Simulation Explained

Is Your Contact Center Capable?

Most of today’s retailers have some sort of contact center department which supports customer service and sales. This is not surprising with the shift of customers shopping via their home computers and smart phones. Retailers now need to ensure their contact centers are capable to compete within this multi-channeled world.


Agents must be able to take service and sales via phone, email, online chat panels and even social media. As greater emphasis is placed on the contact center, it becomes ever more the critical factor to delivering the company’s reputation, achieving projected sales and meeting customer satisfaction. This is where a simple simulation model can provide key insight into the design and operational effectiveness of your company’s contact center.


Why Simulate your Contact Center?

As customer demand increases and shifts towards multi-channel capabilities your company might be faced with an outdated contact center structure.
The dilemma remains the same: calls come in and must get handled within an appropriate amount of time. The difference is that there might be greater volume and that this volume might come from various contact channels. This is why many companies are revamping their analogue contact center departments and internal processes with digital routers.


Despite these changes most of the performance metrics, such as inbound volume, service level, abandonment, agent occupancy, and average answer speed amongst others, remain the same. With this change in customer shopping habits, upper management should be asking these questions:

1. Is our contact center capable of multi-channel orders?
2. What is our priority when it comes to multi-channel orders?
3. Do we need more agents with additional contact channels?
4. What is our current contact channel volume?
5. What is the trend via contact/order methods?


Let’s take a look at a typical contact center that might be widening its multi-channel usage. We will start with an example of a retailer with an online catalog currently using 5 to 8 contact agents. The agents are capable of servicing customers and taking orders via phone, email and through online chat panels. All forms of contact methods can be categorized according 1 of 4 types of contact: New Orders, Order Status, Returns and Questions. For example, when you place a phone call you often get a recorded message specifying such a menu which categorizes the call. Utilizing a digital router it is possible to have categorized queues feeding each contact channel.


In this example we have two routers sending contacts to agents; one designated for email and the other for phone and chat panel communication. Therefore, we can prioritize how contacts are handled by respective agents. We might place phone calls ahead of chat panel communication, which in turn is considered higher priority than emails. Knowing your company’s internal performance metrics will definitely help identify the design of the priority scheme. For example, the desired Level of Service might be that 80% of calls and chat inquiries are answered within 1 minute and within a working day for emails. This metric is closely related to abaondonment, where the customer drops the contact inquiry if they are waiting too long.


Design an Efficient Multi-Channel Contact Center with Simulation

Like with most simulation models we start with the process. Most contact center’s overall processes are quite straightforward. Contacts, in the form of new or existing customers with a specified type of inquiry, enter the system upon a certain arrival rate, such as an exponential distribution, or through an internal empirical dataset. In our simulation (see Figure 1), these contacts arrive either via phone calls or online. The categorized contacts then can be queued and routed to agents upon various conditional statements.


Figure 1 – Contact Center Simulation Model


Each agent, or group of agents, might follow a certain set of operational guidelines (an Agent Flow Chart) for handling different types of contacts. These can follow a complex set of process flow patterns to enable the modeling of Skill Based Routing. In our case study simulation they are captured within a unique sub-window for each contact channel with all agents handling all types of contact. As we can see from Figure 2, questions can be routed back as new orders capturing the agents’ unique ability for making sales over the phone. Finished contacts are then either routed to the warehouse or to the appropriate “Work Complete” object to track and collect vital statistics on the volume of each individual contact type.


Figure 2 – Phone Process Agent Flow Chart



Figure 3 – Agents On Shift Dialog Box


In this simulation we can also readily change the number of agents via a simple user dialog box depicted in Figure 3 to facilitate the testing of varying staff allocation. This is particular can support contact center directors in their Work Force Management and Optimization. The agents assigned work an 8.5 hour shift each day and the contact center runs Monday to Friday and is simulated over the course of 4 weeks. By incorporating further simulation modeling it is also possible to:

  • customize agent shift patterns
  • vary the volume and distribution of incoming contacts
  • designate groups of agents to specific contact methods or types
  • supplement the process with additional contact channels
  • construct specific or more detailed agent flow charts


All of these variations to the simulation can be used to test “what if” scenarios. The results of these scenarios can then in turn be used to perform Call Handling Analysis by comparing the recorded performance metrics. This allows contact center management to hone in on the best use of their agents in order to maximize the contact center’s performance metrics!


Contact Center Performance Metrics

Most of the important metrics can be captured within a table and displayed on the simulation model’s desktop. Below Figure 4 shows the resulting metrics for a 4 week period with 5 agents on duty. We can see that 16,229 contacts came in with a Level of Service of only 75.6%, yielding an abandonment of 24.3%. However, we are exceeding the target of 80% agent occupancy, with an average of at least 122 contacts per day for all agents. Additionally, in the same table we can see the average contact time for each of our agents on shift. We can also dig further into the results and see the statistics per channel, such as the average speed to answer. This shows that the call and chat panel speed to answer is on average about 45 seconds while emails are being attended to after roughly 136 seconds.


Figure 4 – Scenario 1 Performance Metrics


To investigate how we can improve our service level we can now make alterations to the number of agents available on-shift. Our question is: what would the performance metrics look like with all 8 agents working? Using the Set Number of Contact Agents button we can make this change and then simply rerun the simulation. Well… we can see in Figure 5 that our agents are now easily meeting our target with a service level or 93.4%.


Figure 5 – Scenario 2 Performance Metrics


We can compare all of the previous metrics but the key one that we should notice is that the agent occupancy is now not meeting the 80% target. This trade-off is one that contact center management will be familiar with and one that can be further investigated using this simulation model. We invite you to use the simulation to find a suitable balance between Level of Service and agent occupancy by testing different staffing allocations and comparing the performance metric results.


Concluding Thoughts…

Whether your contact center has 8 or 8000 agents the simulation model and analysis would remain similar to the depicted case study simulation. For larger contact centers, analysis could be reported on shorter intervals in the simulation. Most of today’s digital contact center software and routers will provide a vast amount of data that can be easy imported to the simulation. The benefits of providing a simulation is to leverage the data at hand and dial-in your contact center algorithms and improve the utilization of your agents.


The simulation can enable the contact center team to stay ahead of the curve when it comes to maintaining customer satisfaction. As multi-channel ordering dynamics continue to evolve and change over time, it’s important that your contact center reflects what customers are demanding with respect to placing orders.


Want to Try it Out?

Download the contact center simulation.

Don’t have SIMUL8? Get in touch to request a trial.


Learning Points

Explaining how we use Exponential and Poisson Distributions when it comes to contact arrivals


There often is confusion on which distribution to use for incoming contacts which often follow a “Poisson Process”. We actually use the “Exponential Distribution” for the inter-arrival times within our “Start Points”. This distribution is actually the time between events within a Poisson Process. If we were to take some time interval such as an hour and actually count these events we would come up with a discrete value. If we collect all of these counts for the duration of the simulations run it would fit a discrete Poisson distribution. Theses counts are captured within the spreadsheet “ss_Ph Call Counts Poisson” and can be imported into Stat::Fit to build and visually display the distribution of the arrivals.