Prediction Analysis Conundrum
- This topic has 3 replies, 1 voice, and was last updated 5 years, 10 months ago by Ron.
November 10, 2015 at 9:56 am #55174
If anyone has advice on this, please let me know. We are trying to (been “coopted to” is a better description) do some forecasting for a call center. It took us a while to figure out that the two different datasets we have been given are quite different: The incoming call volume is captured on a monthly calendar basis (business days) by the call center manager. The second dataset is their billable hours which is captured by a financial analyst on the other side of the state. She uses an accounting system called the 4-4-5 system, which divides the year into four 13-week quarters. Each quarter will have two 4-week months and a 5-week month. The 5-week months are March, June, September and December. (As you can see, June and Sept. are two of the four “5-week months” when they are actually 30-day calendar months.) An example: The incoming calls for June are lower (calendar system), but the billable hours are higher (accounting system). They want us to help them forecast incoming calls, and how many FTEs they need to always reach their service level agreements. We’re pretty good at what we do…but we’re not that good. Any ideas how to get these two differeing data capture systems to work in tandem?0November 10, 2015 at 12:52 pm #198993
Shelby JarvisParticipant@ShelbyJarvis Include @ShelbyJarvis in your post and this person will
be notified via email.
I am not certain I understand all the details of your situation. If I understand it correctly, you are seeking a way which to forecast your staffing at a call center in order to fulfill your SLA. Your data is in 2 systems; one customer facing and one internal finance.
If that is correct, I would begin by using the customer facing data. Once you understand the actual demand, then you can map it to your financial plan.
Without knowing the details, I would suggest treating the data in a few methods. (Some of these may or may not apply depending upon the type of data. Consider understanding Takt Time, avg. wait time for answer, number of handoffs during calls, and avg. time to resolve calls. You should also perform time series analysis on the call data to understand if patterns exist. Examples of influencers on patterns may be time of day, week, or month. Others may be special causes like promotional programs etc.
By starting with looking at the data (graphically first then analytically), you may see information which can help you assess headcount vs. demand vs. Service Level Provided.
I hope this helps you get started.
Shelby0November 12, 2015 at 6:40 am #198999
Thanks Shelby. That does help. We were thrown, just ever so slightly, that an accounting system (4-4-5) could carry so much weight in a calendar-observing world. The lady who does the finances is situated hundreds of miles away from the call center and really has no interest (skin in the game) on how busy they are on different months. Her job is just to make the numbers work in her system. The poor call center manager is left trying to explain that just because the finance people think the four “5-week months” of Mar/June/Sep/Dec are charging a lot of hours, it doesn’t necessarily correlate to the amount of incoming calls they report for those months. We will forecast on the calendar, customer-facing, system. Cheers.0November 23, 2015 at 6:05 am #199022
Great guidance in previous comments. I would also check for seasonality in the data. I have found in many projects (not all), it’s best to remove the seasonality manually, run your forecast using the non-seasonal data, then reapply the seasonality factors to the forecast. This should increase your accuracy.0
You must be logged in to reply to this topic.