Let's not get caught up in the numbers

I recently had the misfortune of a 3 hours delay on a 40 minute flight. Luckily years of travelling a few times per month has changed my attitude to such inconveniences and I was able to take advantage of the delay to catch up on my reading.

I had brought Ger Koole’s latest, excellent, publication – An Introduction to Business Analytics with me, and I thoroughly enjoyed being challenged around how we use numbers in the business. What surprised me most, was in a book with some serious mathematical methods in it, it was a comment with a contrary view that resonated most with me. I’ll come back to some of the mathematical methods in a future blog – but for now let’s look at the antithesis.

In one chapter Ger comments that the forecaster who relies on mathematics will over time become less accurate than one who takes a more multi-faceted approach to forecasting. For some of us that is a bold statement and challenges that ultimate quest for the perfect algorithm. We want that perfect forecast, one which the operation can’t rubbish, can’t use as evidence of why us forecasters don’t know what we are doing – all feedback that contact centre forecasters hear all too often.

As I have worked and assisted in many different contact centre operations, I have more and more become convinced that a good forecaster has the ability to move beyond the spreadsheet and the numbers. We all know that WFM systems have limitations in that their forecasting algorithms take the numbers at face value. If we received 3,000 calls last week, 2,000 the week before and 1,000 the week before that – then the obvious forecast for the week ahead is 4,000. Without intervention of some sort the algorithm won’t necessarily spot that last week was pay day, or the country had severe storms, or some reason which seen a volume surge that is not going to re-occur in the week ahead. Yes, this is oversimplified as most systems have been designed to learn over time and identify trends – but that only if everything else remains the same. How many businesses have not introduced or cut products, changed their support offering or even changed the support channel mix over the past few years? These changes challenge the ability of software to compare ‘apples with apples’ and maximise their accuracy.

So what should a good forecaster do? I personally believe it comes down to one question: “Why would someone want to contact our business today / tomorrow / next week / next month?” When I was working with Ulster University to design the BSc (Hons) Customer Contact Planning & Management (incidentally the only BSc programme globally offering a career related qualification to planners – contact us if you want more information) – one tutor, Arthur McKeown (https://www.linkedin.com/in/arthurmckeown/) used to talk about the macro-environment in which we operated. It took me some time to realise what he meant. However it is quite simple – a good forecaster understands the world in which his business operates.

A good forecaster reads the news, stays aware of the environment in which his or her business (and their competitors and companies in their business’ supply chain) operates. They also think about the social aspects of the world around them and how it may effect customers and potential customers. At this stage we all need to work out what Brexit is likely to mean to our business, whether we operate in the UK or not. We may be lucky enough to have analysts looking at this within the business, but like many others they are looking at what Donald Rumsfeld would have called an Unknown Unknown (https://en.wikipedia.org/wiki/There_are_known_knowns) Something we couldn’t have predicted happening a few years back, and something for which we can only guess at the impact it may have. But as a forecaster we need to account for it happening in what may be only a few short weeks from now. Gut feel and guess work (educated guess at least) will have to come into play as it is unlikely many of us will have any relative analysis to fall back on.

Another example is the recent Bank Holiday in the UK – the warmest Bank Holiday at the end of August in history. I feel I can confidently say that no WFM system or Excel algorithm could have predicted the impact on contact volume over that weekend. There was no data to rely on – maybe with global warming we can start to gather data for the future – but it probably meant an unusual inaccuracy reading for many of us. Yes those working in the 111 Health Service lines would have guessed that volumes would go up due to an unprecedented number of calls for advice of dealing with sunburn. Similarly many other centres would have realised it may be quieter than usual as people decided to enjoy the last days of summer. But the fact that this was a weather related volume deviation means we could not predict it more than 5 – 10 days in advance. No-one could have built it into the annual budget.

So this brings me full circle to the initial premise of this blog. A good forecaster has to reply on more than their numeric ability, their skill in building complex and detailed algorithms to predict the future. A good forecaster will understand the wider macro-environment and be able to interpret what they see in the world around them and made an educated decision on what that will mean to volume.

Bring the human element into our forecasts and we will increase accuracy over time.