RESEARCH & RESOURCES

Business Intelligence and Data Warehouse Technology Models Under Fire

Economic crisis invites a referendum of BI and DW technology models

Deserved or not, the travails of software-as-a-service (SaaS) business intelligence (BI) specialist LucidEra Inc. raise questions about the viability of the SaaS model. More importantly, they invite a referendum on the proper -- or most cost-effective -- way of doing SaaS. Unrest in the SaaS BI segment parallels that in the data warehouse (DW) appliance space, where hardware-centric players such as Dataupia Inc. are also finding it tough going (see http://www.tdwi.org/News/display.aspx?ID=9493).

In both cases, it comes down to a question of dueling technology models -- not of business models, competitors argue.

To a degree, industry analysts also concur. For example, says Wayne Eckerson, director of TDWI Research, while the SaaS BI business model itself is sound enough, there's a concern that SaaS BI start-ups, much like LucidEra, could be adversely impacted by a brutal venture capital (VC) funding climate, in which VC backers are chary of costly or (to this point) unprofitable technology propositions.

Likewise in the DW space, where -- from the foundational success of Teradata Corp. to the appliance pioneering of Netezza Corp. -- the DW appliance now has broad acceptance, even if certain implementations of the DW appliance model (e.g., Dataupia) are struggling.

Why are some companies able to secure funding -- and, thus, longevity -- while others aren't? This is a ball that SaaS players, in particular, are anxious to run with. At issue, they allege, are the best SaaS BI (and, by extension, DW appliance) technology models. Not surprisingly, no one seems to agree on precisely which vendor has the best technology model -- although if events in the DW appliance segment are any indication, the SaaS BI market could ultimately coalesce around vendors that market the most platform-independent offerings: in other words, SaaS BI offerings that can be variously deployed on internal (enterprise) IT resources; in a hosted (service provider-based) context; or on top of a subscription cloud service (such as the increasingly ubiquitous Amazon Elastic Compute Cloud, or EC2). Such offerings -- like the DW appliance that can run on a mix of in-house, proprietary, and cloud-based hardware or services -- are probably best positioned in the long run.

Dueling VisionsTDWI's Eckerson sketches a multi-tiered SaaS BI-scape, citing first the classic -- or application-centric -- SaaS BI model, pioneered by Salesforce.com. After SaaS, of course, comes what Eckerson calls the "infrastructure-as-a-service" (or IaaS) play; EC2, he notes, is a textbook case in point of IaaS in action.

There is also something called platform-as-a-service (or PaaS), in which a vendor enables customers to create custom applications via the Web. In the BI/DW market, this means providing a platform that lets users create Web-based reporting or dashboard applications along with underlying data marts entirely via Web services. LucidEra, for the record, fit more comfortably into the general (or Salesforce.com-esque) SaaS model, inasmuch as its offering took the form of a full-blown BI application-- specifically sales pipeline analysis. One problem with respect to the classic SaaS approach is that it’s enormously costly and assumes big upfront costs (associated with both designing and building out a platform and purchasing, deploying, and managing a multi-tenant hosting infrastructure); a second issue -- and one that’s likely to give potential VC backers pause in the current climate -- involves the slow ramp-up to profitability, with (initially) slow customer uptake and small deal sizes.

That's likely the case at LucidEra, says industry veteran Mark Madsen, a principal with BI and DW consultancy ThirdNature. He cites both the cost of designing, deploying, and hosting the LucidEra solution and the comparatively small deal sizes that (at this stage) characterize most SaaS BI engagements.

An abundance of SaaS competitors, many of which tout what they claim are superior refinements to LucidEra's model (particularly with respect to platform topology or infrastructure underpinnings) didn't help, Madsen adds.

Add to this the fact that LucidEra is said to have consumed more than $20 million in VC funding and you have a scenario in which VC backers became reluctant to engage further.

Based on our interviews, that's the consensus on LucidEra's travails. The company's SaaS BI competitors, not surprisingly, have distinct (and self-serving) spins on this take. What is surprising is that several LucidEra competitors position the company's flame-out as a referendum on its traditional SaaS technology model, even as they disagree about which model will triumph once the dust settles.

A Platform Play

Take industry veteran Dyke Henson, chief marketing officer with SaaS analytics specialist PivotLink. Henson believes that SaaS BI is "inevitable," even as he stresses that there "will always be a place for traditional reporting and analysis." PivotLink, like LucidEra, is a privately-held, VC-funded enterprise. (PivotLink generated $10 million in VC funding during Q1, according to Hensen.) Unlike LucidEra, he says, PivotLink doesn't shop a full-fledged BI solution, complete with DW, reporting, analysis, and analytic applications; it pushes, instead, a Web 2.0-based PaaS offering that gives users a means to perform reporting and analysis against financial, operational, and other data sources.

Earlier this year, PivotLink announced a new dashboarding capability (PivotLink Gadget) based on Google Apps. PivotLink Gadget lets users build BI dashboards in the context of the Google Apps Web environment -- effectively recasting Google Apps as a BI front-end, Hensen argues.

"Our target market in general hasn't been the 24x7 dedicated three super analysts within the organization. We target the equivalent of a power user that might've been used to Pivot Tables and just couldn't scale [Excel] to what they needed to do," he explains. A typical PivotLink experience might involve a user "dropping" (via a Google Gadget) "one or multiple views of data into a Google iPage," Hensen explains. "The target here is obviously [information] consumers as opposed to hardcore data analysts."

Hensen believes that PivotLink's PaaS-based approach is a -- if not the -- viable model for SaaS BI. He anticipates, however, a topology in which PivotLink's PaaS-based offering ultimately migrates over to a IaaS service, instead of (as is currently the case) being hosted internally. "Right now, we don't run on Amazon services. We do our own hosting. We are looking very closely at moving beyond that. Right now, believe it or not, in our pricing, [Amazon is] just too expensive. The EC2 value proposition is not there when you're dealing with the composite [mix] of customers that we have today," he says.

Roman Stanek, founder and CEO of SaaS BI start-up Good Data Corp., says the economics of PaaS running on top of IaaS already make sense -- for his company, at least. Good Data, like other SaaS or SaaS-oriented players (including DW specialist Vertica Inc.) exposes its offering via Amazon EC2. He says the benefits of EC2 -- particularly with respect to the "elasticity" that's unique to an EC2-like service -- far outweigh the costs.

"We would like to automate as much [of the management associated with the service] as possible. That's the way to make SaaS work. If you go to GoodData.com, you can actually sign up for a project and you will have a project provisioned in a few seconds," he comments. "There should be no affinity between the software and the underlying hardware. Any time you have to provision the hardware, you have ongoing fixed costs," Stanek continues.

"Your own hosting gives you no elasticity. If you host on your own hardware, you're limited to your own capacity, so how do you size that? Do you size that for a minimum amount of load, do you size it for the average amount of load, or do you size it for a big load? However you look at it, you always end up with some dissatisfaction."

Good Data is very much a start-up: Stanek is reluctant to discuss either the size of its customer base or the extent of its capitalization. (Good Data has raised approximately $2.5 million from VC funders, he confirms.) In addition to its bread-and-butter Good Data PaaS offering, Good Data markets two application-specific analytic offerings: Good Data for NetSuite (a reporting and analysis application designed to work with SaaS player NetSuite’s on-demand application stack) and Good Data for Amazon CloudFront, an analytic app designed for Amazon’s CloudFront content delivery network

Stanek says Good Data will introduce other analytic services in the coming months. It takes its pitch to partners, integrators, and enterprise customers, he explains, contending that this approach enables it to both mature the base Good Data infrastructure and develop a portfolio of industry-specific analytic applications.

"What we actually provide is fully customizable ETL, data warehouse, analytic engine, and dashboarding [services]. All four of those components can be turned into projects [by users]," he comments. "We pitch the generic platform to people who have vertical knowledge. We have a partner who builds some different types of applications and we will be coming out with those applications when they are built. Once these vertical applications are built, then we go and pitch it to business users -- for example, NetSuite analytics is a high-level [customer analytic] application that we pitch directly to users," Stanek continues. "We work with NetSuite integrators to help us with the NetSuite platform and we are working with companies in hospitality and logistics and IT management and so on to build applications."

Stanek concedes that Good Data is still a gestating entity, but argues that its hybrid model -- which consists of both a PaaS BI infrastructure offering (Good Data) designed to layer on top of a cloud infrastructure service (e.g., EC2) and a still-evolving portfolio of industry-specific analytic applications -- augurs the future.

"When someone says to me that Amazon is too expensive, I say 'No, if your system is not able to take advantage of that elasticity that Amazon gives you, then obviously it's too expensive!' The way we designed [Good Data] is that we can actually change the number of resources that we get from Amazon on the fly. Because we pay them by the hour, we have dynamically responsive [capacity]. We don't have to plan for seasonal fluctuations or business cycles. We don't have to pay to host or manage [the service]. This is the model the industry will adopt, even if EC2 isn't specifically [the service] used."

TDWI Membership

Get immediate access to training discounts, video library, BI Teams, Skills, Budget Report, and more

Individual, Student, & Team memberships available.