[NOTE:  The following article was available on CMP's TechWeb, December 1996... but has since been removed.   I have placed in on my site because I referenced it in my How Many Administrators are Enough? article, and lots of people asked me about David's article.]

Know The Types: Sizing up Support Staffs

By David Cappuccio, Gartner Group Inc

One of the questions corporate networkers like to ask concerns LAN support staffing. Specifically, they want to know how many people are needed and what their skill set should be. The answer? It depends. It depends on a number of key factors, from the technical sophistication of users to the number of applications that are being run to whether the network is centralized or decentralized. It depends on whether or not there are service-level agreements with departments, and on what service levels need to be met. In short, there's no simple answer to the question.

The first thing corporate networkers should keep in mind is that cutting support staff will actually increase costs. Sounds odd until you think about unqualified end-users tackling complicated problems themselves.

Does this mean it's time to devise a complicated formula for figuring out how much bigger the staff should be? Not exactly. Instead, corporations might want to look at their end-user and server environments in terms of technical sophistication--and then develop staffing plans around each user or server type. Just break them into easy -to-remember categories according to technical implementation or utilization--such as Type A, B, and C.

End-Users and Abusers

Take a look at the Type A users first. They're the ones continually pushing the leading edge of technology--so-called power users like traders, systems developers or engineers, and, for lack of a better word, geeks. They require (or desire) the newest and fastest systems.

Contrary to popular belief, sup-porting these kinds of users is tough. They're technologically sophisticated and handle most problems themselves--so that when they do go to the help desk it's usually with a problem that's truly bizarre and difficult to isolate, duplicate, and resolve. Diagnostic and resolution times are high, affecting the levels of service given to other clients and forcing down the overall ratio of technical staff to end-users. At the companies we've surveyed that have Type A end-users, support ratios of 1-to-30 are about average.

Type B's are users--not abusers--o f technology. They typically have a common suite of applications on their desktops; use servers for database, decision support, and workgroup applications (e.g., Notes); and, in many cases, have access to department-specific applications, intranets, and the Internet. They're skilled at their jobs and generally view technology as a means to an end.

And as such, they don't engage in the kind of peer support that Type A users do. Service-level agreement response times are critical for these users, since any time lost to technology problems could mean lost business. Companies that want to keep their Type B's happy generally have a staffer-to-user ratio of between 1-to-60 and 1-to-100.

The typical Type C user is a general office worker with a desktop system installed, a standardized suite of applications, access to file and print servers, and in most cases, access to some host-based legacy applications. These customers are the easiest to support, at least after the initial introduction to technolog y. Ratios of 1-to-125 are common.

Serve Somebody

Figuring out support requirements doesn't end with categorizing the users. The ratios addressed above are for support staffs performing Tier 1 or desktop support functions. Tier 1 personnel are business-focused and concentrate on end-user desktop and applications support. They normally can solve up to 85 percent of the day-to-day problems without having to call on their technological superiors--the Tier 2 personnel.

Assuming that Tier 1 workers take care of what they're supposed to, Tier 2 staff can concentrate on technical work, and a similar metric can be developed for this group. Instead of focusing on end-users, this support staff should focus on network infrastructure. Consequently, the target ratio is no longer staff member per user, but servers per staff member.

Levels of labor

Use the same ranking scheme that we used above for categorizing users. In this case, traditional file and print services can be categor ized as Type C's; they're the simplest to support, with ratios of 20 to 1 not uncommon.

This ratio changes with the introduction of higher-level directories and automation tools. Application and Web servers require support that's a bit more focused, and ratios are typically 12 to 1. While a Web server looks and feels like a file-and-print server, the ongoing maintenance of Web pages (and all of the associated links) is labor-intensive right now. Once complete sets of Web management tools become available, these servers could drop to Type C status.

Finally, there are the database servers. These are considered Type A; they generally require the most ongoing support.

The bottom line is that the most cost-effective and high-performance support organizations are those that make sure to focus on business solutions first (yes, we support end-users, not technology), and therefore position the most appropriate staff members at the right functional levels. Network engineering experts don't staf f help desks, and PC support people don't handle network operating systems.

Once this repositioning occurs, and once well-defined job descriptions with the appropriate salary levels are laid out, the overall cost of support will begin to stabilize. Notice I didn't say that costs will all of a sudden start coming down. But at least we can begin to add more desktop and network functionality without adding significantly to support staffs.


David Cappuccio is vice president of the Network Computing Service at Gartner Group Inc. (Stamford, Conn.).