Plan for the Possibilities

By Ric Kosiba, Ph.D.

I know it that has been a while since the SWPP conference in Nashville but, oh my, was it fun and productive. Kudos to Vicki, the SWPP Board, and all her many helpers. What a great event!

Themes and T-Shirts

When I was at Bay Bridge, every year we had a theme, a catchphrase that made it into our presentations and onto our awesome customer t-shirts. One year we had a theme “Fast and Accurate” and the t-shirt was a cartoon World War 2 fighter pilot diving with guns a-blazing. Another year we had a cute Disneyesque theme and shirt, “Plan to Infinity and Beyond” for our user’s conference at Disney World. But my favorite was “Plan for the Possibilities” because those words encapsulated a whole new way to look at strategic planning and managing business risk.

When chatting with Tricia Payne from American Express we came upon an idea. When we are forecasting and planning, we are usually trying to find the probabilistic best forecast. We can use it to staff to hit the probabilistically likeliest scenario. But is this right? Maybe we should understand our business risk under various possible or likely scenarios.

I had previously seen a demonstration of a very sophisticated long-term forecasting system. Part of the time-series forecasts they produce include a 95% confidence interval, which I believe shows that if they simulated the forecast over and over, 95% of all outputs for every data point would be within that wide range. I don’t think this is particularly helpful, because this confidence cone becomes super wide quickly. In other words, the forecast has very wide probabilistic variability.

If you pointed to the top of the interval and ask “Is this super high volume data point even possible?” your considered answer would be “probably not really.” However, the confidence interval says that the outcome is as likely as any other point in the confidence interval. Why the disconnect? This is because the math that is used to produce this forecast and uncertainty prediction only has time-series history as its guide, whereas the intuition we all have includes years of experience and countless other data points (say, a discussion with marketing and sales) that we internalize. The math has less information than we have.

Whenever I chat with folks about forecasts, I make sure to tell people to trust their gut. I’ve created enough math mistakes to know that simply eyeballing a graph and looking for weird curves is helpful. Our eyes can see output that looks out of place and is likely wrong. Asking a common-sense question to challenge an analysis can be super helpful in finding problems with the math. Our instinct is a tool for understanding what is possible in our plans.

Quantifying Uncertainty

This year at the fabulous SWPP conference I had a gentleman who worked for an insurance company ask for help in understanding how to build capacity plans when their demand spikes were driven by weather events. What an interesting business problem.

His “regular” forecast followed nice and consistent seasonal patterns, but if there were major weather events, their phones would light up. So we discussed how often these events happened, whether their frequency and intensity could be quantified, and whether there were sources of history for him to pull from. Weather is impossible to forecast months in advance, yet there is a lot of data on severe weather events in his area, and it is very possible to be able to develop probabilities of events happening.

For instance, we could find for the last fifty years:

• How many severe (defined, say, as hail-producing) weather events happened in every season, every year.
• The probability of 0 to 15 severe events happening in a single season.
• Staffing strategies (agent overtime budget, outsourcers ready to take calls, other channels) for the various weather scenarios, using his capacity planning spreadsheet.
• The service cost of choosing the wrong level of storms to build staff plans for.

When you had this information, it was simply a matter of presenting the probability of seasonal events happening, and their likely cost. The decision-makers could look at the cost versus risk trade-offs and choose the strategy that they felt good about, using both the math and their intuition.

In effect, we were using the old “built for the 1000-year flood” analytics that civil engineers use when building a dam or a bridge. It is a lot of work, but it would certainly help answer that question.

Economic Uncertainty

Similarly, all of us have economic uncertainty to some degree that we need to think through when building a strategic/capacity plan. Having been a planner through many severe recessions, I have seen individual companies face wild swings in demand for their product and contact centers.

An approach to modeling tumultuous scenarios is almost like wargaming a plan. Of course, we start with the standard forecast and staff plan — this represents the status quo if nothing were to change in our environment.

But what if we notice a swing in demand for our contact center? How would we respond?

A good capacity plan would allow us to mock that up, and would allow us to pretend that these changes to our demand would start to happen. The question is, when would we notice it, and how quickly could we affect a response to it?

We could:

• Develop the status quo plan.
• Add (or take away) volume in the future to represent the possibilities associated with an economic shift. Use your best judgment, there is little math to rely upon (except any previous examples of similar events).
• Guess as to how your organization would react. Would you start hiring right away if demand increases? Would you wait and see? Try and mimic your company’s decision-making.
• Using your planning process, determine how bad service would get or how overstaffed you’d be.
• If you model that your response would not be sufficient, go back to the first step. Is there something you could do today that might help you recover better (e.g., sign a new outsourcing partner, or run a little fat for a while?)? Are there changes to your organization’s decision-making process that would help?
• Present all of this, and all assumptions to senior management.

Some Thoughts

If anything, our recent experience with Covid and our contact center response taught us is that much of the math and standard operating procedures can go out the window pretty quickly. How do you forecast when history is no longer a guide? Automatic forecasters become sort of useless during and after these sorts of events. And after the event, we have a huge hole in our data that makes it hard to develop time series forecasts from.

At this time, judgment and experience take over. We plan for the possibilities.

Ric Kosiba is a charter member of SWPP. He can be reached at ric@realnumbers.com 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/)