Get a bunch of planners in a room and the discussion inevitably turns to forecasting and then to the fact that we just love to do our forecasting in Excel and upload to our WFM systems. Why? Because we believe the WFM is unable to forecast as good as we can in Excel. I was a believer and always argued the case with anyone who would listen. However I have had recent experience to think that WFM can forecast and can forecast very very well.
I’m currently reviewing WFM practices for a client and have had access to a WFM tool I never used before. It has the usual functionality – forecast / schedule / adherence / annual leave and as usually happens the company in question use a subset of the software. As often happens they also have a suite of spreadsheets, indeed complicated and comprehensive spreadsheets containing the forecast.
I was drawn to the forecasting tool and noticed some issues. Tuesday’s were forecast as higher volume than a Monday on one queue – but the other main queue was the opposite. The latter reflected reality. I spoke to the vendor, prodded behind the scenes, excluded some data (usually around Bank Holidays) and suddenly things started to change. The WFM forecast was getting accurate – spookily accurate. Week 1 was 0.4% inaccurate for the week. On week 2 we were within 8 calls (In a 4,000 calls per day scenario) on one day and then a couple of days later disaster spoke. The actuals were within 1 call of the forecast. Given I often ask on courses “who this year has got their forecast right for a day?” and never had a hand raised – I almost achieved the Holy Grail.
Suddenly within a couple of weeks the business perception changed from WFM can’t forecast so we don’t have confidence in something we pay for, to, this is very good. I almost don’t have the heart to tell them that this honeymoon will only last for a few weeks due to the system under forecasting January. Weekly volume will reset after Christmas and increase 25%, and since there is only a year of historic data since a major operational change, it is only natural that the system hasn’t reacted. It needs a second year of data to be able to identify the reset in volume and not continue with it’s downward trend. But the week and month pattern seem logical, and I hope right.
Using a WFM again ‘in the flesh’ instead of just seeing it in action when visiting clients has challenged me. The forecasting algorithms are improving and improving fast. This particular WFM is using hundreds of algorithms to analyse historic data and find the method that suits each clients individual circumstances. I guess that comes down to the pure power of computer processors now compared to just 10 years ago. But that doesn’t excuse us from helping the software. Some reminders…
- Mark special events – we do billing on 1st working day of the month and I marked every 1st working day as a special event, and also every 2nd working day as a separate event. The software reacted and is now projecting both of these.
- For Bank Holidays – remember to mark at least one day after as well.
- Look for cultural events that may affect your customer behaviour – or indeed your agent behaviour – as their absence can lead to repeat volume. Our staff and even our customers love Halloween – not a public holiday but I mark the days around it as volumes drop.
- Know when to reset the base volume – especially if you have had significant operational change.
I’m a convert – I have a WFM predicting my weekly, daily and even intervals with better accuracy than any spreadsheet I could design. Let’s give WFM a chance, use it alongside our spreadsheets and live the efficiency gains of good initial forecasting.