Ways to Mess up your Capacity Plans

By Ric Kosiba, Ph.D.

I have a dubious claim to fame. I may have seen more capacity planning spreadsheets than anyone on the planet (except Jen Dziekan). In the larger world, say at a cowboy bar in Nashville for 362 days of the year, that claim would sound like a most boring description of a career. But at the Omni Hotel in Nashville, March 4-6, that fun fact could make me actually sound pretty interesting. (Make sure you come to the SWPP conference this year!!)

I thought I’d relay some items we noticed when looking at spreadsheets over the past 20 years or so. Note that these aren’t the universe of issues we noted, or even the worst we’ve seen. They are simply some mistakes that come to mind after so many years.

Before we start discussing common problems with capacity planning, I’d like to make a couple points.

First, capacity planning is a critical workforce management process. When you think about it, your expensive and time-consuming workforce management software serves, basically, to manage the number of agents you already have on staff. It is the capacity planning process that works to determine how many agents you have available week-over-week. Get this right and you have exactly the right number of agents trained and ready to go. Get it wrong and…

Second, big spreadsheets are hard to manage, and capacity planning spreadsheets are among the most complex spreadsheets in all of business. A proper capacity planning spreadsheet involves linking many metrics, many formulas, and so many assumptions.

But also, a proper capacity planning process involves

(at least) three separate classes of mathematical models. First, using historical data, we forecast volumes, handle times, various shrinkage categories, attrition, wage rates, etc… Next, we simulate the contact center to determine the number of agents needed and the service levels, ASA, occupancy, and abandon rates at various staffing levels. Lastly, we need to optimize the hiring and overtime levels of our agents, and allocation of contacts to our outsourcers. You can get a Ph.D. in forecasting, simulation, or optimization. These models, properly built, are not simple. From experience, I know it is difficult to be an expert in any one of these classes of models, let alone, three. But if you do these three steps well, your contact center will run more smoothly and at a lower cost.

So here are some common problems seen in many spreadsheets.

Learning the Process

In one of my introductions to Erlang-based spreadsheets. The aforementioned Jen Dziekan and I were working with a large, household-name company, to help them convert their spreadsheet to our cap planning software. To do our task well, it was important for us to understand the spreadsheet and all of its calculations and assumptions.

Jen found a function in their spreadsheet that was a complex Erlang model, but when she finally figured out how the model worked, she found a weird calculation. The staffing equations basically expressed: “Staff required is equal to the results of the Erlang C staffing calc, which is a function of volume, handle time, and the service goal. Plus 15 FTE.” So, she asked, “What’s with the plus 15 FTE?” The analyst turned white and told us that he had inherited the spreadsheet three years ago, and never noticed that before. When we looked at the other cells, there were more fudge factors like that that added maybe 50 heads to the budget (or $3MM).

Many, if not most of us, inherit a complicated spreadsheet and a critical planning process from the analyst who came before us. This can lead to some sticky situations, as we usually do not have much time to learn it in between inheriting the process and being required to use it day-to-day. But it is still very important to understand it well. To be an expert, I suppose you could either build your own new cap planning process, or you can do the hard homework of understanding all the equations and inputted data in the inherited system. Either approach is not easy.

Not Planning Often Enough

On my LinkedIn page, I recently wrote about a time when I worked for a large airline whose annual budget was the law. The process did not allow for any changes within the year, even when the business environment was known to be shifting. It led to all sorts of political and operational problems. Frankly, it made the annual budgeting process way too important and time consuming.

I would argue that capacity planning updates and meetings need to happen monthly, but cap planners should make sure that their plans are scoured and analyzed as they get new data every week. As demand for your operation’s services change, the capacity planning process is a place where your organization should notice important fluctuations first. The cap planner should be ready to propose updates to staffing or other reactions to the changing operation when these inevitable shifts occur. The capacity planner is the operation’s best big-picture business analyst.

Planning Too Often

I once worked at a bank, where the operation’s EVP waited impatiently for a new long-term forecast every day at noon. What was scary was that the variability of the performance data was such that the forecast could actually bounce significantly, one day to the next. This EVP focused on the delta of that day’s data relative to that which was forecasted the day before to make decisions. Big-picture decisions could be made without enough due consideration of the natural variation in the operations performance.

This is no way to run an operation — one day’s data should not alter a forecast.

It is important that we as strategic planners understand the difference between a trend and variability. We need to have the courage to ask our senior management to be patient when they might be pushing for a decision right away. Don’t develop new long term plans every day or week.

Straight-Lining Seasonal Variables

I have always been fascinated with the idea that much of shrinkage is seasonal. Not always, but there are often categories of shrinkage that have seasonal peaks and valleys. Sick time, for instance, can follow the weather in some locations. Agent attrition often follows seasonal ebbs and flows.

It is always a good practice to try and capture these, and to forecast them. If you assume shrinkage is a single percentage week in and week out, when there is clear seasonality, you add error to your plan.

All Locations are Not the Same

In a similar fashion, you can gain accuracy by isolating the shrinkage at various locations (including home offices). While you might not be able to discern seasonality in rolled up data, you may be able to note and track seasonality in different locations.

I was also taught, early on, that different locations may have different cultural personalities. For instance, an early customer of ours noted that in some locations it was easier to ask for and receive overtime from agents, and in others it was easier to get agents to go home early. Knowing these characteristics of each location’s agents and using them to help shape your staffing levels is a winning strategy.

Making Staff Levels Tie Out

There are two important week-over-week staffing metrics. The first is the number of agents on payroll, and the second is the number of bodies-in-chairs agents. This is a simplification, but the number of agents on payroll should be equal to the number of bodies-in-chairs agents plus shrinkage.

It is important to make sure — in your history — that the two numbers are exactly tied and that you can explain all of the non-productive time. In history, all hours should be accounted for, so you can be sure that your shrinkage is exact. If those two measures are historically not tied together by validated and accurate shrinkage rates, then there is a math discrepancy that will make the strategic plan fiction.

Does Your Planning Model Tie to Reality?

The first thing they teach you in modeling school (I used to joke that I was a super-model-er) is that your model is proved to be accurate. I think that is something our workforce management industry has forgotten. Have you ever seen any accuracy validation associated with your staffing calculations/models in your WFM software? I suspect not.

The way a validation works is very simple. You plug in last week’s actuals and see if the model predicts service levels and abandons accurately.

Everyone assumes that Erlang C or Erlang A or whatever your WFM vendor or planning vendor calculates is accurate.

But I’d bet they are off, maybe significantly. For workforce management software, it isn’t quite so important, because we use it to allocate the agents we have already hired. But for your capacity planning software, having an accurate model is critical — it drives our budgets.

I will make this a topic in one of the next On Target articles.Anyway, I am looking forward to seeing you all in Nashville!

Ric Kosiba is a charter member of SWPP. He can be reached at kosibas@comcast.net or (410) 562-1217. Please know that he is *very* interested in learning about your business problems and challenges (and what you think of these articles). Want to improve that capacity plan? You can find his calendar and can schedule time with him at realnumbers.com. Follow him on LinkedIn! (www.linkedin.com/in/ric-kosiba/)