In my last blog, I had talked about agent level dashboards (iDAT), and the vital role they play in the judgment about the 3 P’s (Presence, productivity and performance) of call center agents. Hence iDAT is an immensely useful tool for TLs to asses individual agents based on their scores on various parameters like AHT, Occupancy rates, Phone time variance etc.
In this blog, I will discuss about iDAT’s feature which also includes forecasting Call volumes (could be daily, weekly or monthly) for a particular process/skill set wise/agent department wise etc. Call volumes forecasts are essential for any call center process to get a “night vision” of the expected number of calls in an hour/day/week/month and hence agent resources could be allocated accordingly, so that “supply is always sufficient to meet the demand”. These forecasted call volume numbers are not just randomly guessed , but, they are analyzed from past historical call volumes data, put under rigorous analytical process to fit significant Time series models and in turn precipitate out reliable and statistically significant call volume forecasts.
From what I have learnt doing some past projects dealing with forecasting numbers and volumes, is that the process of forecasting call volumes is a fine blend of creativity (studying patterns and trends in the past data) and technical knowledge (Fitting of appropriate ARIMA/ARIMAX models to the historical data to produce sensible and statistically rigorous forecasts).
In this discussion, I will talk only about the creative aspect of Forecasting call
volume numbers. Studying the various patterns in the call volumes data like, seasonal patterns (lesser calls in December due to Christmas breaks, etc.), weekly patterns (more calls on Mondays and Tuesdays consistently over months), daily patterns (rush of calls in the morning hours, less calls in evening hours), general trends (call volumes are increasing gradually over time), cyclic patterns (which occur over large time intervals, may be 5 or 6 months), and detecting random variations is a test of the analyst’s creativity and art. Accurate analysis of the patterns in the call volumes is very crucial for a robust and accurate forecast. Under/over assumptions about the patterns in Call volumes might lead to unreliable numbers which in turn would result in inefficient staffing models.
So from the above discussion, it’s clear that it is creativity that counts when it comes to deciphering patterns in the past data. Thankfully, the removing of patterns/cycles need not be done manually, as there are standard statistical packages like SAS, R etc which fit a Time series model automatically (provided we identify the patterns correctly and decide on the most probable models to be fit) , which next would lead us to the technical aspect of Call volumes forecasting.
In my next blog, I will discuss about the technical aspect of the call volumes forecastingprocess which would close in the dimension of an analyst’s capabilities to treat historical data of calls and come up with reliable and significant forecasts of call volumes.