3 lessons learned to improve your forecast accuracy on disrupting events
Recently a lot of disruptive events, like the pandemic, have occurred which many of us, myself included, didn’t see coming. The cost of doing business shows a strong increase driven by the higher energy - and raw material -prices. Despite their advanced business intelligence, many businesses weren’t prepared, and are now facing unexpected challenges to overcome.
Instead of asking ourselves “Why didn’t we see it coming?”, I would like us to challenge ourselves with the question of “How can we see it coming?”.
Learning from the past will help us to react faster and more effectively in the future when new disrupting events occur. Therefore, I would like to reflect on how we could take disrupting events into account, in order to increase our forecast accuracy. My key answer to this is that we should expect the unexpected, keep our forecasting up to date, and move to long-term thinking.
Lesson #1: Expect the unexpected
There is always somebody in the room that has critical arguments on how you think the market will evolve based on your forecasting models. Sometimes we put these arguments aside because they tend to be based on gut feeling rather than factual reasoning. Think about what could happen when we expect the unexpected and be more open to these critical arguments. Crowd forecasting uses the wisdom of people inside your organization to predict the future. A prediction poll brings personal expectations together and then smart algorithms consolidate and optimize everyone’s guesswork into a reliable collective forecast. Watch out for the signals in the noise that all the data is giving us.
Sentiment measurement can give you an insight into what people are worried or excited about. Consumer sentiment is a strong indication of future behaviour, for example if we are very excited about the newest innovation, we are more likely to embrace it. Work out different scenarios, including ones with more disruptive events than expected, and build different forecasting models to be prepared for the unexpected.
Lesson #2: Keep your forecasting up to date
When a disruptive event occurs, ideally you would respond as fast as possible based on your up-to-date forecast. In essence there are two foundations that together hold the predictive power of your forecasting models. On one hand you will need good quality data and on the other hand you will need a strong algorithm to run on this data. In many cases we see that datasets are updated manually and not frequently enough. This increases the risk of human mistakes and outdated data to start with.
Automated data integration with regularly scheduled ETL jobs will keep your datasets up-to-date. You will always have the most recent data in your data warehouse to work on. We also see that methodology seldom changes, and that similar algorithms are being used as they were two decades ago (see Xavier Marti Audi's article).
These old algorithms lose their predictive power, since times are changing; especially, when disruptive events occur. Think about revising your algorithms, using new methodologies like machine learning and artificial intelligence. They self-calibrate when the market context evolves, and help to keep your forecasting up-to-date.
Lesson #3: Move to long-term thinking
A manager once told me that the 1st of January is just one day after the 31st of December. By this, he meant to say that little has changed after the calendar year has ended. Although in business we are very much focused on yearly targets, especially when we want to achieve quick results to build our personal careers, disruptive events don’t fit this timeframe; since they build up over time and have long-term effects once they occur.
Long-term thinking is therefore more important than ever, and we should remember our long-term strategy, even when the annual results might be disappointing. Long-term trends need to be included in our forecasting models, by incorporating external drivers and macro-economic trends.
A disruptive event like climate change seems to be gradually developing but think about the long-term impact, when you are running an energy company or selling air conditioners. Sometimes, we even need to consider external drivers that have indirect impact.
The demand for cardboard packaging has tremendously grown, with the rise of e-commerce, and cardboard has become scarce. Thereby e-commerce growth has indirectly impacted raw material prices for toilet paper manufacturers and book publishers.
Finally, many forecasting models are built on monthly or even weekly data which makes it challenging to predict the long-term future on these short-term analysis periods. Consider the use of longer time periods in your datasets or transform your datasets in order to make them more consistent over time.
Do you want your own success story?
Contact Engagement Factory today!
Engagement Factory has years of experience in consulting and developing marketing strategies that engage and delight your customers, no matter your industry or challenges.