Home

Ric Kosiba

Vice President of Interactive Intelligence’s Decisions Group

Last month, we discussed the first five common contact center planning mistakes:

· One: Flat-lining shrinkage is always a bad assumption

· Two: Manual hiring/OT/UT plans are too hard to do right

· Three: Treating all centers as though they are the same is dangerous

· Four: Stretch goals are evil

· Five: Always hitting your service goals may not be optimal

Let’s dive into the next five.

Six: Presenting one plan is a disservice

A concept that we all can buy into is this: more analyses is always better. Given the variable nature of planning, it is always better to do several analyses up front in order to try to capture the risk associated with each one of our planning assumptions.

For example, given all of the business variability over the last few years, it makes more sense to run the following simple scenarios:

1. What is the staff plan associated with the current most expected forecast?

2. What is the risk associated with this plan (meaning what service will we see, and what costs will we incur?) if:

a. Calls come in higher than we thought?

b. Calls come in lower than we thought?

These several plans are much more beneficial to senior management- and may lead to a different resource decision- than the linear “one plan” approach.

But, to be honest, because we are using time consuming and error prone spreadsheets to develop plans, we often don’t have time to develop these cost saving risk analyses. This is a reason to either automate your planning process or to purchase a strategic planning system. You need this process to be quick enough to allow for quick and accurate analyses.

Seven: Weekly versus monthly plans: weekly wins

Many plans, usually those generated by finance, are developed with the finest level of hiring, volume, and shrinkage detail as monthly. But we all know that these decisions are best developed knowing the intra-month patterns.

Blending an average month together will lead to shortages some weeks and overages other weeks, simply because of the assumption in our spreadsheets.

What is the best compromise, if finance likes to see their plans reflecting monthly costs? It is to develop weekly capacity plans, and roll them up (with detail on subsequent sheets) so finance gets your plan in the form they like, but your operational plan includes the detail that the operation needs.

Eight: Long-term questions deserve long term analyses

Every so often, we see a spreadsheet process that only determines staff plans for ninety days or so. While real-world activities, like outsourcer locks, mean that these timeframes are important, they do not preclude the importance of using long-term analyses for long-term business problems.

Many of the decisions our operations make are long-term decisions. For example, hiring is not a short-term decision, or even a ninety day decision. Because of this, we need to understand the implications of our hiring decisions with a seasonal view.

Coming into a seasonal peak is a lousy time to look at hiring a few months out. We might decide to hire to the peak, even if the peak is short-lived and a valley immediately follows. However, it may be optimal for the business to either 1) miss service levels at peak (see number five, above) or to 2) use overtime to staff to peak staffing requirements.

Nine: No variance analyses, means you don’t know what to do next

This is a hard one. Many businesses don’t have time to look backwards, and many operations do not provide meaningful variance analyses. But the process of determining changes to the operation or variance in the business environment is critical to running a smooth operation and to avoiding service catastrophes.

While many businesses will have forecast review meetings, usually the point is to reevaluate the forecast (usually only volume forecasts get reviewed), put pressure on the forecast team, and possibly change the plan. For many other organizations, variance analyses are a luxury, rarely consumed.

The best contact center organizations look at the variance differently. They view variance as a core piece to their planning process and an item that is dutifully performed. These same organizations view variance not as a forecast error, but as a change in their environment. Variance analysis is the canary in the operational coalmine.

Their variance analysis meetings serve a specific set of purposes. They look to 1) track what is changing in the operation’s performance drivers, 2) to determine if these changes require business decisions, 3) fix variance items that are controllable, and 4) make resource decisions about items that are not controllable. These variance meetings are management decision-making meetings (and not beat-up-the-forecaster meetings). They track variance to all of the main center performance drivers, certainly volumes, but also handle times, attrition, sick time, sales per call, etc…

Central to variance analysis is also what-if analysis. It is not good enough to know that an item (like attrition) is changing; it is also important to know what will happen to the operation if it continues to change, or what the resource cost will be to react to this change if it is determined the change is not temporary.

It is arguable, that this piece of the planning process is the most important, and it is enabled by an automated and optimized planning process.

Ten: Validation breeds confidence

One step we rarely see in planning spreadsheets is a validation step. Meaning, if you know the number of people available, the handle time achieved, and the number of calls offered last week, it should be an easy exercise to plug those real world values into the spreadsheet to determine whether the spreadsheet predicts the actual service achieved. If it does, the model is accurate. If it doesn’t (over a reasonable timeframe), then it is not accurate.

Validating your spreadsheet is not as easy as I make it out, only because most analysts build their spreadsheet to determine agent requirements and not to provide service analysis (given agents staffed). My suspicion is that most of the methods employed in spreadsheets (usually Erlang, or an assumed occupancy based workload equation) are not really all that accurate (I’ve tested several and my suspicion always bears this out).

Validation of the model that determines your requirement is very important. If you know that your method is accurate under different service standards, and this is something you can show your management team, then they will have confidence in your plans, your analyses, and you.

How about I make you an offer; call me up, and I will offer to take a peek at your spreadsheet. Perhaps together we can find sources of hidden value that you never knew were there.

Ric Kosiba, Ph.D. is vice president of Interactive Intelligence’s Decisions Group. He can be reached at Ric.Kosiba@InIn.com, (410) 224-9883 or Twitter: ric@decisionstech