Business intelligence has become fashionable again. As ever more users want access to self-service tools, do they always know what to expect and get what they wanted? David Reed looks at how important it is to define exactly where reporting, intelligence and analytics meet.
Imagine you are working on a $125 million project that will take 286 days to reach its objective. At that critical point, something goes wrong and your precious work crashes. You’d want to know why, right?
So did Nasa when its Mars Climate Orbiter disappeared into the hostile atmosphere of the red planet. It turned out that while the space agency uses the metric system, the engineering team at Lockhead Martin had been working in imperial measures. When the orbiter was supposed to fire its rockets 96 miles/160 kilometers above Mars, it was actually at a height of just 36 miles/60 kilometers and the system didn’t work.
The story has entered legend as an example of just how critical it is to make sure the numbers within a business add up - and to ensure everybody is working to the same set of definitions. This issue is at the heart of business reporting and business intelligence, which produce the critical measures and indicators by which a company understands its performance.
It therefore came as a surprise to Ross Simson, managing director at Insight Republic and a member of the IDM Data Council, to discover that there are no fixed, agreed definitions for what these activities actually mean. “There is a blurring in the understanding of those terms,” he told DataIQ. “I started to look for a definition of business intelligence and couldn’t find one.” A round-robin to fellow members of the Council failed to shake one up, either.
Understanding what is involved matters when developing a project, since it conditions the expectation of how measures will be set and tracked. “One client wanted to get into this issue because they were starting a business transformation. They wanted to know what the operational intelligence would be, what management and business information would be delivered,” says Simson.
Demand for better intelligence is rising everywhere as companies absorb the message, promoted by the likes of Accenture, that competing on analytics delivers five-times more profit and multiple percentage point increases in efficiency. This is blurring the line between predictive analytics - which used to be the domain solely of revenue management and customer insight teams - and business intelligence - which has enterprise-wide traction.
“There is a desire to get clarity and insight, as well as to get the numbers to add up across the business. Managers want to know what to focus on in order to get better and improve activities that drive profitability,” he says. Unfortunately, creating a project to meet that appetite is not cheap - Simson estimates a £100 million turnover business could spend 12 per cent of its revenue to fix the problem. “It is hard to get a change project to deliver a benefit to business intelligence within the horizon line of most organisations. Few companies can wait three years,” he warns.
Robbie Burgess, data and technology director for marketing at Reed Business Information, knows all about the pain involved. “We have just completed a data warehouse to support the marketing department and have put in place a BI capability over the top so we can understand what is happening, whether our marketing activity is having an impact and where we can go forward,” she says.
Although wary that insight and analytics have a degree of “Emperor’s New Clothes” about them, she recognises that the appetite for them has grown. That also means a demand to evolve standard reports into more complex business intelligence. “We had people who are skilled in SQL who could generate that, but it was not scalable,” says Burgess. Users were also being kept at arm’s length from data that they could benefit from exploring in their own way.
“You have to think about the costs as reporting starts to overwhelm your skilled people. But we have moved on from reporting to business intelligence and we want to get to insight, so we are delivering a self-service management information solution,” she says. “I’d like to say making the business case for that was easy, but it had to be made on the basis of getting x more leads with y less revenue.”
Using predictive analytics to make the case for business intelligence might seem like entering a hall of mirrors. It is one of consequences of the lack of clear definitions, however, that what ought to be understood as a cost of business - like keeping the lights on or the intranet running - is viewed as an incremental addition.
Together with the costs identified by Simson, this barrier to entry is one of the reasons why Colin Rickard started Insight to Interaction, which delivers “business-intelligence-as-a-service” solutions onto mobile devices. “Until now, BI and analytics have been the domain of large enterprises, because they were the only ones who could afford it,” says Rickard.
As a result, he argues that, “definitions are only of academic interest - the really important thing to a business is what impact it has. Managers want to know what the business knows about its customers and how to increase its revenue. That comes down to having your data in the right place and making it available, as well as ensuring that those numbers add up,” he says. “They don’t care what the definition is, they only care about its effect on the business.”
Most SMEs struggle to get beyond “rear view mirror” reporting based in spreadsheets, but they increasingly want to ask more complex questions, such as what products are bought by the top quarter of customers. “That is when they want to get into managing information, which is more difficult,” says Rickard.
Insight to Interaction has already delivered a solution for Joy, which distributes premium haircare products to independent retailers across the UK. The company was looking to improve the efficiency of its field sales teams as well as to optimise revenue from its customers. The solution has been created using data feeds from its sales and CRM systems to create a knowledge store which non-technical users are now accessing via iPads, all in a company with no IT department. “To build that, we had to create a data model and identify what the client wanted to know,” he says.
Anybody who tracks the data industry will have noticed a recent increase in interest around business intelligence, driven in no small part by the consumerisation of IT and the demand by managers to have the kind of reporting on their devices which they find within consumer-oriented services. That presents a challenge for organisations of all sizes, even those with a mature data and analytics capability.
One high street retailer had undertaken business intelligence projects twice in the last four years and both had failed. Following a small-scale pilot with Blueberry Wave using a six-month extract of loyalty card data, it has now been able to establish a system shared with over 100 users and 18 external suppliers. Given the scale of the data involved - over 35 billion transactions - part of the challenge was simply to tame the information before turning it into intelligence. A similar solution has now been put in place for Sally Salon Services, also a distributor of health and beauty products for salons.
“Reporting has evolved,” says co-founder and head of systems Stephen Schneekloth. “The retailer had huge teams of analysts, but what they delivered was very static and slow. They wanted to go to self-service so those analysts could become paid-for consultants at a higher margin for other clients.”
That provided a useful - and unusual - way to make the business case. At Sally, the environment was less sophisticated, but no less business-critical. “We ran workshops with them and their suppliers to identify what they wanted, which threw up a lot of different demands, so we had to do a cost-benefit analysis to narrow it down,” says Schneekloth.
He notes that most marketing database projects ask for Excel and FastStats-based reporting in order to by-pass asking their IT department to put together a reporting system. “IT wants to deliver fixed reports with no changes and they might not understand the complexities of the data and how dynamic it can be,” he says.
One of the failings of marketing has historically been its inability (or unwillingness) to provide the board with reports that align with what other functions deliver. The argument has always been that its environment is complex and the goals constantly on the move. As a result, there has been a lot of scepticism about the financial rigour being applied by marketing.
In the new era of reporting and business intelligence, the same underlying data from the core business systems can now be extracted, sliced and diced to suit every function, while still rolling up to the same totals. However you choose to define reporting and business intelligence, what the board really cares about (metaphorically) is whether there is life on Mars.
According to Schneekloth, “I personally don’t believe there is any need for fixed definitions.” True or not, expectations are created whenever a project is described using terms that its users think they understand. Delivery may or may not then meet what was being expected. DataIQ offers the following definitions to help close that gap:
Reporting: “How many beans are in the jar?” Counts of static data items (sales, assets, customers) whose definition has been previously agreed, created to a defined and fixed timetable. Distribution to specific roles can be in any form, from spreadsheets to dashboards.
Business intelligence (BI) or Management information (MI): “Who put the beans in the jar?”, “Did the beans arrive singly or in combinations?” Correlations between counts generated for reporting and comparison of counts to pre-agreed benchmarks, targets or between different views of the counts (ie, financial value v acquisition cost). Frequency may be both fixed and variable. Distribution is typically to more senior roles, increasingly in interactive formats. Performance measurement and reward may be based (in part) on indicators in BI or MI outputs.
Predictive analytics: “Will the same number of beans be in the jar in the next period?”, “Can the same number of beans be gathered using cheaper methods? Models based on static data items using correlations identified in BI or MI to provide forecasts and scenarios. Distribution is usually to specific functions, often in real-time or ad-hoc and allowing train-of-thought querying. Yet to be widely accepted or established by senior decision makers as providing indicators with the same robustness as BI or MI.