An interview with SkuCaster’s Adam Toledano: An AI-powered approach to inventory forecasting in the automotive aftermarket

SkuCaster, an inventory forecasting company, employs a team of professionals specializing in predictive analytics, including data science, machine learning and statistical modeling.

SkuCaster, an inventory forecasting company, employs a team of professionals specializing in predictive analytics, including data science, machine learning and statistical modeling. Their focus is on providing inventory forecasts and monitoring solutions to businesses, helping them make informed decisions.

Adam Toledano, the CEO of Skucaster, has a background that spans 10 years working with various industries. Before establishing Skucaster, he built experience in software, AI, and consulting, and worked as an FP&A analyst for multinational corporations, such as Deloitte. His work included forecasting projects aimed at optimizing organizational spending.

Now leading Skucaster, Adam applies his experience to help supply chain teams enhance their forecasting efforts, minimize forecasting bias and transition to a more proactive planning approach.

How does SkuCaster’s demand forecasting differ from traditional ERPs and how does it enhance the reliability of supply chain management in the automotive aftermarket?​

SkuCaster demand forecasting is different from traditional ERPs because we don’t just factor internal data but take an outside-in planning approach by factoring market data such as Vehicles in Operations (VIO) Data, Point of Sale (POS) data, Parts Data, Customer Data, Vendor Data and apply our proprietary machine-learning/AI models on that data set to generate accurate SKU forecasts.

How does SkuCaster utilize artificial intelligence?

Our models are auto training models which select the best performing forecast models from multiple machine-learning algorithms. Our AI forecasting model continually improves itself every month, thus improving the accuracy over time of the SKU-level forecasts.

How does SkuCaster incorporate a company’s specific knowledge of products, market trends, and cycles into its machine learning models for inventory forecasts?​

For every AI model that we create, we sit with different stakeholders of that specific distributor, including the business teams, and ensure that we consider the specific inputs unique to that business and specific inventory parts (customer segment behavior, SKU specific behavior, etc.). Those inputs are then transferred into our AI model.

For market trends, we factor all third-party data, like macro-economic data (inflation, climate/seasons, employment by region etc.), vendor data and VIO data to complement a company’s internal data.

Given that automotive aftermarket can be quite dynamic, how does SkuCaster’s AI-powered forecast models adapt to rapid changes of demand at the SKU level?​

Our models are not manually trained only once. Our models are automatically self-trained (monthly), and thanks to the inputs of our AI models, we already account for the changing nature of the aftermarket industry.

How can SkuCaster help automotive aftermarket businesses prevent low fill rates and avoid penalties associated with them?​

Low fill rates is a problem when actual demand is higher than the forecasted demand and thus low supply occurs. SkuCaster ensures that forecasts are closer or slightly above the actual demand, hence the users are always in a better position to proactively order enough quantity in advance of leads times so that they do not encounter low fill rate issues.

Can you walk me through the process of automating inventory forecasting with SkuCaster, from data collection and cleaning to accurate forecasting?​

Step 1: we collect the data from various databases (customer’s ERPs, vendor data, etc.)

Step 2: we “cleanse” the data (i.e., contextualize it according to the client’s business model) so that it can be ready to be trained with our AI models. At this step we must ensure that there is zero gaps in the data.

Step 3: we manipulate the raw data so that we are able to create multiple features of the forecasting algorithm (i.e., seasonality, bias, anomalies, accounting, trends, SKU behavior, customer behaviour etc).

Step 4: We train our proprietary models and do multiple rounds of test and validations to select the best performing AI models.

Step 5: We productionize the AI models so that the forecasts can run automatically without any manual intervention and is not impacted by normal IT issues.

For more information on SkuCaster, visit the company’s website.

For the latest news and information on the global automotive aftermarket industry, visit Do you have news? Contact Aftermarket Intel Editor Mark Phillips at He’s on LinkedIn here.

Is your company exhibiting at an upcoming show? Let us know!

Join your colleagues in the global automotive aftermarket. Get the Aftermarket Intel Briefing, edited by Mark Phillips, AAP.

* indicates required