Forecasting your supply chain and role of AI

Forecasting your supply chain and role of AI

By Rohan Rendalkar

Supply chain management (SCM) is the game of demand and supply. Goal is to increase the demand and fulfill it with adequate supply, hence avoiding stockouts and overstocking that may lead to wastage. Actioning on the demand post order cut-off window may lead to stockouts because of shortage in supply or loss due to higher cost buying (To fulfill the demand).

Hence, it’s very important to forecast your demand based on multiple trends. Overstocking, to and extend is not a big issue in the non-perishable items, but it’s a big issue in the space of perishable items. Dealing in F&V sector makes it very important to foresee your demands and plan the procurement accordingly to achieve that perfect balance between demand and supply. Foreseeing the demand upfront also helps to optimize your cost and improve margins.

Forecasting not only help in the maintaining the fruitful balance between demand & supply, but also helps in other factors of SCM like operation management, warehouse planning, delivery optimization and more.

Forecasting plays an important role in supply chain management, but the question arises on the accuracy of the forecast, better the accuracy, better the balance.  Technology can play an import role on helping to achieve that accuracy as close as possible. Factors that play an important role in the accuracy are, but not limited to:

  1. Historic data: More the historic data, better the chances of achieving that accuracy. Good amount of historic data exposes multiple trends in forecasting, giving better results.
  2. Selecting the right data: All factors that can impact your demand prediction, that should be the ask to be getting closer to that prediction accuracy.
  3. Narrow down data: From the lot of SKU’s choose and concentrate on those which contributes to higher volume or value. Analysis says 15-20% of your SKU lot contributes to 90% of your volume. We can take care of that remaining 10% volume and 80% of SKU’s in traditional way.
  4. Forecasting technology: Technology has taken a big leap in forecasting, earlier using basic algorithms and rules towards self-learning AI & ML models. Choosing the right AIML model based on your data is the key.

Though all the above factors are self-explanatory but lets deep dive more into the forecasting technology.

Technology plays important role in analyzing and providing better forecasting results, but one key is human factor that plays an important role in achieving the desired results. Choosing the right AI forecasting models in itself is a tedious task and it takes multiple iterations to choose the right AI model. The result of these iterations is compared with the actuals to detect the % error and then the model is adjusted & tunned to reduce the error %.

                                                 *Figure is for reference only

Ninjacart F&V supply chain model extensively make use of the AI models to achieve the desired demand forecasting results. Our journey started with 60-70% accuracy and have reached to a level of 95% above.

At Ninjacart we use multiple AI models, and choose the top performing ones. We further consolidate the outcomes of the top performing model to perform adjustment and reduce the error %. These models are trained at regular intervals to keep fine tuning them and improve their self-learning capabilities.

Key reasons to use multiple models also includes the variety of SKU’s we deal with. SKU’s dynamics are different in terms of perishable/non-perishable, their selling patterns, seasonality etc.

Hence at Ninjacart we train 10 different time series models which was a combination of different univariate and multivariate models and built an ensemble model on top of them to get the final demand for each SKU’s.

With the initiative of Ninjacart Tech Venture, we are interested in joining hands with similar start-ups across globe and help them with our tech and learnings, which is built with a strong base of 8 Years.

Interested to join us?

With our increasing global community of Agri-start-ups and partners, we are making our technology, knowledge & experience open to all. And we have started to see the fruition of our endeavour in the Latin America and South East Asia regions. To know more, write to us at or visit our page