I was talking recently to a top CIO of global two wheeler company. In the discussion, it came out that this company has all the best possible Demand Forecasting tools but still was not able supply as per requirement. He clearly pointed out that the time has come for us to leverage the AI,big data and integrate the all possible feeds of weather( floods, drought, typhoon), forced shutdowns, changes in consumer preferences( Colour, fashion, style ), ecommerce sale…
Have we reached the zenith of Demand Forecasting and can’t go any forward with modelling historical data?
Artificial Intelligence Algorithms can detect patterns and predict successfully into the future basis sales history, forecast history, shipment and inventory history, promotion impact and weather impact, Consumer preferences … to decide the short term Demand to increase revenues by 3-5%. Demand sensing may improve forecast accuracy by 50% and reduce inventory by 30%. There are currently AI tools available which can detect a change in sale in short term 3-5 days at item location level and autonomously update forecast and can help you improve near future sales up to 5%. Demand Sensing provide advance warning signals about the gap between the plan and upcoming situation in the Supply chain, detect demand changes proactively, expedite the information flow. How fast and intelligently we can anticipate the changes in Demand will determine our Competitive Advantage.
We all know how Diwali brick & mortar sale go muted due to Pre Diwali online sales. The market size is almost constant in short term. if you are selling big time in Ecommerce then you will sell lower in the offline sales. Can we have a Data lakes to capture all this data which can be leveraged by the companies as per their requirement? Say customised cut of TBs of data from the Petabytes of data? Is the Data Structured & Stored for meaning full recovery?
If we know the how many people are searching online & offline for brand X marathon shoes, type y & colour Blue? It may also by impacted by how many people are searching for the marathons. Again the conversion of both the searches will also impact the sale. Are we capturing this data? Do we have tools to analyse this big data and convert it to meaningful Demand Forecast?
What is Demand Sensing?
It is ability to predict the Demand data using from all the possible sources of online/ offline consumer buying behaviour, market shifts, natural disasters, weather changes and more in the short term. The short term can be hours, days, weeks or fortnights.
In another discussion with Business head of a top Oil & Gas Company, I was surprised with his frustration on the inability to sense the demand in the near future. This Fortune 500 Company has the best of Demand forecasting tools but still it is not able to meet Demand. His Desperation resonates with CIO of the Global Two Wheeler Company? You can grow your business by 5-10% if you can Sense Demand accurately? Are you still working on the past data from the supply chain to gauge demand? Despite the best of forecasting tools you may not able to achieve the Forecast Accuracy of 50-60% at item-location levels. In aggregate you may have a forecast accuracy of 90% to 95% at Item Level. But the devil lies in detail item location level forecast. Are you capturing the Sales at an Item level from Mom and Pop Store in real time?
Aggregation on historical data completely isolates the latest trends and patterns at an item-DC level. The true demand signal is blocked & latency gets into your Demand Plan.
Is my Supply chain ready to start Demand Sensing?
Let’s quickly answer some Questions.
A) Do you have an Ability to See Demand at an Item Store( Reatil/ MOM & POP) or POS level? This will enable you for iteration of Demand Models leveraging real time data and capture Outliers, Trends, Patterns, and Spikes more sharply.
B) Are you Modelling Demand Confidently? It is important to eliminate the noise from the Demand Data. You should understand the true Deviations.
C) Can your Supply chain planning platform process big data of millions of item location combinations, economic conditions, oil price, weather forecast, POS data, collaborative planning, VMI feeds simultaneously?
An early onset of summer may shoot up Cola Sales. Higher rainfall may cause the washing powder sales to increase. An FMCG Company will need to stock the inventory of washing powder at a place where it’s likely to rain significantly higher. If you can integrate the weather feed into the forecast then late stage inventory reallocation can improve forecast accuracy.
D) Does your supply chain platform have a seamless integration between execution & planning as well as the ability to supply the frequent demand with efficiency?
Higher level process automation is required to ensure that the demand signal is driving the execution environment.
Most likely, you have the data to start sensing demand such as sales orders, Ship-to details, replenishment along with respective line-order details. Ironically, most of the manufacturing and distribution companies don’t have retail POS data, but sensing demand using sell-in data has been found to improve forecast accuracy by 30%.
An Article by Sandeep Anand