Sunday, 29 January 2012

Volume Forecasting

All good WFM planning starts with an accurate forecast so for my very first post this is as good as any to start with.

Today I am going to specifically talk about volume forecasting, a critical section for deriving a workload forecast that can be used to plan against.

The WFM forecasting process, like any planning process, is one part art one part science. It is an art because sometimes the accuracy of your forecast will be as a result of your judgment and experience. It is a science because there many step-by-step mathematical processes that can be used to turn raw data into predictions of future events.
Oh and a pet hate of mine is people who believe buying WFO software results in not needing to understanding forecasting. It is just as critical for you to understand these calculations as it is for someone who is doing it by hand, firstly so you can quality check the inputs/outputs and secondarily so you are able to have an educated discussion with business leaders. After all, if you are not able to explain how you came up with the forecast in the first place how do you expect operations to buy-in and support your WFM plan…

There are three main forecasting methods used for Customer Contact Centre forecasting:

  1. Historical Patterns – the focus of today’s discussion
  2. Cause & Effect – a method best suited for a businesses with regularly characterized ups and downs due to causal factors or drivers.
  3. Guesswork – yes sometimes when there is a lack of accurate data gut feel is all you have. Actually in many cases using gut feel can be a better method than using data alone (the artful side of forecasting) but more about that another time….

Today I am going to take you through the main methods used for Historical pattern forecasting, however it should be noted that best of breed uses a combination of all three methods.

Data Gathering
The source of your data of course depends upon your technological set-up and type of work i.e. Front Office – Call Centre or Customer Back Office – processing unit. However, the point I would like to make here is about quality… As any stock broker will tell you that history is not necessarily a true reflection of the future. So make sure you look out for abnormally low or high numbers, missing information and trends that you know will not repeat. For example, data taken during a special event such as the World Cup is not likely to repeat during the next month.

Devising a Prediction for Monthly/Weekly Volumes
Now you need to turn your raw data into something useable for a capacity plan as well as something that can be further refined for scheduling purposes.

Here are the three main methods used, each has its pro and cons depending upon the situation and time available. At later date I will explain in more depth the pros and cons of each…

  1. Point Estimation – The most basic method, that makes the assumption that a future point will reflect a corresponding point in the past.
  2. Averages – The are a number of sub methods that can used here including; simple averaging, moving average. ranging and weighted averages. My recommendation here is to use weighted averaging, putting more weight to recent events than older.
  3. Time Series – This method is derived from the assumption that volumes are influenced by a variety of factors over a time period, with it being possible to isolated each factor to enable it to be used for forecasting purposes. Without trying to get too technical here, in its simplest form a time series is a collection of data points obtained through repetitive measurement over time. There are normally three main components to a time series: long term direction, calendar related movements and short term fluctuations. It is a common method used in call centre forecasting and many WFO software packages use it as the basis of their forecast modelling.

Refine to a Daily and Half-Hourly Forecast
The aim at this stage is to devise a set of percentages for intra-day volume patterns that represent customer opening hours as well percentages for each day of the week. These percentages can then be used to phase the monthly/weekly forecasts into daily and 60/30/15 min intervals which are useful for staff scheduling purposes. Using an average method here is generally acceptable but it is also good to mix in known casual factors, for example for a sales line it might be known that marketing will be running a special promotion on a particular day.

That’s it for today, please feel free to ask questions. Comments and feedback are very welcome.