Suppose our contact centre receives the following monthly call volumes, and you’ve been asked to forecast the June figure:
It should be fairly obvious that the June forecast here is going to be 30,000 calls, as the historical data strongly suggests that the centre gets 1,000 calls per day and June has 30 days.
Now let’s take another slightly more interesting example:
|Number of customers||100,000||120,000||140,000||150,000||160,000||165,000|
|Calls per customer||0.1||0.1||0.1||0.1||0.1|
The secret to being able to forecast the June number of calls is to look at the last row, where there is evidence that the number of calls per customer is very consistent. Given a forecast number of customers (noting that our calls forecast is now dependent on someone else’s sales forecast), we should be able to confidently forecast the June call volume at 0.1 x 165,000 = 16,500 calls.
While life is rarely this simple, the underlying idea always holds true: by finding repeatable, consistent ratios in historical data, and assuming that those ratios are likely to continue unchanged, we can make predictions about volumes based on other information about the future that we believe to be true.
What ratios should we use? Typical values will include calls per customer, calls per new customer, calls per bill, or calls per ex-customer as they leave your business. Some businesses will monitor faults per customer and you might then track calls per fault; as the fault rate comes down (hopefully!) over time, the forecast also will reduce even though the ratio of calls per fault might remain static. Which ratios you use will depend on your business and your ability to access the relevant data consistently.
A monthly or weekly forecast built like this will usually be reasonably accurate if you’re not expecting too much, particularly if your inbound volume is fairly static.