The Erlang formula has stood the test of time in providing a framework to predict staffing requirements for optimal operational efficiency in contact centers.
There is no denying that it is a tried and tested predictor for linear customer journeys, but does it have the flexibility to remain accurate when even the slightest complexity comes into play along a customer journey?
It’s been suggested that the formula is now ineffective in contact centres with more complicated workflows, often over-estimating staffing requirements by as much as 60-70 per cent. However, worryingly, there is still a propensity of reliance on the formula in new and existing prediction software within the industry.
This doesn’t need to be the case, as new approaches in digital simulation are removing the restrictions of the formula to instead develop very realistic models of how a contact centre operates in the real world, incorporating randomness and variety.
To illustrate this point, let’s discuss five scenarios when the Erlang formula won’t help you:
1. Fixed time blocks: Over-estimating capacity needs occurs when using the formula because of the restrictive parameters of using defined blocks of time in which calls can take place. Let’s say that a call comes in at 15:57 and runs until 16:09. An Erlang formula that is set on ten minute blocks will account for this call as running over two blocks, from 15:50 – 15:59, and from 16:00 – 16:09; a ‘12 minute call’ registers as an ‘up to 20 minute call’.
2. Feedback loops: Once a single customer abandons a call and then dials back in, the formula means that they become two separate data points when in reality they are only one. This creates a double data feedback loop that immediately makes the prediction inaccurate. Add to this that there will be a range of time distributions before a caller tries again, whether it’s 30 seconds later or the next day, and the predictions from the formula are once again inaccurate against the real world.
3. Case management: When the complexity of an enquiry increases it often needs numerous individual call operators to help from different departments. The formula makes no allowance for such a prevalent occurrence. As soon as a call is passed to another department the resource use increases but isn’t tracked by the formula. For larger global brands, complexity will multiply further when they require planning across multi-site, multi-lingual and multi-purpose contact centres to all cater to a single brand user experience.
4. Firefighting: The formula may allow for planning around typical peak calls times, such as first thing in the morning or over a lunch period, with the intention to shift resources accordingly. A common occurrence when a contact centre houses several different departments in one place, however, is to use a strategy of shifting resources around to firefight any unexpected and overwhelming spikes in demand as well. Unfortunately, while it resolves an issue in the short term, it can also create knock-on problems. Repeating the strategy means that the problem is just being shifted and chased around, resulting in constant firefighting. A better strategy is to incorporate these additional random spikes into planning by building them into a digital simulation model.
5. Staff experience: The formula makes no allowances for variations in experience and competency levels within the workforce. A new starter and a highly experienced call hander are an identical data point.
The simplicity of the formula has served contact centres well in the past, but in today’s modern world where we have the technology available and grasp of data to evolve this forecasting method into more realistic process analytics, it would be foolish not to adopt and take advantage of these helpful tools.
About the Author
Liam Hastie is a business consultant at Simul8 and an expert in call centre capacity planning that uses advanced digital simulation technology.
