Thursday, March 01, 2007

Early Indications February 2007: The Evolving Enterprise Application

Until recently, enterprise software came in one of two basic shapes. If the firm built an application from scratch, the process was frequently long and expensive. Considerable effort was required for technologists to understand and translate business requirements into code. Once that code was tested and debugged, a particular application reflected the time and place of its creation. Accordingly, each stratum of the heterogeneous software environment often required maintenance from the very people who initially built it because it was often difficult to translate their knowledge of the application into documentation. Maintenance processes thus varied widely across the portfolio, contributing to complexity in the environment that exacted tolls in in time, money, and performance.

Buying an application from an outside vendor solved many of these problems. Microsoft, Oracle, and their kin had thousands of installed programs from which to learn, and used wide experience to build fixes and upgrades for the installed base. Even so, installing and maintaining enterprise packages such as SAP, Manugistics, or even Access could be expensive and difficult: different companies, functions, and, often, individual users require customization. Training remains a challenge even after web-based interfaces replaced many packages' idiosyncratic, proprietary screens and commands. Cost was also a concern as software license expenses are compounded by hardware upgrade requirements, consulting fees, and upgrades over the life of the application.

In both instances, scaling up could be difficult but scaling down -- after a change in corporate direction, spinoff, or shift in the company's market -- was often impossible: 500 software seats, with their attendant human and hardware infrastructure, didn't translate into 250 seats at anything like half the cost.

Six recent developments may portend some major shifts in the enterprise application market. All of them introduce the deep changes to prevailing funding, development, and support models that will be worth tracking.

*Software as a service, vertical flavor

On February 27, announced a 25,000-seat deal with Merrill Lynch, whose financial advisors will use the tool to help manage clients' portfolios. Salesforce had already established a foothold at such sector accounts as Suntrust and Aon, but the Merrill Lynch deal is a clear breakthrough into the top tier of the industry, at scale.

*Software as a service, horizontal flavor

Google Apps Premier Edition launched February 22. Like Salesforce, Google rolled out blue-chip adopters: GE and Procter & Gamble, in addition to many universities and smaller businesses already enrolled. For $50 per user per year, enterprises get

-10GB of storage, about 100 times the typical corporate e-mail box
-99.9% uptime guarantee (which still translates to over 8 hours a year downtime)
-24x7 help desk support
-APIs for single sign-on, data migration, and other integration tasks
-Gmail for Blackberry
-Google docs and speadsheets, which allow for multiple employees working on the same document at the same time but which are not yet fully interoperable with Microsoft Office files
-Administrative access to ensure compliance with corporate policies (who sees which calendars, for example, or what attachments are and are not permitted)

*Corporate mashups

FedEx CIO Robert Carter recently told an audience at Wharton a little about a highly secretive pilot project the company is running. High-value packages, particularly biotech-related shipments such as bone marrow, are tagged with active radio sensors that transmit the parcel's location to the company over public wi-fi networks. FedEx also used the readily available Google Earth APIs rather than building or licensing GIS software, particularly for a pilot project. The resulting application aims to blend mapping data, video from the trucks, and package status (presumably including temperature, shock and motion records, and tampering indicators) to create a new generation of service to shippers with particularly sensitive items in transit.

-RSS inside the firewall

Serendipity Technologies, an Israeli startup I saw at the DEMO conference, encapsulates enterprise application data into RSS feeds. It's not hard to envision an end user with the WorkLight product assembling her own desktop: in-house data such as available-to-promise inventory, customer order history, and current pricing could join Mapquest, Weather Channel, Yahoo Finance, and newswire feeds to get a sales rep ready to do a day's calls. I've asked around informally, and something like 40% of many enterprise's application portfolios consists of reporting tools. Making the relevant feeds available as services for business users to assemble as needed would seem to be a way around the long-cycle, expensive, rigid development process for non-transactional applications.


Building on its integration of analytic functionality on a customized Intel-based device to create the "business intelligence accelerator," SAP will be launching an enterprise search appliance sometime this year. Other vendors have had similar success in bundling software onto special-purpose hardware to speed deployment, lower cost of ownership, and improve performance relative to a general-purpose platform. Familiar examples include antispam (Ironport), data warehousing (Teradata), and network security (Symantec).

-Show, don't tell

The requirements definition phase for an enterprise software project can be a long, frustrating exercise as businesspeople and technologists struggle to define what's possible, what's cost-effective, and what's necessary. There's been a long-running debate as to how well people in IT shops can learn the ins and outs of business process compared to how well users can learn to use yet-to-be-invented lightweight "development" tools (Excel macros are the standard existence case) to build the applications they need. Neither argument has prevailed.

I recently encountered a company that's straddling the line with a intuitively appealing tool that presents an alternative to translating interviews, focus groups, and other forms of task analysis into a text document often hundreds of pages long. By their very size and complexity, these documents themselves inject delays and ambiguity into the processes of both package deployment and custom development. Instead, teams at such companies as Wachovia, Agilent, and Dow Jones are using tools from a company called iRise to see realistic simulations (not prototypes) of the desired functionality, relationships, and usability. Both requirement definition and testing phases are said to be accelerated, developers can cut rework, and time to deployment has decreased in real-world projects.

What traits do these six software forms share? To some degree or another, all six items include an aggressive integration story. This posture is a refreshing change from previous generations of enterprise apps that were built for greenfield deployment yet purchased by established companies with already-complex environments. Whether it's Google's industry-standard APIs, SAP's Netweaver investment, Salesforce's Apex platform, or Serendipity's creative use of AJAX and RSS, there appears to be a general intent to plug things together rather than build siloed functionality (data warehousing appliances may be an exception, however).

In contrast to many traditional applications, software as a service, enterprise RSS feeds, and mashups should, in concept, scale both up and down. If a mashup takes a week or two to build, and then requirements change, is there even a need to call its retirement "sunsetting"? The notion of on-demand functionality has held appeal for several years for precisely this reason, and the combination of a "faucet" on usage, along with throw-away integration, should reduce impediments to change. To this end, one prominent CIO recently told me that services architectures won't save him any money, but not having them would definitely get expensive: in his view, paying the price, in the form of data discipline, to obtain flexibility was well worthwhile.

Finally, payback on initial investment should come quickly for these kinds of software, designed as they are for quick deployment and reasonably simple integration. Google Apps changes less than half of a typical e-mail host for far more functionality, for example. FedEx's Carter said that his mapping functionality would have been potentially too expensive to license, particularly for a speculative effort. This tightening of the connection between cost and value in turn should make the CIO's relations with business unit customers easier to justify. Being able to dial service levels, volumes, and functionality up or down has the potential to improve the chargeback process: instead of presenting business units with lump sums of fixed costs proportionately divided, the CIO can move in the direction of menu pricing for some portion of his or her cost base.

Changing the current model will not happen fast. The life span of application portfolios is frequently measured in decades. Existing cost structures cannot be exited overnight. Appliances introduce their own particular type of lock-in. Many CIOs and other administrators hesitate to move enterprise data off site to a Google or Salesforce. Customer service and support remain question marks, particularly for mashups, which may present users and help desks with unpredictable performance characteristics: is the problem on the desktop, with Comcast's network or the Starbucks wi-fi hotspot, or in some faraway server?

But for all of these impediments, the general direction of the industry is headed toward a blend of customization and standardization, on-site and off-site, lightweight and industrial-strength. The proof is easy to find: the same industry leaders who thrived under the old model are moving to embrace various forms of the new one, whether it's Intel teaming with SAP to build appliances, Dell supplying special-function servers to Ironport, Microsoft's Live suite of offerings, or IBM's support for software as a service. As work, organization, and markets change dramatically, it's fitting that the business toolkit do so as well.