How many heads do I need?

The Art and Science of Forecasting Support – Part 2

In a previous article entitled “How busy are we going to be?” (SSPA News, November 16, 2004), I discussed how to create a reliable forecast for case volume. The next step: transforming the case volume forecast into a headcount plan. Having a reliable headcount plan is critical for creating a budget for support organizations since people expenses typically account for 75% or more of support budgets. Furthermore, if you’re faced with difficult decisions about whether to outsource, you need to start with a good headcount plan so you can accurately compare your options.

As I pointed out in the article about forecasting case volume, it’s both tempting and counter-productive to over-engineer forecasts. Remember that the goal is to create a forecast, not a detailed schedule of who will cover the phone at 3:00 pm on Tuesday. If you find yourself agonizing over a couple of heads, you’re probably overdoing the forecast, at least if your organization is reasonably large.

Start with the case volume
If you followed the steps in the first article, you have a forecast by month or quarter, organized by categories that are meaningful for your business (whether they’re regions, products, new customers vs. existing customers, languages, etc.). Each category should correspond to a unique staffing need.

Figure out utilization
If you’ve never delved into utilization figures before, you may be shocked by some of the figures. You don’t get eight hours a day from your staff members. After you remove vacations, sick days, administrative time for meetings or email, and training days, the number will be more like six hours a day, seven if you’re lucky.
You can probably use the same utilization figure for everyone on the team. Remember that this is a forecast so don’t use each individual’s vacation allotment; stick with averages. You may need to use different figures for workers in Europe who get many more vacation days than U.S. workers. If your support staff is involved in non-case related work, decrease the available hours accordingly.

Figure out case productivity
Here comes the interesting and potentially tricky part of the forecast: determining how much effort time is required to resolve a case. Most support centers don’t keep exact metrics for effort time, so you’ll have to use an estimate. My preferred method, assuming that support staffers work exclusively on cases, is to simply divide available hours (from the previous step) by the number of cases resolved, over as long a period as you can muster to avoid normal wobbles. This top-down approach may be crude but it has the great advantage of a built-in reality check. If you decide to use a bottom-up approach instead, compare your results against the top-down number as a sanity check.

If you use a tiered process, you also need to estimate the percentage of cases that will percolate to higher levels, and how long each case will require, on average, at each level.

This is the time to take into account any planned process improvements. For example, if you’re beefing up the internal knowledge base, you should expect some improvements in case productivity. Be cautiously optimistic: it’s the rare support organization that improves productivity by more than 10% per year – although it is possible, especially if you’re starting from a very low figure. Roll in the improvements gradually in any case.

An interesting situation occurs if you’re planning to beef up self-service support. In that case, your case volume should decrease, perhaps spectacularly. However, at this step you should plan for a decrease in case productivity, as easy cases will be filtered out by self-service and you will get fewer but more difficult cases.

Calculate case-related headcount
At this point, you can calculate headcount for case-related work by simply dividing the labor required (case volume x effort time) by available hours (hours x utilization rate). You will have separate buckets for level 1 and level 2 staff, plus whatever categories you need for products, regions, etc. Round all numbers up since it’s very difficult to hire .15 of a person.

If you have a large organization, the numbers you get will be fairly large and probably won’t need adjustments, but with a small organization, you’ll probably need to make adjustments. For example, if you end up with a lone headcount handling product X, it would be wise to add another one as a backup. If you must cover extended hours, the minimum team size for a given product should probably be four, etc. Build the minima into your formulas or adjust manually.

Calculate other customer-based headcount
If your support organization provides services to customers other than case-related, you need to figure out the headcount needed for those by using one of two approaches.

• If you provide a service that’s time-based, for example, if you provide extensive setup and implementation services for new customers, take the new customer forecast (and in this case it’d better be very accurate) and derive the headcount needed for setups just like you did for cases.
• If you provide a customer-based service such as account management, you typically assign a maximum number of accounts to each account manager, so you can derive the headcount required with a simple division from the number of customers.

Add other headcount
The heavy lifting of customer-driven headcount being done, you can move to other headcount. This includes staff dedicated to the knowledge base, tool maintenance, support planning, training, and of course managers.

Although you can certainly build the numbers manually, I recommend the discipline of driving non-delivery headcount from the delivery (customer-driven) headcount. For instance, keep the number of staffers in non-delivery functions to 5% of the total, maybe a little more for a small team. This automatically right-sizes the non-delivery staff as the organization grows.

As the last step, add management heads, using a fixed ratio of staff to manager. 1:10 or 1:12 works well in most support organizations. If you have a big team, figure out second and third-level managers the same way. Continue to round up fractional numbers as before.

Should you worry about ramp up time?
The staffing model as created works well for large organizations with turnover that’s pretty much constant from year to year. This is because the case productivity numbers take into account the lower productivity expected from new staffers. If your turnover should increase sharply, the productivity figures won’t work well: they will be too optimistic (and the reverse is true if turnover should decrease sharply).

So you don’t need to worry about modeling ramp up time unless you expect to hire massively next year. If you are planning massive hiring, perhaps in response to a new product rollout or as you move the support team from one location to another, by all means adapt the model to hire staff ahead of when you need them. Simply take the figures from the model and move them up by as many months as required for your ramp up time.

That’s all there is to creating a support staffing model! Remember that even a crude model is better than nothing, so start small. Review and revise the assumptions in the staffing model on a quarterly basis, using your metrics to confirm and qualify the assumptions you use. And don’t complicate the model too much over time. Once you can no longer explain its logic to an outsider within a few minutes, you’ve gone too far.

About the Author
Francoise Tourniaire is the founder and principal of FT Works, a consulting firm that helps technology companies create and grow their support operations. She is the author of “The Art of Software Support”, a practical guide to running support organizations. For more information, visit www.ftworks.com or call 650 559 9826.