Big FMCG companies may have a wide range of products in their portfolio, which is making the forecast simulation more difficult with less accuracy. So, what could be the impact of inaccurate forecast on such companies? And how can they benefit from the BI and either AI to enhance this accuracy?
As a Marketer in a FMCG company, I will take this opportunity to briefly share my forecasting experience and process, the problems I used to face along with the reasons behind them, and my analysis on what the solution would be through the progress of technology in this field, and how technology and human would work in parallel to enhance the processes and accuracy in work.
To understand the real problem, I used to face during forecasting process, I will explain what it represents inside the company and why it is so important to enhance its accuracy!
Forecasting is in general an act of calculation within predictions, in other words, to calculate the future! And to be able to do that, we should have the data history, the present data and market place situation, the company targeted growth per product and targeted market share, and expect or simulate the sales to market against market trend, taking into consideration the current stock we have, the stock en-route, the up-coming promotional and marketing events that could affect positively or negatively the sales impact.
What if the forecast was less important than the sales to market need? Therefore, you will have stock shortages not only in your warehouse but also a short in supply to the market, thus empty shelves and loss in sales! And this is a nightmare, since you will miss the monthly sales target, your competitors will take advantage and you will risk losing your brand image!
Let’s now take the opposite scenario, what if the forecast was more important than the sales to market in only 1 month? As a direct consequence, you will have slow moving items in your warehouse, additional stocking cost, slow moving items blocking the space in front of the fast-moving items, thus an unbalanced stocking level.
Not to forget that the forecasting simulation is placed for 4 months ahead as a golden rule in supply chain ; because based on the forecast, Procurement Department should order the raw materials needed for production raw materials need sometimes 60 days to reach the warehouse), then the Production team should produce products and build up the physical stock in order to have a safety stock for Logistic Department to supply the warehouses in different areas and perhaps shipping for different countries!
What if we’re forecasting for a new product to be launched? Isn’t it even more difficult? And what if we’re forecasting for above 600 SKU’s? Can you imagine the nightmare without proper solutions and clear data analysis?
Despite having different quantitative and qualitative methods and approaches in forecasting, they are still not enough to avoid forecast inaccuracy. There must be a harmony between all the processes to ensure a smoother mechanism and accurate execution based on solid data analysis and simulation to get clearer solutions and data readings.
Although having a lot of BI solutions in the market like Oracle, IBM, Microsoft SQL, JD Adwards …etc., that can be used to improve forecasting, it is still the trickiest operation in almost all industries. Market fluctuations, business conditions variation, economic crisis and uncertainty, in addition to shifts in supply and demand that can make from forecasting a more guesswork than science!
So how, through BI and AI, can we optimize this process and increase the accuracy percentage to above 95%?
What marketing and business managers need, is to find a new set of technologies driven by data science; an approach based on predictive analytics as a complementary of forecasting that relies on a combination of data history, statistics, machine learning, data mining and modelling. These processes will enable business leaders to connect data to effective action plans by drawing reliable conclusions and solutions about the current/present situation and the future events.
So, how does it work?
By using data, data and more data from internal and external sources via predictive algorithms:
Internal sources essentially include the company’s internal marketing automation data, sales history by channel by sector by product’s category and SKU, product’s labeling and mapping, current sales situation, stock level data, business plan data and targeted sales growth by SKU, marketing plan data and projected promotional and events activities, projected extra sales deals/new products launching/areas expansion…
External sources consist of data points as market volume, current market situation as market share by product’s category, competition activities offline and online based on seasonality, product’s pricing change, market trend and situation, area/country economic situation, customer’s orientation and surveys to meet customer’s expectations via uploading the surveys data.
Predictive algorithms use data science to spot correlations between thousands of variables (historical data) and the outcome (sales) to predict the likelihood of closing each prospect. These algorithms can rapidly recalibrate themselves in response to emerging patterns of data.
For instance, if a company acquires or undergoes geographical expansion, they can quickly pick up the nuances and adapt to changing circumstances, thereby ensuring the precision of predictions even in the most dynamic business environments.
Performance unearthing actionable insights while gazing into the future is great, the more important issue is to do the internal homework within companies’ departments, not only finding the right data, but to enhance the internal processes and rightly map, label and gather these data to make the algorithm process easier and can help managers to correct various problems in order to easily read the data and improve the solutions.
The main issue of facilitating the data reading and finding easier solutions, is to enhance the business value within the company.
With objective and accurate predictions, it is easier to stay in total control of the pipeline and know exactly what will close and what won’t. This ensures shorter sales cycles, higher rep quota attainment and an increase in average deal size, while reducing sales and marketing costs. Furthermore, it guides smarter decision-making by solving complex business questions in a fraction of time and uncovers new business opportunities. It is for these reasons that predictive analytics is seen by many organizations as an ROI decision instead of a cost factor.
As a conclusion, while it may be galling to discover that a computer thinking in 1second can get a better grip on the data than all our human intuition, one can’t really argue if it works. Predictive analytics is revolutionizing sales forecasting by replacing the constraints of human inference and bias with objective models based on forecasting algorithms.
But at the same time, these models cannot work properly and give us the best solutions and readings if humans don’t do their work properly and provide the right data within an internal business value and synergy within all concerned departments and business operations!