Business Intelligence Best Practices -

Collaboration. Communication. Community.

 E-mail to friend
The Shift to On-Demand Business Intelligence

by Ken Rudin
By providing the solution as a service delivered over the Web, the on-demand approach promised to remove the headaches associated with purchasing, deploying, maintaining, and upgrading traditional enterprise software.

Though initially these were just promises, the industry has matured over the past several years to the point where on-demand applications are indeed delivering on their initial claims and are now widely used by companies of all sizes and in all industries.

Curiously, the on-demand industry has been almost entirely focused on transactional applications. We see successful on-demand solutions for managing sales, processing accounting transactions, providing customer service, managing supply chain, executing marketing campaigns, and managing human resources. The equivalent offerings for business intelligence don’t exist.

However, this situation is changing as people recognize the benefits that the on-demand approach delivers for transactional applications. Enterprises are now looking for the same advantages to be applied to business intelligence solutions.

Why On-Demand Business Intelligence?

Where is this growing interest for on-demand business intelligence coming from? A look at the BI market today points to two main factors driving customer demand: complexity and cost.

Traditional BI solutions are complicated, requiring people with specialized skills to implement and manage them. You need to deploy an ETL engine, build a data warehouse, put a data cleansing solution in place, and implement an OLAP engine. The prospect of deploying a BI solution can seem overwhelming to many. A consulting client of mine once equated deployment with trying to build and manage your own nuclear reactor. Many companies just don’t have the in-house skills to build and maintain a traditional BI solution.

Because of this complexity, BI solutions have historically been expensive. First, there are many hardware and software components you need to buy before you have a complete solution that can pull together data from multiple sources to give you visibility into what’s going on in your part of the company. Second, once you have all the pieces, you incur significant additional costs over the next 6–12 months as the solution is implemented.

Note that new open-source BI solutions don’t solve these issues. Though open source can lower or remove the initial software license cost, the high cost of implementation remains. The real barrier for most companies is not the cost but the lack of specialized skills in-house to manage the solution. Making the solution open source doesn’t make it easier to implement or manage— it’s just a free version of the “nuclear reactor” we referred to.

What options are left for companies without large IT resources and deep IT pockets? Most end up in what I call “Excel hell.” Gaining real insight usually requires data from more than one system. They export data from their various systems, and then cut and paste the data into Excel and do their best to manage their business that way. Essentially, by necessity they end up trying to use Excel as a data warehouse. The process is highly manual, which means it’s also highly resource intensive and error prone.

Since there’s no integrated system in place to create reports and analyses, it typically takes a long time to generate new report types. Elements such as customer names often differ between systems. Furthermore, there’s no way to ensure that the data in a spreadsheet is up to date, so data quality is also a significant issue. Finally, there are no controls over who gets to see what data. These are the types of issues that are driving interest in on-demand business intelligence solutions.

The On-Demand Approach to Business Intelligence

Many people mistakenly think the essential difference between on-demand software and traditional on-premise software relates to where the software is located and how it’s paid for. It’s certainly true that on-demand software is managed outside the customer’s four walls instead of inside its data center. It’s also true that companies usually pay for on-demand software via a subscription pricing model instead of purchasing the software. Nonetheless, to conclude that these make up the essence of the on-demand paradigm is to miss the point.

Both hosting and subscription pricing have advantages, but they’re incremental advantages rather than breakthroughs. For example, taking traditional software and moving it out of a customer’s data center and instead placing it in a vendor-managed data center doesn’t address the software complexity issues discussed. Just having the same software and servers in a remote location managed by someone else doesn’t mean that the software magically becomes simpler to implement or customize. It doesn’t make it any faster to deploy, it doesn’t make it easier to make changes after the initial implementation is complete, and it doesn’t make it any easier to use.

Charging for the solution on a pay-as-you-go basis means you don’t have to pay a large sum for your solution up-front and then hope you receive the promised benefits in the near future. Instead, you pay incrementally as you receive value. Though changing the pricing model on existing traditional software may mean you pay less initially, it doesn’t mean you’ll pay less overall. In the same way that paying for a car in monthly installments doesn’t make the car any less expensive overall, paying for traditional software on a subscription basis doesn’t significantly lower the overall price of the software.

If you don’t actually redesign the software to be simpler and less costly to deploy and manage, it will still be as costly to manage as before (regardless of who physically manages it). This means that one way or another you’ll eventually have to pay these costs.

On-Demand Is a Mindset, Not a Feature Set

On-demand is not just an alternative deployment or payment option for an otherwise largely identical solution. The goal of an on-demand solution should never be to replicate the same set of features that traditional software has and then just deliver it as a service. Designing new solutions to mimic old solutions rarely provides real value.

For example, imagine if the creators of the first automobiles had designed them to look like the horses they were trying to replace. You’d still get saddle sores, and you’d still get wet when it rained.

To truly provide meaningful new solutions that address long-standing market issues, you need a different approach to thinking about the problem. The on-demand approach is much less about focusing on a particular feature set and much more about focusing on a mindset. It’s a mindset that is service-oriented rather than product-oriented, with security, simplicity, and end-user experience as the critical focal points. Since on-demand service providers are responsible for managing customer data, security has the highest priority.

Providing solutions that are simple to buy, simple to set up, and simple to use is paramount, holding promise for customers who previously did not find it practical to implement traditional BI solutions. Following this mindset, hosting and subscription pricing no longer are the defining factors of on-demand applications; they are two components of many for making the application as simple as possible to buy, set up, and use.

Simple to Buy

A key aspect of on-demand BI solutions is to make them simple to buy, considering both price and the purchasing process (see Table 1). Traditional solutions can cost hundreds of thousands (or even millions) of dollars. This introduces significant risk, since an enterprise must invest a significant amount of money long before seeing the value returned.

The high cost also acts as a barrier for many companies that cannot afford the up-front investment. Being required to spend $150,000 to $250,000 (or more) before you can use your solution is at odds with the on-demand “simple-to-buy” focus.

BI Purchase Traditional BI Solutions BI Software as a Service
Total cost of ownership (annual average) Hundreds of thousands of dollars Tens of thousands of dollars
Contract terms Pay up-front Pay as you use
Proof of concept/trial Small portion of eventual application Full application

Table 1. The on-demand approach simplifies purchasing a BI solution.

On-demand BI solutions don’t require you make any capital expenditures up front, since there’s no hardware or software to buy. Instead, you pay a monthly subscription fee. Prices vary, but a true on-demand BI solution can have an annual cost between 5 and 10 percent of the up-front hardware, software, and implementation costs of a traditional BI solution. Also, since an on-demand BI solution should be easy to configure, there should be no need to pay significant up-front implementation fees.


There’s really no magic in how on-demand BI solutions can be delivered for lower cost than traditional solutions deployed in a customer’s own data center. One of the key cost-reducing architectural techniques used by ondemand vendors is multi-tenancy. The concept is fairly simple, though the behind-the-scenes implementation can be quite sophisticated. Instead of every customer getting a dedicated deployment of the solution (the “single tenant” model), multiple customers (“multiple tenants”) share the same physical infrastructure. To ensure that customers see only their own data, each piece of data is carefully tagged with its owner’s information.

In addition, a set of configuration data is associated with each customer that indicates how that customer wants its solution to behave. Whenever a user makes a request to the BI service, each part of a multi-tenant BI solution (such as the ETL engine, data warehouse, analytics engine, and UI) reads the configuration data associated with that customer and processes and returns the appropriate results.


Figure 1. On-demand business intelligence architectural diagram.

Costs are lower because it is more efficient to manage a larger platform shared by a large number of customers than it is to manage a large number of individual deployments, where each is used by a single customer. For example, from an operations standpoint, each operational activity (such as creating a backup or upgrading the system) needs to be done only once per platform instead of once per customer, so you benefit from economies of scale.

Additionally, the hardware and personnel are shared across many customers, so they are used more efficiently and therefore the amortized costs are lower than if customers had to buy their own hardware and hire their own operations staff.

Try before You Buy

On-demand business intelligence solutions lower your up-front costs, but they can also help simplify the purchasing process. Customers want to be sure that any application they buy will meet their needs. Though sales meetings, conversations with other customers, and technical discussions can help, the best way to know if a solution will meet your needs it to try it yourself. Traditionally, this means that you carry out a proof-of-concept.

Because there is no pre-existing platform you can use, you must build and deploy a scaled-down version of the intended business intelligence solution. This can take weeks or months and requires a significant investment of time and energy.

With on-demand business intelligence, the solution is already built and running as a service that you connect to over the Web, so a proof of concept is easier to build. In fact, in many cases such a test drive is simple enough that on-demand BI vendors can offer a customer a free trial instead of the traditional resource-intensive proof of concept.

Simple to Set Up

A significant difference between traditional and ondemand BI solutions is how the solutions are set up by the customer. With the traditional approach, you first buy several sophisticated tools you can use to build your own unique solutions. These tools include an ETL tool to extract the source data, a data cleansing tool to remove duplicate records and match fields (such as customer name) across systems, a database engine to store the cleaned and integrated data, and reporting and analysis engines to create your reports, charts, and dashboards.

You use these tools to design, build, and deploy your business intelligence solution. This involves defining ETL scripts, designing star schemas for your data warehouse, and creating OLAP cubes, and often requires writing custom code as well.

The ability to build a solution completely tailored to your every need is certainly an attractive goal. However, just like designing and building your own home, building your own BI solution comes at a price: it can take 12 months or longer, and it usually requires hiring contractors or consultants with specialized skills to help you customize it.

Pre-built, Configurable, On-Demand BI Solutions

The on-demand approach to setting up a BI solution is different. The architecture has a generic hosted “analytic services” platform that performs all standard data-gathering, cleansing, storage, reporting, and analysis functions, but you don’t have to directly design or modify these components to create your solution. Instead, the focus is on leveraging pre-built solutions that sit on top of the platform and which you can configure easily (see Figure 1).

The pre-built nature of on-demand BI solutions makes them a great fit for companies that want visibility into common business processes and want to answer their employees’ most common questions about those business areas. For example, one pre-built solution can provide visibility into marketing-campaign effectiveness, another can provide visibility into all aspects of the lead-to-cash business process, and another can provide visibility into suppliers and inventory. Unlike designing and building a home from scratch, the “pre-built and configurable” ondemand approach equates to someone buying an existing home and then redecorating it.

Of course, the notion of pre-built BI solutions might not seem new, but traditional pre-built solutions (sometimes called “analytic applications”) are very different from the new breed of on-demand pre-built solutions.

Traditional pre-built solutions typically focus on only a few components of the overall BI technology stack, such as supplying pre-built data warehouse schemas to answer certain types of questions about particular business areas. All the other components that make up a BI solution (such as ETL scripts or OLAP cube definitions mentioned above) still must be custom built.

Modifying the pre-built components to meet your needs requires using development tools to modify database schemas and write custom code. In contrast, true on-demand BI solutions deliver pre-built solutions that include everything pre-built, not just some of the components This includes connectors to data sources, pre-built ETL scripts, predefined data warehouse schemas and OLAP cubes, and a set of pre-built reports that analyze key metrics. The focus is on providing a complete end-toend solution to truly simplify the setup experience.

Customization versus Configuration

The pre-built nature of on-demand business intelligence doesn’t mean you get a set of reports or dashboards that are rigid and fixed. Different companies have different needs, which is why pre-built elements are modifiable. You can also create a new report from scratch.

The process for tailoring on-demand BI solutions to meet your needs differs significantly from the process of tailoring traditional BI solutions. I refer to the traditional approach as customization, and the on-demand approach as configuration. That is, the traditional process for tailoring a solution to meet your needs involves using development tools to open up the analytic platform and customize the components directly. For example, you might edit a data warehouse schema, change an ETL script directly, and write custom code.

The benefit of this approach is great flexibility, but the downside is that it requires significant skill to make, strenuously test, and debug these modifications to ensure that you get valid results.

The on-demand approach is different. Rather than using development tools to customize the underlying platform by changing schemas and writing code, you use a point-and-click approach to configure the application. Essentially, you walk through a number of set-up screens, answering questions, selecting options, and providing information such as user names. It’s more like configuring your MyYahoo! home page than writing custom code.

BI Set-Up Traditional BI Solutions BI Software as a Service
Design and deployment time Months Days
Application design team In-house IT and/or BI consultants SaaS vendor
Deployment team In-house IT and/or BI consultants End user
Special skills required for set-up DB, DW, ETL, OLAP, reporting None

Table 2. Comparing traditional BI solution set-up to on-demand BI set-up.

For example, to tell an on-demand BI solution the type of source system of your data (such as whether it’s a CRM or financial system), you select the type from a pick list of all the types of source systems that the solution supports. To indicate which data fields you’re interested in using from those source systems, you view a list of all the fields and check boxes to make selections.

The same configuration approach is used to tailor the on-demand BI solution to your company’s business processes. For example, there are a few common processes for recording order cancellations in order entry systems. Some companies use a process that updates the original order record and changes its status from “Open” to “Cancelled.” Others leave the original order record untouched and create a separate cancellation record that points to the original order record.

Clearly, a BI solution needs to know which of these processes is being used to handle your data correctly. Traditionally, you would use developer tools to customize the BI solution’s order-cancellation logic to match your company’s process. With the on-demand approach, the pre-built solution has support for both processes built in; you configure the solution by telling it which of these processes your company uses.

The Benefits of Configuration

The on-demand approach favors configuration over customization in two important ways. First, the pointand- click nature of configuration means individuals don’t need specialized knowledge, and they don’t need to be skilled in database schema design or writing custom code. This can greatly simplify setup. Second, the approach simplifies upgrades. The traditional model of going under the hood and customizing the underlying schema and business logic creates complexities when upgrading to a newer version of your BI solution. If you’ve modified the schema and your vendor provides a software upgrade that includes a new version of the schema, none of your customizations will appear in the new version.

Upgrading involves installing new software and schemas as well as recreating all your customizations in the new version. Often the vendor will make architectural changes to the platform that require you to redesign how your customizations are implemented.

Using a configuration approach instead of a customization approach avoids this issue. With configuration, you specify the rules and select the settings that define how the solution will behave, leaving the underlying platform untouched. Since your configurations are not directly tied to the underlying platform, your configurations are preserved across changes in the underlying platform. Any new version of the platform just needs to look at the rules and settings you’ve already specified to set the solution’s behavior.

Using the order cancellation example discussed above, even if the vendor completely rewrites the underlying platform and database schema, the new version need only look at the choice previously made about cancellation processing to know that the company processes them by creating a separate cancellation record.

Why Configuring Pre-built Solutions Works

The traditional BI industry has not significantly focused on pre-built solutions. Instead, it has grown up building custom solutions. All of the dominant traditional BI vendors began life selling toolsets customers could use to build their own solutions. The vision has always been that you could have a customized solution tailored to your exact needs that could answer any question you had about any area of your business. There’s even a thriving ecosystem of consultants and systems integrators to build these custom solutions for you.

The problem is that if you ask employees at most companies what visibility they want into the business data to succeed in their jobs, it’s unlikely that they’ll say they need to be able to answer any question about any area of their business. In fact, as a consultant, I was regularly asked to create over 100 pre-built reports for clients. Out of all those reports, rarely were more than a dozen ever used regularly.

Rather than wanting to answer any question about any area of their business, most employees want a handful of relatively straightforward reports that answer some key questions about their area of responsibility. They’d also generally agree that the questions they ask are quite similar to those asked by others in comparable roles in other companies.

That’s the key that enables pre-built solutions to succeed. Despite the “everything must be custom-built” mantra we’ve been practicing and preaching since the late 1980s, people in similar roles in similar companies ask similar questions. Though most people wouldn’t find this statement to be controversial, most are still building custom solutions using traditional BI tools.

More controversial is that people in similar roles across companies in different industries also tend to ask similar types of questions. When I was a BI consultant, I helped build customized solutions for an airline, a telecommunications company, a sports apparel manufacturer, and a fast food chain, to name a few. I’m sure I’m not the only ex-consultant who realized that at least 80 percent of what I built again and again was the same across clients, because the questions they wanted to answer were similar.

For example, a sales manager at a software company wants visibility into her team’s forecast compared to their actual bookings, how much her sales reps are discounting, trends in the lead-to-cash cycle length, and the win rate compared to top competitors. Those are the same types of questions that sales managers ask in the manufacturing industry and the professional services industry.

As another example, a marketing manager in a consumer electronics retailer wants visibility into customer segmentation and customer profitability, marketing expenses compared to budget, the ROI of marketing campaigns, and Web site traffic analysis, as do marketing managers in several other industries (including travel and clothing).

The point is straightforward: when there are similarities in the solutions people are seeking, there is a great opportunity to provide pre-built, configurable solutions to meet their needs.

BI Usability Traditional BI Solutions BI Software as a Service
Feedback on usability Periodic surveys Online usage statistics tracked by application
Link betweeen usability and vendor revenue Indirect Direct
Percentage of functionality actually used Low High

Table 3. Usablity is a core part of the on-demand BI mindset.

Configuration Requirements

The idea of making set-up simple by leveraging configurability works best if all components of your BI solution are integrated and can be configured from a centralized location. Real visibility comes from pulling together data from multiple sources, so missing any of the required components significantly increases your set-up time and complexity. If any piece requires separate customization, you can’t run through the configuration process once and be done with it. You’d also have to separately customize each of the other components.

For example, imagine that the configuration settings on your MyYahoo! home page aren’t integrated with backend content providers. You can change the configuration settings to return, say, 10 news headlines instead of five, and though the Yahoo user interface might now expect to display 10 items, the back-end content providers are only providing five headlines. You would have to change the interfaces to the content providers so they would return 10 headlines.

If you have the ETL, data cleansing, data warehousing, reporting, and analytics engines all sharing the same configuration information, all components are updated with any change you make.

Simple to Use

The on-demand approach to BI enables much simpler set-up and initial maintenance. The success of BI in your company is by no means guaranteed just because you’ve successfully set it up. Of course, BI is only valuable if people actually use it, and that depends on two factors.

First, the content has to be valuable. People must feel the visibility they are getting is helping them succeed in their jobs. Pre-built solutions can be helpful here, because they can represent the feedback from hundreds of different companies about their best practices and most valuable types of reports, charts, and KPIs.

The second key factor is usability—if a solution is confusing or hard to use, internal adoption will drop, and the value you gain from that solution drops. For many traditional BI vendors, the approach to increasing the value that companies receive from their solutions is to deliver an ever-expanding list of more powerful features. Experience has shown that this strategy backfires. It makes the solution more complex, and the result is that most BI solutions now require users to attend long training classes.

As mentioned earlier, the on-demand paradigm is more about a mindset than a feature set. There is more emphasis on building a user interface that requires little or no training, that makes it easy to bring in external data, and makes it easy to share results with a community than there is on building a long list of power-user features. This is the right approach, since the majority of people would rather have a clean and simple solution that works well and is easy to use than a powerful and highly sophisticated solution that’s hard to use. That’s why iPods are so successful. They do just what you want them to do, and they’re a delight to use.

Of course, every BI solution (whether traditional or ondemand) tries to be easy to use, and over the years there have been important improvements in usability across the BI industry. With on-demand, providing solutions that are simple to use takes on an even more critical role. Usability and user adoption are core necessities built into the on-demand business model—if users find the solution too hard to use, they’ll stop using it, and customers will stop paying for the service. The direct link between usability and revenues has a strong impact on how ondemand solutions are designed and built.

Common Concerns about the On-Demand Model

Though people certainly value how the on-demand approach focuses on making solutions simple to buy, set up, and use, there are some common concerns related to this model.

Security. People are sometimes concerned about the security of their data since it’s being managed by a third party. As a related issue, people want to know if outsourcing the management of their data will create Sarbanes-Oxley compliance issues.


As mentioned earlier, security is a key focus for ondemand vendors. In 1999, the idea of having a third party manage your data was a foreign concept that made many companies uncomfortable. The environment is very different today. There are many successful companies (several of which are public companies) such as that now act as proof-points for having third parties safely manage customer data.

Second, many industry best practices and government standards have been defined for securely managing and protecting customer data. There are also certifications (such as SAS 70) awarded by third-party security and auditing firms to provide a level of trust and to ensure that all security standards, Sarbanes-Oxley compliance guidelines, and best practices are being followed by on-demand vendors.

In addition, because on-demand vendors can leverage economies of scale and amortize their security costs over all their customers, they can spend more on building a world-class security infrastructure than most companies can. So, for many companies, the data being managed by their on-demand solution vendors is often more secure than the data in their own data center.

Network bandwidth. Because on-demand business intelligence means that there’s a data warehouse hosted remotely, people are occasionally concerned about how much bandwidth it will require to populate the data warehouse.

First, bandwidth requirements depend on how much data a company generates daily. If you’re thinking about populating a data warehouse to analyze all the point-ofsale purchase records for Wal-Mart, then the bandwidth requirements would be very large, and currently on-demand probably would not be a good fit for that application. For most other companies and most other types of analyses, however, the bandwidth requirements are easily managed.

In fact, the only time that bandwidth requirements might be high is during the initial load of the data warehouse, when you want to upload all relevant history. In this case, it might take many hours to complete the initial upload. After that, if you’re refreshing the data warehouse daily, you only need to upload the data that has changed in the last 24 hours.

Even if you managed to modify 10,000 records in a day, that usually translates into less than 100 MB of data. It doesn’t take much bandwidth to upload that amount of data once per day. If your uploads are more frequent (twice a day, perhaps), they involve even less data.

Vendor viability. Another common concern is data loss if your on-demand vendor goes out of business. There’s little to validate this concern. To be sure, just like all types of vendors, some on-demand vendors have closed their doors. When they do, you don’t lose your data. The vendor doesn’t own the data—you own it, so at any point you have the right to ask for your data back, and it’s usually a simple process of exporting the data and sending it back to you. Also, this becomes less of a concern with ondemand business intelligence, since the data managed by the on-demand BI vendor is just a copy of your data—the original data is still in your operational systems.

Historically, new approaches to solving business problems have brought with them new concerns, and on-demand BI is no exception. However, the main concerns have been addressed by vendors, enabling the on-demand approach to spread rapidly.


The business intelligence industry has much to be proud of. It has helped commercial, educational, and government organizations gain insight into their operations and improve their overall effectiveness. The introduction of the on-demand model to this industry will have far reaching effects, opening up business intelligence to a far broader population of companies and individuals. Many successful on-demand solutions have already paved the way, and the on-demand model has already been applied to a broad range of software categories.

Over the past two decades, business intelligence has changed the way we do business. It will be fascinating to see what business intelligence on demand will enable in the next two decades.

Company On-Demand Focus
Arena Product lifecycle management
Concur Technologies Corporate expense management
Eloqua Marketing automation
LucidEra Reporting and analysis (BI)
NetSuite ERP and CRM
Omniture Web site analytics
Postini E-mail security
RightNow Customer service and support CRM
SuccessFactors Employee performance management
Xactly Incentive compensation

Table 4. On-demand vendors are addressing a wide range of market need.

Ken Rudin -

Ken Rudin is a veteran of the rapidly growing on-demand/"software as a service" industry, bringing to LucidEra more than seven years of experience as an executive with leading on-demand software vendors--including SVP of products at, a founding board member of NetSuite, and VP and GM of Siebel's CRM OnDemand division. As Co-founder and CEO of LucidEra, Ken is focused on redefining the business intelligence market by providing a complete business intelligence on-demand solution.