Five Really Bad Ideas in Business Intelligence

Introduction

The goal of Business Intelligence (or "BI") should be to assist business in maximizing the value of their data. Whether companies know it or not, their data is one of their most valuable assets for strategic and operational decision-making. When done right, BI has the potential to deliver highly relevant, highly targeted applications, reports, and dashboards, designed to maximize an organization's ability to gain specific, actionable knowledge from their corporate data. That means valuable tools like statistical analysis and predictive analytics, and cross-referencing data from different departments to mine for trends. Valuable stuff! When done wrong, BI has the power to create organizational, political and technical debacles of Biblical proportions. Not good.

So, how do you navigate the minefield of complex topics like Business Intelligence? What about important supporting topics like data warehousing, data governance, data stewardship, etc? How do you make sure that the decisions your organization is making are the best possible ones - data driven, insightful, and packed with value for your business?

In this edition, we'll take a slightly different approach in discussing BI best practices. Let's talk about a few of the things you really shouldn't do, if you want your Business Intelligence initiative to succeed.

Bad Idea #1: Pick the Technology First

Myth: The technology is the hard part

Many see Business Intelligence or Data Warehousing as technology problems. They get budget and approval, do lots of research, have vendors give demos and promise endless mentoring ... all in the industrious (if misguided) attempt to "buy" Business Intelligence.

Unfortunately, BI is not, and likely won't be soon, a technology platform. It doesn't come shrink-wrapped ... at any price. It can't be bought from anybody like you'd buy a copy of Windows from Best Buy. BI is a new way of thinking, best practices molded by nearly two decades of successes and failures, architectural paradigms for data and software, changes to your organization, overcoming political hurdles, and much more. The truth is that the technology is actually the (comparatively) easy part.

Here are some hurdles I'd recommend overcoming before you even talk about the technology you'll use to undergird your efforts...

• Securing executive commitment (including funding)
• Building consensus among business functions (and conforming data dimensions)
• Establishing effective data and IT governance
• Developing a change control management plan to control scope and growth
• Accurately defining requirements
• Figuring out how the system will be supported after it's built
• Etc, etc, etc

So, let me suggest that there is an order of priority to focus on when launching a BI initiative or project:

1. The Plan — What exactly are you trying to accomplish, and how does it support the organization's KPI's?
2. The People — Who is going to accomplish your plan? What roles will they play?
3. The Process — How will this awesome plan be accomplished?
4. The Technology — Now that you have all this in place, let's talk about what tech you'll use to support the plan, the people and the process.

Bad Idea #2: Acting on a Wrong View of the Data Warehouse

Myths: Everything has to be in the data warehouse. The data warehouse is a copy of my operational data. Data warehouses are fundamentally the same as transactional systems. Etc.

There are few IT-related concepts more poorly understood than the data warehouse. Dozens, if not hundreds, of disparate definitions and overloaded terms abound. The data warehouse is a highly-specialized, highly-flexible data architecture, which stores the history of dimensionally-modeled observations of business processes, optimized for online analytic processing (OLAP).

Data warehouses shouldn't contain every piece of data you can think of. They aren't copies of your operational data run on another box for the sake of performance or disparate QOS requirements. They should contain exactly what is needed to represent observations of business processes within your organization in such a way that you can make decisions based on that data. The data thrown off by these processes is collected in the data warehouse in a highly-specialized, optimized way (multi-dimensional form, not normalized; I'll write more on that in another issue), and then used for reporting actionable decision-making knowledge to end users through a targeted BI presentation layer.

Every piece of data in the warehouse should be for a specific purpose. If you can't "justify" why a piece of data is in the data warehouse, then it shouldn't be there. You "justify" this data by tracing it back to the business case / strategy / top-level KPI's of your BI initiative, a particular project, etc. If you can't do this, then you're probably dumping data in the warehouse just because you can, not because there's an actual reason to do so. And that means you run the risk of creating clutter which will seriously hinder the value of BI, and will likely cause major regrets down the road.

Bad Idea #3: Ignore the Political Landscape

Myths: Good planning, smart people, and hard work can overcome anything

There are political realities associated with building out a BI initiative. Populating a data warehouse and building cross-functional reporting applications and dashboards isn't nearly as much about the technology as it is about all the people involved. Not only will you have to work with staff across the organization, but you'll be working on processes which are the heart of the business. The same inherent value of a business process which makes it a high-value target for BI (warehousing the data thrown off by it, then reporting on it cross-functionally) is exactly what will make it politically challenging to interact with that process. This isn't a bad thing, just a reality that has to be wrangled with as you consider BI in your organization.

Here's the bottom line. Don't invest too heavily in your BI initiative until you have:

• Plentiful executive support, which is demonstrated not just verbal
• Won over some of the skeptics in your organization
• Bribed staff in key positions all over the business
• Created an air-tight, objective business case / value statement for the initiative
• Thought through a solid objection handling plan
• Expert, outside help on speed-dial for areas which aren't your strengths
• Identified quick wins you can knock out early in the project

Bad Idea #4: Leave out Critical Support Functions

Myths: Build it and they will come

Make sure you have the following four critical support functions in place. Without them, the best conceived data architecture or BI presentation layer in the world will not be successful over the long haul.

1. You need governance. How will you decide how to decide? Who will provide services? Who will write the rules? Who will enforce them? Carefully consider the people, policies, processes, permissions, etc which need to be in place for your success.
2. You need marketing. A lot of people think that "if we just have this awesome data warehouse, then everyone will fall over themselves to use it". Sounds nice, but it's not true. Build a communication plan. Figure out who will tell what to whom. When? What will it accomplish? Expect most people to be skeptics, including some of those who helped you get your funding. You're going to have selling to do.
3. You need user training. Business users won't use BI tools unless their presentation is mind-numbingly simple. If you need paragraphs of explanation on the screen, then rethink your design. Spend time in the classroom, not generating help files. You're way better off putting users in a room and training them than being overly verbose in written explanations of functionality.
4. You need IT support. Your BI architecture (especially the data warehouse) won't take care of itself. Data warehouses need more care and feeding than other apps, not less. Consider external managed support, which allows the experts to be the experts. It's unlikely ongoing support of your data warehouse qualifies as a high-value innovation for your internal IT team.

Bad Idea #5: Don't Address the Question of Data Ownership

Myth: IT owns the data

In many organizations, the default for data ownership is their IT group. If you want to realize the full potential of BI in your organization, this just won't work. Think of a plumbing analogy. Data is like water. IT infrastructure is the plumbing — faucets, pipes, water heaters, etc. It's IT's job to maintain the infrastructure, not the data — the pipes, not the water. And don't let a single business unit own data warehouse data either. The whole goal of BI is to produce cross-functional reporting capabilities. That means that the whole business has to own the data, typically managed through representative governance in the form of a governing council (also another topic for another time).

Wrapping Up

There's no question that executing a successful BI initiative is hard. It's also expensive. But the benefits of BI (when done right) dramatically outweigh the cost. Don't let the complexity and the special skills required scare you off. But don't bite off more than you can chew either. You can do it. It will be worth it. Just make sure you're properly prepared and have the right experts on hand to help.

- Jeff Block, BI Practice Director
Capstone Consulting

Back to Top