In pre-pandemic days, the answer to the question “Why should I care about digitalisation and artificial intelligence?” was very often “predictive analytics”. This refers to the ability to harness the big data being produced by an organisation to predict the future, through the magic and science of data mining and mathematical processes.
Predictive analytics was posited as an essential future capability for businesses that would allow them to spot trends, identify the relationship between behaviours, and forecast events and market conditions, accurately and reliably. This would be with the intention of enabling companies to grow their businesses by being prepared for challenges, mitigating risks and spotting opportunities at the right time. That was the carrot.
The stick, on the other hand, was that if you didn’t unlock the predictive power of your data, your competitors certainly would unlock theirs and leave you trailing behind, by an ever-widening gap.
Pretty much every sector stood to benefit from predictive analytics: airlines could predict passenger numbers, energy companies could predict power demand, retailers could offer consumers what they wanted when they wanted, and manufacturers could avoid downtime and lower costs by accurately predicting when parts would need servicing or replacing.
And then came COVID-19. The impact of this has, as we are all aware, been wide-ranging and has, to a great extent, undermined predictive analytics. Predictive analytics relies on the ability of machines to spot patterns in huge amounts of data, better than we ever could, and then learn from these patterns to accurately and reliably forecast the future. But today, any predictive analytics efforts are impaired by two to three years of atypical data. And when things normalise—hopefully in 2022 or 2023 depending on vaccination rollout and its ability to adapt to variants of the virus—the data generated at that point is not going to pick up where 2019 left off. Indeed we will, in all likelihood, have a new baseline from which to reset our future expectations as things may well have fundamentally shifted since 2019.
Take travel. Now that it’s clear that it is possible that business can be carried out around the world without getting on an aeroplane, is it likely that business travel will return to pre-2020 levels? The realisation that long-distance travel is not always a necessity will intersect with businesses’ sustainability goals, and companies will no doubt also realise how much money they can save (a win-win). A shift away from pre-pandemic levels of corporate travel will have a knock-on effect on hotels, car hire, venue hire and a host of satellite businesses that supported these trips.
On the other hand, a sector such as eCommerce, which has undoubtedly received a boost from the pandemic, is likely to see a lot of this new business remain. During the pandemic, companies accelerated their eCommerce rollouts and improved their offerings, and many late adopting consumers tried online shopping for the first time due to lockdown and health concerns. No doubt some of these consumers will return to bricks-and-mortar stores when they can, but, I would suggest, it is highly likely that many new converts will continue to do at least a portion of their shopping online, thanks to the convenience and wider choice.
The big question in both these cases is how many businesses and people will make the change permanent and to what extent, and this is what potentially undermines the predictive power of pre-2020 data.
But today, all the data we have is outlier data, in other words, not reliable as baseline data upon which to predict future outcomes. We’re in uncharted territory here, hence my position that 2022/3 will present as a new baseline. If we need to wait to build up a baseline of “normal” data to feed into predictive analytics models, how do companies plan today?
Accountants actually have a model for this already: zero-based budgeting. Taking a zero-based approach to budgeting and forecasting using financial modelling, companies can side-step predictive analytics’ reliance on historical data. Instead of looking at last month’s or last year’s data, and extrapolating from that, companies can look at their own predictions, coloured by their own experiences and expectations, relevant to their unique set of circumstances.
Another benefit of this approach is that your financial model is now rooted in reality, your numbers can be tested against a myriad of flexible assumptions, at very granular levels, and across your entire organisation. These best estimates may paint a picture of what your tomorrow may look like in a somewhat uncertain future.
Yes, this is, to an extent, a back-to-basics approach and it may even be a bit more manual. But when you have no historical trend data to rely on you, unfortunately, you need to start from scratch and this will undoubtedly mean challenging the established norm.
As published AccountingWeb - March 2021
Comments