Q&A: BI Buy versus Build

The age-old question of buy-versus-build is alive and well for developers. We look at the current state of the decision-making process for BI professionals.

Much like the chicken-and-egg conundrum, one of the age-old questions of the relative merits of build versus buy continues to dominate discussions in the BI world. With the emergence of newer companies developing next-generation suites of products, BI This Week spoke to asks Peter Chirlian, CEO of Armanta, Inc., about the current status of the discussion.

BI This Week: How has the buy-versus-build decision evolved regarding the BI?

Peter Chirlian: At its highest level, this decision process now is no different for BI than it has been for any other technology. If IT professionals think the "buy" path provides all of the required capabilities, as well as being quicker, less costly, and less risky than a "build" solution, they will take that route. However, with the current crop of BI products on the market, that decision becomes more complex given the fairly generic state of the available technology.

Traditionally, this decision to purchase has been relatively straight forward: simply assess the product on its features, performance, and scalability. With BI, however, the products are not turn-key solutions. Rather, they are software tools that facilitate data visualization, thus making the assessment process more involved. To make an informed decision, one must consider how the BI installation would be used to deliver the desired capabilities. Typically, a reasonable amount of development work is needed to get those desired end-user features, and that means that even in the "buy" route, there is still some amount of "build" required.

How does the current economic environment influence the BI buy-versus-build decision?

As budget pressures mount, IT decision makers are increasingly looking to reduce expenditures on internal software development. This is a boon to the well-placed BI vendor. Incorporation of BI tools into a development plan can greatly reduce the time and cost involved in a large system deployment. An ideal BI solution would provide the speed of buy with the capability of build. It's as simple as that.

Although most BI tools are very good at visualization of clean and static data, few are capable of handling the dynamic and interactive needs of many of today's more complex applications. As there is now almost universal acceptance of BI tools in more straightforward applications, the current economic conditions do not have a great impact on the decision to go with a vendor solution in those cases. For more complex applications, IT executives would like to move from a strong internal-development bias to inclusion of BI tools.

Unfortunately, most available BI tools are not flexible or capable enough to play a role in these sophisticated systems. As a result, the internal IT effort required to incorporate these products into a large-scale application often outweighs the potential cost savings of a vendor-supplied BI solution.

What shortcomings in the current BI offerings lead to the "build" decision?

In the financial services sector, for example, we see many broad issues, including, the complexity of time-series data, real-time calculations, interactive requirements, and performance demands that have led to failures at the hands of traditional BI vendors.

That being said, we believe existing BI solutions fall short in many areas:

  • Solution can't work with data spread across multiple databases and geographies (and possibly multiple database vendors)
  • Unable to work with data that is not in a clean, relational format
  • No caching of both data and calculated results to reduce latency and hardware expense
  • No live calculations in both synchronous and asynchronous manners
  • The end user cannot manipulate the data and see the impact immediately
  • The solution cannot scale to handle very large amounts of data and calculations
  • No support for grid computing techniques to drive performance and enable the handling of very large datasets
  • No 'what-if' capabilities
  • End users are not able to greatly change the data that they are viewing without involvement from IT support personnel

Can these BI shortcomings be addressed without a build-from-scratch approach?

Not until recently, and -- of course -- my view is biased. While none of the traditional BI vendors has been able to address these significant shortcomings, there are now products available that marry sophisticated data integration capabilities, live caching, incorporation of customer-defined calculations, grid computing, and interactive BI capabilities all in a real-time environment.

I emphasize that the BI capabilities must be interactive because read-only reporting on static data from a single database is not sufficient in many circumstances. Since Armanta is one company that does provide these state-of-art solutions, we can attest to the fact that our customer base believes that to stay competitive they must provide the needed integration capabilities and capture all the data and computation residing throughout the enterprise, regardless of the source system upon which it was created. That ultimately leads them to us.