The Call Distribution Object
· Provides a way to distribute calls in a call flow randomly through a switchboard. The number of output options that can be selected from and the weight given to these channels can be modified. The distribution can be changed, but the sum of the options must equal 100%.
Use Cases
· A small sales team wants to make sure inbound calls are shared evenly across their team. They place a call distribution that splits the call volume and gives all the salespeople a fair chance at incoming leads.
· Ability to distribute load to a different contact center if the volume is too high. You could create a distribution that sends 80% of traffic to your main contact center and 20% of calls to a different call center, or ring group.
How does it work?
Distribute calls between 3 sales people
· Call Distribution is setup with 3 different legs weighted at 33, 33, 34. Since the distribution can't sum to 99 we must have one equal 34.
· Call 1: Call reaches Call Distribution and generates a random number between 1 and 100. The number is 53. This call will skip the first agent who has slots 1-33. The second agent is 34 to 66. Since 53 is between 34 and 66 the call is directed to agent 2.
· Call 2: Generates a random number between 1-100. The number is 10. This call is routed to the first agent who covers any numbers between 1-33.
· Call 3: Generates a random number between 1-100. The number is 88. This call skips the first agent 1-33, the second agent 34-66, and routes to the third agent who is covering 67-100.
FAQ
What if I want to have a single call flow of 100%?
This may be unnecessary. However, this may be valuable for situations where you want to have 100% of call flow to one contact center and then be able to failover to another on by updating the distribution.
When I tested my call flow it went to the same leg 2x in a row. What is going on?
This is because the call distribution is not the same as a round robin system. The current call has no knowledge of the previous call, they are completely independent events and new random number is used on every call.
Over multiple calls the distribution will match the distribution that you have specified in the switchboard. The likelihood of a 50/50 distribution randomly sending all of its traffic to one leg or another is very unlikely. The likelihood of a 50/50 getting sequential calls to the same leg is show below.
(probability_of_leg)^number_of_sequential calls in a row
Probability of reaching a 50% leg X times in a row
(.50)^1 = 50% of call reaching leg 1
(.5)^2 = 25% of 2 calls reaching leg 1 in a row
(.5)^3 = 12.5% of 3 calls reaching leg 1 in a row
(.5)^4 = 6.25% of 4 calls reaching leg 1 in a row
While it is technically possible for a 50/50 distribution to send 4 calls in a row to one leg, the probability drops rapidly over sequential calls to ~6.25%.
If one leg of the call distribution is very heavily weighted, (example 90/10) then it may be very likely for that leg to receive multiple calls in a row. This is demonstrated by how a leg weighted at 90% will have > 50% likelihood of getting 6 calls in a row.
Probability of reaching a 90% leg X times in a row
(.9)^1 = 90% of call reaching leg 1
(.9)^2 = 81% of 2 calls reaching leg 1 in a row
(.9)^3 = 73% of 3 calls reaching leg 1 in a row
(.9)^4 = 66% of 4 calls reaching leg 1 in a row
(.9)^5 = 59% of 5 calls reaching leg 1 in a row
(.9)^6 = 53% of 6 calls reaching leg 1 in a row
(.9)^7 = 48% of 7 calls reaching leg 1 in a row
How can I know if the call distribution is working?
The random number that was used to determine which number to direct the call is output in the call log. It can be seen directing the calls between to two different call legs below.