If your organisation is still using the same data forecasting tools and techniques that you were before the COVID-19 pandemic, it’s time to think again.
This pandemic is what we call a Black Swan event, characterised not only by its unpredictability but also its rarity and severity. However, the problem is not how to predict black swans, but how to forecast accurately after one has occurred.
As businesses scramble to adjust to life after COVID-19 – the ominous new normal, and what follows, now referred to as the ‘next normal’, forecasting data will be an essential component of every business survival strategy.
More than ever before, companies need enterprise data like, what is our bottom line? How many layoffs can we really afford? Are we missing out on opportunities to take advantage of new market demands? Leaders need to know what best and worst-case scenarios the business may be facing.
And the answers lie in the past, but more so in the future. In solutions that enable planning for new demand and supply, and managing risks in an unpredictable climate. Survival depends on innovation.
The limitations of algorithmic forecasting
Many forecasts, particularly algorithmic, rely on historical data, analysing past trends, averages and seasonality to produce a statistical prediction of the future. And that is precisely where the problem comes in – forecasting tools that use a historical-based algorithm.
To illustrate, let’s take a look at Vizlib’s predictive forecasting tool, an advanced analytics capability in Vizlib Library. This solution provides a sophisticated statistical prediction of the future without the need for any coding or complex installation.
Vizlib’s predictive forecast uses an algorithm based on the popular Holt-Winters method, which takes into account seasonality (when a variable typically goes up and down at different times of the year), the overall trend, and the average. For example, the sale of air-conditioning units in summer compared to the winter season.
However, what happens when seasonality goes out the window, or rather, down the toilet?
The recent global spike in toilet paper sales had nothing to do with seasonality and everything to do with the global pandemic. Which is why the Holt-Winters method is no longer helpful in the near-term as it assumes that history repeats itself and models accordingly.
Predicting business needs based on past data with Vizlib Line Chart’s Forecasting feature
How do you forecast data accurately in the ‘next’ normal?
Businesses need advanced analytics solutions that consider this new status quo. The accuracy of their long-range forecasting depends on it. The current models, using historical data, fall into the black swan chasm and are rendered useless from a BI perspective. And that’s where methods such as linear regression and scenario analysis enter the playing field.
To provide users with direct access to these forecasting scenarios, without the need to rely on data science teams, Vizlib has released two exciting new features.
Let’s take a closer look at these two methods and how they can impact your business.
1. Linear regression forecasting
A linear regression model looks at the relationship between two variables (the date and your chosen metric) by plotting a trend line through the observed data. This type of analysis works best when there is a known relationship between the two variables, for example, between age and height, sales and advertising, or travel and term-time holidays.
And, though this approach still pulls data from the past, it is a simple model and can be useful in guiding decision making. Why? Because you can define or limit the historical data to a particular period such as two months, making it more accurate than algorithmic forecasting.
It is best suited to forecasting data that continues at a constant rate, such as fixed income or costs. Data teams can also use this forecasting model to compare different outcomes. For example, How would we fare if we charged all customers a flat rate, vs charging them for usage?
Businesses can also use the linear regression feature to show two linear lines based on 6 months data compared to 18 months to benchmark the performance of various metrics. By plotting a regression line, it helps visualise the difference and where data “meets in the middle.”
Predicting a ‘best guess’ case with Linear regression forecasting in Vizlib Line Chart
However, considering the accuracy limitations under the current conditions, this model may not be the best choice for supporting major business decisions.
2. Exponential forecasting
Another approach is to look at different possible growth scenarios using compound or cumulative growth. With this feature, you can set up different future scenarios yourself based on various assumptions, and work your way back to what that would mean for the business today.
This type of forecast is particularly effective for visualising and comparing financial forecasts or different investment outcomes as they accumulate month-on-month or year-on-year.
These two new features also enable users to forecast in several different ways directly in Qlik, without the need to export into Excel.
Predicting growth scenarios with Vizlib’s advanced analytics capability in Vizlib Line Chart
Managing uncertainty with scenario analysis
Whether you’re leading a team or an entire business, you need a sense of stability and control in uncertain times. Your decisions affect not only the livelihood of the business but also that of each worker. That’s a lot on your shoulders.
And the best analytics tool to support these big decisions and help the business evolve and grow is scenario analysis. This innovative approach looks to the future and does not rely on historical data for input. The inputs come from the knowledge, experience and foresight of the organisation’s leadership.
With scenario analysis, leaders can evaluate and compare best and worst-case scenarios to manage potential risks. And in post-pandemic times, it is the best approach, when machine learning offers lesser quality input.
Vizlib’s new forecasting solutions for more accurate planning in times of uncertainty
For example, in the military, scenario planning helps with the successful handling of any unlikely situations. Because they can anticipate the consequences of every event. In instances like this, scenario planning can switch the odds between life and death.
Or in agriculture, farmers use scenarios to help them predict whether their harvest will be good or bad, depending on various factors, like the weather. As a result, they’re able to forecast their sales and future investment needs.
Both of Vizlib’s new forecasting tools now enable users to plot multiple scenarios, such as the best case, worst case or growth case scenarios, either on a linear or exponential scale.
Particularly when evaluating a course of action or a new decision, managers can leverage this tool to test plausibility and planning when facing a high degree of uncertainty.
For example, in the Vizlib chart below, we look at the absolute number of COVID-19 cases in the UK compared to the cumulative growth rate since the lockdown first started and after easing measures were introduced. This helps to show what a second wave might look like compared to what happens if we continue with social distancing.
Scenario Analysis with Vizlib’s Advanced Analytics for more accurate planning and decision-making
To ensure your business remains future-proof in a COVID-19-authored next normal, harness solutions that provide accurate data and are easy-to-use. Nobody can read minds, but your BI should help you predict and plan for success, even during times of uncertainty!