For the next post I am really pleased to have the pleasure of introducing a guest author; Blair McGavin. Blair is an absolute expert in his field, both from a WFM system perspective but also comes with first hand experience running a workforce planning team. I mentioned in my latest article, My workforce planning journey – the early days that I am starting to see a few WFM software developers push the boundaries and suspect that we will start to see major upgrades in the years to come, and the following article really zero’s in on that thought. By the way be sure to look out for a couple of follow ups on Capacity Planning which Blair is working on. Doug
What is the FUTURE of WFM….? Or do we first need to go back and fix the PAST?
Which of the two puzzles below do you think represents the complexity of modeling staffing requirements and service levels in your contact center?
The WFM Industry would like us to believe that it is as simple as the puzzle on the left: Calculate some historical averages, plug them into a formula, and Voila! Here is how many agents you need, and here are your service level forecasts.
REALITY tells us that it’s a bit more complicated than that. Have you ever wondered why Actual SL’s are always so different from the original forecast in your WFM software. There are 4 basic pieces to the staffing model process, but is the industry ‘managing’ ANY of them with enough accuracy and detail to nail the forecast? Do they even bother to reconcile each piece of the puzzle so that we know how much each contributed to our service level variance?
Since the inception of the WFM software industry, the Erlang formulas have been at the heart of determining staffing requirements and forecasting service levels. Have you ever tried to reconcile actual service levels with what your software forecast? How frustrating! It takes time and effort, raises some questions about which numbers to use, and rarely turns out well. Erlang was a great mathematician, but unfortunately these models just don’t work very well in our complex contact centers. Why is that?
- In it’s one dimensional, single skill environment – Erlang overstaffs (can easily be up to 10%)
- Erlang can’t see the efficiencies that occur when agents are multi-skilled (oops – it’s 1 dimensional).
- Erlang doesn’t account for abandons (luckily customers don’t abandon in your CC).
- And then there is schedule efficiency, which also has an effect on our occupancy – super difficult to calculate.
Even if you found a queue where the implied assumptions behind Erlang held true – which is rare (Poisson distribution for arrival rates, and exponential distributions for Handle Times and Mean Time between arrivals), the model just doesn’t work very well. And is it realistic to think that ALL of your queues would behave consistent with the same rigid assumptions? NO, it isn’t. And ‘MY’ queues probably behave differently from the queues in ‘other’ contact centers, so why does the industry think that the same model would work for every customer.
What we need are customized models that reflect the ‘unique variability’ in each of our queues and skill groups, and that can recognize the efficiencies when our agents are multi-skilled, and even working in multiple channels.
So, you have probably already reached this conclusion. If we can’t trust the WFM industry’s forecasts for staffing requirements and service levels, then should we have any confidence in their forecasts of full FTE requirements (capacity planning)? Nope, we shouldn’t, and the problem isn’t just with Erlang. There are 4 basic pieces to a capacity plan. All that information resides inside our WFM software. Yet, they still struggle to piece it all together to tell us our true and accurate FTE requirements. If it were easy, it would have been figured out by now.
It has become incumbent on the end user to build and maintain their own capacity plans. Every WFM team is doing it differently – which means different assumptions, different definitions and different formulas – which lead to different requirements. Beyond Erlang, error is also being introduced with how the WFM industry forecasts shrinkage and handle time. Unfortunately, the methodologies that forecast every piece of this puzzle either need to be redesigned or enhanced to improve their accuracy.
If there is a word that is overused in the WFM industry, it is the word ‘Optimization.’ Every vendor says that they ‘optimize’ their schedules – and since there is only ‘one’ way of doing things, they are all coming up with the same ‘optimal’ set of schedules. Right? Wrong!!
How optimal your schedules are depends on a handful of obvious variables:
- Days and hours of operation
- The types of schedule templates being used,
- And most importantly, the scheduling optimization methodology being used by your WFM
The proof is in the pudding
The two graphs below show the ‘optimal’ results from two different scheduling tools. Do you think that one would be more efficient than the other? Clearly.
If our WFM software can generate an efficient set of schedules relative to a flawed and inaccurate forecast, have we solved the problem? Not yet. We first need to improve the accuracy of forecasting how every piece of the puzzle behaves at each interval, THEN generate a more optimal set of schedules to match our more accurate and complete forecast.
Let’s say that your schedules are more like those in the 2nd graph – less efficient. Does that have a cost associated with it? If so, how do you quantify that cost? One more missing piece in why this industry is unable to fully support us in staffing our contact centers.
Service Levels vs. Financials
We sure spend a lot of time talking about service levels and staffing requirements. “How much are we over / understaffed?” That’s our job! And we take pride in that process. We spend a lot of time crunching numbers trying to provide those answers to our management team.
But more important than meeting Service levels is our company’s profitability:
- Are we making any money?
- And can our contact center be run more efficiently without compromising quality and service?
- How could WFM software help with that analysis?
When we sit down with management to review last month’s financial performance, do we really understand what is causing our budget variances? It’s easy to say, “volumes were higher than expected,” or “we had 23 business days in the month.” It’s just NOT that simple. There are dozens of variables that drive the budgets of our contact centers, most of which ‘should’ be managed and reported on by our WFM software, but they are not => Lack of vision.
“You can’t manage what you can’t see”
What if we assembled the staffing and capacity planning puzzle so well that it would enable a financial reconciliation at the end of the month. This would allow us to itemize, by variable, how much each piece of the puzzle contributed to our overall budget variance, including unit cost variances by skill. Imagine having 4 or 5 automated and well-designed reports that outlined exactly which variables caused our financial variances by site, by department, by skill, and even by agent.
“Sounds futuristic ☹. We can’t even determine what our staffing requirements need to be.”
If we could see detailed financial reporting that showed us where our money was going – we would know where to focus our efforts, and easily manage out the inefficiencies.
- Customized staffing models for every skill? – No Way.
- More efficient schedules? – “But they said they ‘optimize’ our schedules”
- An accurate Capacity Planning tool? – “I don’t buy it, it will always be our burden”
- The ability to reconcile SL variances? – “Is that even possible? It’s a ‘random’ process.”
- Integration with the financials and automated reporting so that I can see what caused my variances –“I need taller boots, it’s getting pretty deep. Come back to earth.”
As a frustrated consumer in the WFM industry, I have always had a couple of questions in the back of my mind:
How come we haven’t demanded better forecasting?
Why don’t we have the ‘vision’ to connect WFM with the financial planning and budgeting process?
Because ‘they’ haven’t figured out how to assemble the puzzle yet, and it’s light years more complicated than any vendor has given it credit.
What if all this functionality had already been designed and proven, and was just waiting to be developed? Would you want to be a part of it?
Doug and I welcome your thoughts and feedback.