sits on top of a proper data layer. This in turn is backed by proper data warehousing, an ETL layer, and the hardware that will support it. None of this comes cheap. “So we looked at the stack in its entirety while presenting it to senior management because we knew that a piecemeal approach was unlikely to succeed,” says Purushotaman.

In order to keep expectations real, it is also necessary to present the advantages of BI in broad-based overviews. “Much of the promise of BI is based on great examples from other organizations,” says Banerjee. “We clearly informed business leaders at India Infoline that results are not just a consequence of the quality of the tool but also depend largely on the users who handle it.”

There’s another lesson Banerjee learnt early on: The more the number of variables, the greater the number of insights. “If your organization functions only on six to seven streams of incoming data then there is very little scope for a BI tool to bring out exceptional insight. But a stock broking firm like ours encounters hundreds of thousands of different streams of data and there is a lot of exciting opportunities to extract something interesting,” he says.

A BI project without a senior sponsor is like a ship without a captain.

By its very nature, BI is not a CIO’s baby. “A CIO who simply acts as a gatekeeper to the data is not a good BI sponsor,” says Howson. It needs someone at the very top, ideally the CEO, to drive home the enterprisewide ramifications of a BI project.

A board-level sponsor for BI can loosely be called the ‘cheerleader’ whose job is to boost the project forward. Without such a sponsor it is likely that business interest will fade quickly. Angel Broking, a stock-broking and wealth management company, for example, was keen to protect its customers by implementing a system that flagged fraud in real time and allowed business users to stop it. A BI project could accomplish that but it called for large investments, changes to core business processes, and the creation of new roles. “So the sponsor had to be someone at the board level who was suitably empowered to take calls across departments and cut across divisional boundaries,” says Dr. Mudit Kulshreshtha, who until recently was ED for enterprise applications and business intelligence and analytics at Angel Broking.

“In our case,” says Purushotaman, “the CEO chose to be the strategic head, so we needed someone a level below him to handle the project at a tactical level. So the selection of a sponsor depended on the departments that we planned to initially target with the BI project and was based on the scope of the project that we had chalked out. While a CIO can play the role of catalyst and implementer, the sponsor has to drive it.”

Project Time

Business units are surprisingly less willing to part with information than you think—and not always intentionally.

A fundamental aim of all BI projects is to bring more clarity about the business to its leaders. At the Braj Binani Group, bottom line pressures that the company, like all cement players faced, led to a realization that growth would come from increased transparency, accountability, and the ability to take better informed decisions faster. But it also meant that a lot of people would need to give up whatever they might be withholding. And that, Rajesh Mohan, joint president-IT and systems, learnt, was easier said than done.

The information platform at Binani provides users with a basis to investigate the business process inefficiencies and execute quick and effective corrections. “These corrections affect both the top line and the bottom line quite directly in some of our businesses,” says Mohan. However, as the head of the group’s business analytics initiative, Mohan realized that believing that IT had access to all of the company’s data was a misconception. “Despite a top-down mandate, we still needed to get buy-in from users who knowingly or unknowingly put away information in silos. We only had access to information that was shared with us and was part of structured systems,” he says.

A case in point is Mohan’s experience with Binani’s internal marketing department. “They believed that they had benchmark standards for the industry based on what the CMA (Cement Manufacturers Association) provides,” says Mohan. But this did not even include information from major players in the industry who were not part of the association, which skewed their data. “That apart, this information was not even made part of SAP because the moment they put it in, their performance would be measured against those benchmarks,” he says.

At his old job at Angel Broking, Kulshreshtha faced a similar hurdle with his HR department. During a BI implementation, the HR department had a problem submitting all the data points the new system required because they did not have accounting skills to do the job. “Hence bringing out a multi-dimensional monthly balance sheet which included information from HR was difficult to produce and even when it was brought out it was not very accurate,” he says. Similar cases can come up with other departments and need a sponsor to arrange for training.

But, at the same time, there is a danger in swinging to other end of the information sharing pendulum. Care must be taken as to what information is shared with shareholders—especially regulators and auditors. “The business needs to be ensured that information being fed into the system is meant to work for them and that they will continue to remain in control. Losing control is an immense fear,” says Mohan.

If you think you have staff that’s qualified to handle BI, think again.

A BI project is a potpourri of sorts. It involves a large number of factors that the IT team could be faced with for the first time. During a BI implementation there is a lot of unstructured data that comes to the fore which technology folk are not used to dealing with. “In our experience with the marketing department, our team wasn’t trained to understand the importance and relevance of data that was either intentionally or unintentionally being withheld from them,” says Mohan. His advice to CIOs who think that BI is a technology project: Be warned, you could face large skill-set gaps.

A CIO would be ill-advised to believe that outsourcing eases all skill-set worries. “You will need the right skill sets within the organization to handle a BI implementation,” says Purushotaman. If nothing else, a CIO will need skilled database architects, ETL architects, and business analysts. “Despite using an outsourced model, we still had a team that knew exactly what was required by the business and had the project management skills to keep a tab on the system integrator’s progress,” he says.

Mohan seconds that thought. “I know that the actual project will be run by a partner since I don’t have the in-house resources. But I still needed people within who will manage my partners and map the expectations of senior management and business users. Since BI is not just about technology, I am hiring business consultants with a track record of being transformational, while my technology team helps with the backend,” he says.

In either case, in-house or outsourced, it is advisable to have a dedicated internal team working on the project. “As BI is a resource-intensive project, in the case of an established organization, a wide variety of legacy systems require the integration of multiple devices. So I tried to keep my team constant without too much shuffling around,” says Kulshreshtha. “Program management, an analytical know-how, domain expertise, and business process mapping are key skills to have in your team,” he says.

An understanding of the business is a key requirement for someone to be a part of a BI project. “Staff a BI team with the right people, put a business-IT hybrid in a leadership position, and then ensure that a BI team comprises both business and IT resources,” says BI Scorecard’s Howson.

The insights you seek to reveal are hidden within reams of data that are at best unclean and at worst missing.

Another roadblock is dirty data: Data that is incomplete, inaccurate, incorrect, outdated, redundant or misleading. “A lot of money is spent on a correctly modeled data warehouse or well-implemented BI will be totally pointless,” says Desai.

As a first step toward this, Purushotaman gathered folk from the various business units and asked them for the rules by which they would like data to be inputted into the system. This prompted them to come up with a detailed set of rules. “Then we took their own data and it failed to meet the standard. That opened their eyes,” he says. Getting business units to acknowledge that there is a problem with unclean data is a primary requirement.

Mohan was faced with a different challenge at Binani. “Consultants came in and told us that everything—right from our transaction engine to our CRM systems—lacked the necessary fields and data-points; essentially the whole thing needed to be changed,” says Mohan.

Mohan scoffed at the idea because he knew the importance of protecting his existing investments. “Instead, I took a top-down approach to fixing data. If transaction systems didn’t provide the view that senior management wanted then we initiated smaller projects to add the required functionalities,” he says. And as the process kept maturing they continued to clean data and add new features.

According to Berkeley College’ Desai, in order to enhance the quality of data, it is important to discern which data is wheat and which is chafe; what’s worth using for actions, profiling and predictions and what needs to be discarded. “Many times, being data-driven, companies require a shift in culture, and, for that purpose everyone should be on the same page about the relevance of data,” says Desai. Hence, it is important to create a policy of data governance. “Developing data governance policy is not a very easy task; it involves clear, step-by-step definition of the whole process of data cleaning; and it is important to involve both IT and business users in this process,” she says.

The more legacy, the more complex an organization’s data cleaning issues. Getting the design of a data warehouse right is very important as it helps be clear about the variables on which questions business is going to ask will be based on. “At the point when we invested in BI, we also invested in the clean-up tools. We defined authoritative data stores and all subsequent information is channeled into it. While it was not possible to clean up the existing database because many other applications accessed it. We ran a constant clean up so that new DB had information within five minutes of it entering the old DB,” says Banerjee.

The key is to identify major areas of data errors, clean them to exacting standards, and fix processes at the source to avoid further occurrences. Even after the go-live there may be issues. “However, it is important to remember that good quality data is not an ideal; good quality data won’t always ensure good quality decisions,” says Desai.

POST PROJECT

The revelations you’ve delivered are no longer what the business wants.

If you are a non-believer, it would be advisable to join a flock of the faithful. “Pray there are no changes,” says Purushotaman with a chuckle.  “Because if changes do present themselves after an implementation, then no matter how difficult it may seem, you have to be ready to incorporate them.” 

One way to avoid this is being thorough upfront. “At HyperCITY, we went ahead with the entire set of information and did not leave out any data points in the rush to finish the project,” says Purushotaman. “But if an entirely new set of data points comes into the picture due to added applications, then it needs to be pushed to the concerned people. A fresh ETL layer has to be created to include this,” he says.

This problem reinforces the needs to own in-house skills sets. “If your functional team understands the business then it ensures that whatever requirement the business has overlooked, it can be brought to light before it’s too late,” says Purshothaman.

The dynamic needs of business also ensures that a big-bang approach to BI also doesn’t work. “In order to cater to ever-changing business requirements, the only thing to do is to avoid mega-releases and break a BI project down into four or five mini-releases spread across the year,” says Bhavish Sood, research director, Gartner India. He suggests taking a cue from the telecom industry, which breaks down their service offerings into smaller bits and makes them available at regular intervals. “That is more or less the direction which the IT industry will also have to take,” say Sood. This has become especially pertinent post the slowdown where no CIO can ask for a year or two to bring out a new solution. “So prioritizing the needs of each unit and taking it one department at a time is the way to go,” says Sood.

You will waste a lot of intelligence to prove nonsense if you don’t have a dedicated analyst team.

Mastering the art of analytics is no walk in the park. No matter how fancy the user interface is, the tools, for a non-tech-savvy user, are complex. “And to draw the intelligence out of business still requires human judgment and imagination,” says Banerjee. Purushotaman agrees. “Analytics is the next step of BI and it’s not inherently a part of the tools out there in the market. So you need people who understand data and are capable of using these tools to interpret it.” 

Once the BI is in place, the IT team should no longer cater to report requests. Business analysts and statisticians working on the data should be the ones in charge of actionable intelligence. “And when they are backed by senior management, like in our case, it ensures that their insights are put to fruitful use,” says Purushotaman. 

One of the ways in which HyperCITY utilized such insight was when food purchases from one of its store dipped below its all-store average. Based on that observation, HyperCITY targeted customers (whose food shopping bill was below the average) with incentives. The move garnered a 31 percent response and resulted in incremental sales of over Rs 7 lakh for the store.

“After investing in the analytics layer, we now have an outsourcing partner who comes out with insights, works with the business and executes them. It is their core competence and it makes life easier for the organization,” he says. And when asked if third-party information is taken more seriously, Purushotaman makes no effort to deny the fact. “In the Indian context, you essentially pay outsourcers to say what you want to hear but it is thought to be more reliable,” he says.

 At HyperCITY, every team is clearly told that they need to utilize the expertise delivered by the analyst team and come out with deliverables that are mapped to an individual’s KRA. This ensures that inputs are not taken lightly. “A percentage of an employee’s KRA is focused on executing promotions that you drive based on insights derived from the system,” says Purushotaman .

While Banerjee agrees with the need for a separate team, he also feels that power users are best left to use the tool themselves. “Investment bankers or financial analysts, who are comfortable with technology and perform analysis on their own, are equipped with the tools while others are helped by a competency team that shows them the way forward,” he says. 

“By using both business and technical experts working together in an analyst team, it can be possible to see outcomes sooner and better,” says Desai.

Success, especially in the case of BI, is illusive and largely immeasurable.

Even once you have made reasonable progress with a BI implementation, experts suggest holding your breath for spectacular results.  If nothing else, that could lead to asphyxiation. In spite of having the best BI tool it is possible that while you are looking for patterns in the data, a particular pattern just might not exist for you to figure out. “In the risk metrics, for example, it is very possible that through your analysis you are unable to catch a fraud because, to begin with, there was no instance of fraud to catch,” says Banerjee.

“Attributing say a certain increase in sales to the insights from a BI project is possible ex post facto. But even this recognition comes after due endorsement by respective business units,” says Banerjee. And BI does not always create something new. More often than not it lets you do, what you have already been doing, better. This again adds to the intangibility of its success. “So we got the business to back our claims. If the business units acknowledge the contribution of BI in their performance improvements then you have a very strong case,” says Purushotaman .

“One best practice that I recommend is the use of the balanced scorecard performance management system to design a set of strategically aligned measures that help to gauge performance,” says Howson.

Desai suggests another alternative in what Dr. James Thomann proposed in his webinar titled Determining the Success of Your BI Initiative. According to Dr. Thomann, there are three types of measures of BI success: Political, technical, and business-level.  The political metrics show that the application usage is healthy with measures like number of users and number of requests for enhancements. The technical metrics show how well the application itself is performing measured in response time, down time, etcetera. These metrics are relatively easy to use, but they do not tell the whole story. The more important one how ever is the business-level measure that shows the business success of the implementation. These measures are generally some form of ROI. “Since BI is kind of an enabler, contribution to revenue and even to indirect expenses will be imprecise estimates. However, reasoned allocations of these streams can give a good idea of the contribution of BI to the bottom line,” says Desai.