SaaS and BI – September 2008

Software as a service represented about 5% of business software spending in 2007, but analysts are predicting that this trend is going to accelerate. Gartner has predicted with a reasonably high
degree of probability that by 2011, 25% of new business software will be delivered through a software-as-a-service (SaaS) model.

Some people compare SaaS to the ASP (application service provider) world, but SaaS is a very different animal from ASP. For one thing, times have changed. The rich browser-based, highly interactive
application experience has come a long way since the days of static HTML and Java applets. New and innovative application development methods such as asynchronous Java and XML (AJaX) – the
magic behind Gmail and Google Maps, and Adobe Flash-based technologies for example – have helped transform Web apps into highly interactive entities. For this reason and a few others, some
industry analysts say the time might just be right for SaaS BI – particularly in the SMB segment, although recent statistics are proving that on-demand penetration is not being limited to the
SMB sector. In fact, in a survey conducted by Aberdeen research, it was found that 13% of large organizations surveyed are planning to start on-demand business intelligence (BI) projects in the
next 12 months. This was at a higher rate than mid-tier (8%) or small (11%) companies surveyed in the study. Aberdeen research has also shown that “limited BI skill sets” among users is
the top pressure companies are facing as they attempt to expand the delivery of actionable information to the enterprise.

Over the last 2-3 years, on-demand BI has gained adoption because it’s appealing to smaller companies that can’t afford to invest in a full-blown BI solution, and to larger companies that are
looking for some simplicity in the way things get done. It would be very prudent to think that the companies that are most likely to attempt SaaS BI deployments will be the ones with existing SaaS
investments, such as CRM or ERP.


What is On-Demand BI?

Business intelligence (BI) refers to products or solutions that help companies store, find and analyze the information they need to make better business decisions. Business intelligence used to be
a rearview mirror showing where a business had gone off the road, but today the face of BI has completely changed. Business managers today want BI to serve as a global positioning system, showing
them where they are and the shortest paths to their destinations to stay ahead of the competition.

On-demand BI includes BI applications accessed by business users over the Internet. Some companies also have a set of platform services for IT users in support of on-demand business intelligence.
Examples include BI development tools, data warehousing integration tools, security services, analytical database processing services, and so on. These platform services are also known as
PaaS-based services (platform as a service). 


Architecture Models for On-Demand BI

Before we delve deeper into the cases where on-demand BI may be more applicable, let us look at the various architecture models for on-demand BI. Although vendor offerings from SaaS vendors might
look the same, there are essentially 3 key architectural areas along which they generally differ –infrastructure, data separation and versioning.

Some vendors adopt a shared infrastructure (multi-tenant) where the same infrastructure is shared amongst all customers. It has lower costs, but sometimes the perception here is that
performance can be affected as usage increases. In some cases, this can be turn out to be true, but performance levels can be managed even in a shared infrastructure by ensuring that the data
center provider does not over-provision the infrastructure and puts controls in place to ensure that no single customer can monopolize the infrastructure resources. A dedicated
infrastructure
is built on a per-company basis and is typically associated with higher costs. This option, just like the shared option, also has to deal with spikes in usage and be able to
scale linearly and perform intelligent performance optimization.

Data privacy and data separation relates to SaaS vendors providing logical or physical separation for their customers’ data. Logical separation implements data separation by via logical
separation
of customers’ data within a common database. This allows better sharing but can be perceived as a security risk. But at the same time physical separation does not
implicitly mean that it is more secure. Some vendors charge an incremental fee to provide physically separate databases for their customers to cover the additional costs incurred by them.

The third architecture component in the way SaaS vendors differ is the way they support versions of their software. The popular SaaS vendors adopt the single-version model while some
others allow multiple versions of the software to be co-existent. If a customer doing an on-demand BI implementation is thinking about a lot of customization and integration, then a model that
supports multiple software versions might be a better and more flexible way to go. The biggest advantage of the multi-tenant / single-version architectural model is that for commodity applications,
bug fixes can be addressed more quickly.

If you really look at it, a model that has a dedicated infrastructure, supporting multiple software versions and providing separate physical databases, is in fact the old “managed hosting
service” model. So it cannot be truly called a SaaS offering. It would be very expensive for SaaS vendors to support multiple models as they need to manage both shared and dedicated
infrastructures, which brings in a level of inherent complexity that can very well be avoided by offering lesser options. 

Another architecture model that seems to be emerging is the hybrid on-premises/SaaS capability enabling large enterprises to retain their data warehouse in house while using the power of cloud
computing to collaborate on BI data and enhance the BI user experience with sophisticated, yet simple, but highly interactive Web 2.0-like functionality.


When to Use On-Demand BI?

In my opinion, SaaS is a valuable addition for SMB companies and departments in large companies that don’t want, can’t afford or don’t have the expertise to install a full-fledged BI or data
integration infrastructure. The on-demand delivery model allows SaaS vendors to deliver products in shorter release cycles. Companies that leverage innovation developed in on-demand services,
including the ease of use aspects of the browser interface, will take huge strides in establishing themselves for tackling new competitive frontiers. An on-demand BI solution offering should be
considered if you have a straightforward set of requirements around scheduled reporting, or are trying to monitor a particular piece of data or are handling repetitive analysis of X measures.
On-demand BI works best in environments where most software and processes are standardized and a high level of customization is not needed just to make things work for the organization.  With
BI on demand, the SaaS vendor builds the data model and provides a specific set of reports and/or dashboards. Some degree of customization is possible, but 70-90 percent of what you’ll do
would be according to what that BI on-demand vendor has created.

In situations, however, where loss of control is an issue, where significant customization is needed, where there is a significant pushback from IT, where data is too sensitive to be stored outside
a firewall, or where your company runs the risk of outgrowing the current BI application provided by the vendor(s), it is better to go with a more traditional on-premises BI approach.

Although the primary usage of on-demand BI is to drive down cost of ownership and the cost of expensive IT and user skills (i.e., provide business value at a lower price point), it can also be a
competitive differentiator providing capabilities that could not have been easily possible or, in some cases, impossible in the traditional BI approach. Companies should keep their options open and
never get locked into a multiyear agreement, especially if this is a new approach for your company. Going with the minimum period – usually a year – is probably the more prudent
approach. Businesses grow every month and every quarter, so you might need to change your approach over time. Thus, having a flexible contractual agreement with your on-demand BI vendor gives you
the needed flexibility and freedom to do so.


Is On-Demand BI the Panacea for all BI Problems?

It is interesting to study the reason behind why companies that have adopted on-demand BI have stuck to it. The reason that most companies give is the ability to get answers to tough business
questions when you need them and in the most palatable web-based format without having to devote a team of BI professionals from IT. But even on-demand BI demands a thorough and in-depth
understanding of its benefits to the enterprise. In other words, understanding the business problem being solved is of paramount importance and will almost always be the deciding factor in the
success or failure of the BI implementation. On-demand BI needs to be based on clearly documented business needs just like any other technology that gets implemented to solve a business problem. If
the business problem wasn’t clearly defined or not well understood, then even an on-demand BI offering will not be of help to the company. Data quality and data integration challenges that
plague the on-premises world do not automatically get solved in the on-demand world. They have to be carefully thought through with a clear path chalked out to resolve the issues involved. The path
to issue resolution has to be dealt with from a strategic standpoint and not a reactive tactical standpoint.

One of the biggest challenges with hosted BI solutions is data integration – involving areas such as data access, data profiling and data cleansing. This is because of the adapters that must
be run within the organization’s infrastructure to get access to clean data and data of different shapes. Some companies are trying to solve this problem or at least alleviate it to a
reasonable extent by providing a “data integration as a service” offering. On-demand data integration would be a great enabler for on-demand BI, in my opinion.

Customization is another challenging area as a lot of the packaged BI applications are tuned with a business model of standard packaging. It is hoped that the product capabilities for customers to
customize the product offering and integrate with ease will get more and more mature over time. 


Future of On-Demand BI

In this section, we will look at some key factors that will affect and influence the future of on-demand BI (see Figure 1).

Figure 1: Factors influencing On-Demand BI

  1. Hybrid application architectures: Of the 3 key trends in on-demand BI that have been described in detail in my earlier article – SaaS and BI: Overview of On-Demand BI and Its Key Characteristics, I believe the most important one is the emergence of hybrid
    application architectures where on-premises data, applications and processes are linked through web services-based integration APIs (application programming interfaces) to SaaS-based
    applications. I strongly believe that large enterprises will probably never throw away their customized on-premises software but will most probably look at SaaS-based applications to complement
    them in an effective way to improve the bottom line. And if you think about this closely, this strategy makes perfect sense as these enterprises have invested millions of dollars customizing
    on-the-shelf software or building homegrown BI applications to suit their particular business needs. Real-time collaboration “in the cloud” around business data will be an exciting
    area. Tighter integration with productivity tools like Google docs and Google Apps will put BI and productivity tools more closely together. The combination of corporate data mashed up with
    public data for greater insights will also be interesting and so will be new RSS functionality, enabling users to subscribe to changes in new powerful ways. 

  2. Inter-Organization Benchmarks: As on-demand BI gains more adoption, an interesting area to look at as part of BI 2.0 would be inter-organization performance metrics in areas
    other than finance (which they can do today for public companies that have to report financial metrics). The ability to easily augment internal data with external benchmarks, such as market
    share, economic predictions, etc. would be very useful. Executives will continue to get the internal numbers derived from their expensive corporate information systems, but what would be more
    compelling for them would be to determine how the internal numbers compare to external benchmarks. They would want to be able to keep track of financial numbers (revenue, stock price or return
    on capital), standard industry metrics (revenue per square foot of retail space) and, increasingly, non-financial metrics such as healthcare outcomes and customer satisfaction. The gathering of
    this sort of benchmark data is already a big industry, with major players such as Hoovers for business reports; D&B for credit information; and Thomson for financial, healthcare, legal and
    scientific information. Then there is also the concept of the “deep web” – that refers to World Wide Web content that is not part of the surface Web, which is indexed by
    search engines. It is estimated that the deep Web is several orders of magnitude larger than the surface Web. The rise of the “deep web” means that there are vast stores of information that
    could be used to benchmark almost every conceivable economic activity. The emergence of new data standards such as XBRL, web services and other new information providers will greatly ease the
    process of comparing benchmark information with internal data.

  3. Adoption of Data Quality Functionality: This will gain more adoption in the hosted model as certain aspects of it such as customer address cleansing and validation, matching,\
    and data profiling have a well-defined scope and are largely repeatable across enterprises. Increasingly mature integration middleware and SOA are making it easier for BI applications to be
    integrated regardless of whether they reside inside or outside an enterprise firewall.

  4. Focus on Operational BI: BI, in general, is moving from the strategic side of the enterprise – which focuses on financial performance, costs, customer profitability, and
    similar corporate analytics – to operational business intelligence, which focuses on such transactional information as whether to give a customer a discount, whether inventory levels need
    to be increased or whether a new promotion is driving sales in the current day. On-demand BI will follow suit.

  5. Ease of Use to Drive Adoption: A key factor that would be affecting the success or failure of on-demand BI vendors would be the ease of use factor across 3 areas – setup,
    integration and training. In fact, training should be minimal. If the user needs to open the 500-page manual to be able to navigate around the UI, then the UI is probably not built right. A
    February 2008 Gartner report on self-service options for business intelligence concluded that users find BI tools difficult to use and consume. Anecdotal evidence suggests no more than 20
    percent of users in most organizations use reporting, ad hoc query and online analytical processing tools on a regular basis. Lack of both end-user and developer skills is frequently cited as a
    major barrier when deploying BI applications. But on-demand BI has the potential to change this.

  6. Changing Role of IT: With the emergence of easy-to-install on-demand applications, IT’s role as gatekeeper will get further minimized. By 2012, Gartner predicts that emerging
    technologies such as on-demand and SaaS BI tools will make users less dependent on central IT departments to meet their BI requirements. However, it is also essential to be pragmatic here and
    realize that for bigger enterprise-wide implementations involving thousands of dollars IT’s buy-in and support for on-demand BI applications would not be a bad thing to have.

  7. Pricing Models: Another key trend would be the way BI software gets sold. The perpetual license option will probably disappear and the subscription model will probably take
    over for all almost BI applications. This would have a huge impact to the adoption of on-demand BI for both SMB companies and large enterprises as fewer approvals would be needed as there would
    be less upfront costs involved. And then the customer can always change their mind if they don’t like the offering or if the needs have changed drastically so there is greater
    flexibility.

  8. Emergence of Predictive Analytics: As companies perform more and more business functions following a metrics-based approach, and mature along the BI curve, they will need more
    and more sophisticated capabilities to predict how things will shape out against the competitive landscape. There are tremendous possibilities around the business value of combining techniques
    such as data mining and predictive modeling. When time will tell as to how this market will get shaped in the on-demand BI world.Conclusion

For customers, the reality is that for them it’s not really a choice between on-demand and on-premises –they just want it all, just like as consumers we don’t want to choose between the
internet and our PCs – where we can keep an offline copy of what we want. Organizations ultimately desire a solution that can be installed on-premises (if needed), used on demand, or any
combination in between. With SOA (service-oriented architectures) and data services-oriented architectures, the boundaries between on-demand and on-premises solutions are becoming increasingly
vague and for a good reason. The BI users of the future really don’t care where the data comes from – it could be accessing a seamless combination of web services that are executed on
systems inside their organization or systems that are outside their firewalls.

Sad to say, but the stark reality is that on-demand BI is not a magic wand in any shape or form. It would not be prudent to think and fantasize that the data is going to integrate itself or
magically become error-free if one goes with an on-demand BI application. However, having an on-demand BI vendor do it for you, on their platform, as a service offering can certainly make your life
easier and ease your pains to a reasonable degree.

Share this post

Diby Malakar

Diby Malakar

Diby works as a Principal Product Manager for Informatica, the data integration company. His prior experience includes working as Acting-VP of Engineering and Senior Director of Product Management for Cloud9 Analytics, a SaaS-based startup focused on building sales performance management applications. He has also worked for companies like KPMG, Neoforma and TiVo in various engineering management and product management capacities. He has more than 14 years of experience in the information technology industry and specializes in business intelligence, data warehousing and data quality management. He holds a Bachelors degree in Computer Science and a MBA in Information Systems.

He is an active member of IAIDQ and the Program Director for the San Francisco chapter of DAMA. His articles have been published in a variety of places such as TDAN, Wharton’s Leadership Digest and Oracle Magazine.

scroll to top