.png){kind=logo}
This article is the third in a series focused on how Remi AI deploys demand forecasting to help clients improve their operations, increase revenues, find optimisation opportunities and reduce wastage. In recent years, deep learning, probabilistic programming and machine learning have delivered significant improvements on traditional time series methods in generating accurate forecasts. In e-commerce there are additional data streams that can be utilised to leverage these new approaches and to deliver even greater accuracy.
The need for demand forecasting in e-commerce business is ubiquitous. It’s a fast-paced business environment that needs accurate planning and regular adjustments to react to market changes. With buyers able to easily compare prices across competitors, Amazon changing customer expectations around delivery times, and the challenges of conversion attribution, Demand Forecasting needs to be leveraged to ensure you meet customer expectations.
At Remi AI, we’ve delivered Demand Forecasting for many different types of E-Commerce Businesses. At the stage of writing we are currently running daily forecasts for over 100 million different products across the United States, United Kingdom and Australia.
Typically, we leverage demand forecasting for two critical planning processes:
Inventory Management
Many E-Commerce businesses see up to 70% increases in conversion rates when a product is in stock and expected delivery is less than 30 days. Simultaneously, many E-Commerce Companies are utilising International Suppliers, which means potential delays in supplier deliveries.
With these two factors (keeping stock on hand and managing supplier lead times) to be considered, E-Commerce businesses should be leveraging AI demand forecasting to help support their replenishment decisions each week. Accurate and regular forecasts allow companies to make better strategic replenishment decisions, have fewer stockouts, and more satisfied customers.
Driver and Staff Planning
Amazon, as mentioned earlier, has changed expectations around product delivery times. This creates difficulties for other E-Commerce Operations — especially those that manage their own deliveries. If you are understaffed, your deliveries schedule will fall behind.If you are overstaffed, you run the risk of wasting money on unnecessary wages. Accurate demand forecasting for your delivery team will support them in their strategic decisions and also help you optimise your FTE costs.
The uniqueness of demand forecasting for E-Commerce
There are two unique factors in establishing good demand forecasting for E-Commerce businesses:
Data sources
Unlike brick-and-mortar businesses, an E-Commerce store has many digital triggers that can be fed into the forecasting model to help improve accuracies:
- Product Page views
- Which products are on the Homepage
- On-site search across Product Categories and Brand
- App Page Views
- Saved Items
These five, and many other triggers can all be leveraged to more accurately understand upcoming demand. This is especially powerful for expensive, low-velocity products in your category range, I.e whitegoods. It is uncommon for someone who is looking to buy a new fridge to complete the purchase on the first visit to the site. Instead, they’ll look at a few products in your fridge range, maybe print off dimensions to check that each will fit in their kitchen, and come back a few times before they complete the purchase. To ensure you can make the most of this important data, our Platform directly integrates with Google AdWords, Google Tag Manager and also the E-Commerce Platforms, allowing you to quickly setup AI Demand Forecasting across your store.
Competitor Pricing
While technically a data source, the power of knowing this is important enough to differentiate. Although it’s rare to know very far in advance what your competitor’s pricing will be, it should still be leveraged to better understand short-term demand. If you’re not utilising AI price optimisation, like Remi AI’s Dynamic Pricing, it still makes sense to at least be factoring in where your products are priced in comparison to your competitors. This helps you predict the likelihood of sales in the coming week.
Finding the best forecasting model for each of your products
Our Platform comes with 22 different AI models, each with their own unique strengths. Through working with both small and large business, from low to high velocity goods, we have come to understand the power of finding the best model for a particular data set, instead of just relying on classic, standalone forecasting methods. With access to any combination of models that are relevant to your store, you’ll have the confidence to make smarter business decisions.
If you’d like to learn more about Demand Forecasting and how it can help your business, please let us know.
Or if you would like to just jump into the APIs yourself, then signup!
