How to get started with Supply Chain Analytics.
If you’re running an eCommerce business, you no doubt already have processes in place to handle common workflows. Things like inventory management, warehouse utilization, omni-channel pricing, and more. But the best companies are leveraging their operational data to optimize the way they make these types of decisions. Whether you’re just taking the first steps or you’re a pro looking to squeeze the last percent of improvement out of your data, we’ll cover some ways you can use data analytics to improve your supply chain efficiency.
Demand forecasting
Unless you’re making orders in response to out of stocks, you’re already using some level of demand forecast for inventory management. There is an incredibly wide range of techniques, however, from “last month’s sales” all the way to deep learning models that take a month to train. You’re probably somewhere in the middle of the pack.
If you’re just starting out, spreadsheet based forecasting will likely give you the best ROI initially. You can set up a monthly forecast fairly easily using a TTM moving average, a 1yr lookback with a growth factor, or other simple techniques.
More advanced analysts will use statistical forecasting and/or machine learning approaches. This is what we do for most of our customers in this space. The Prophet library (open-sourced from Facebook) is a good place to start for easy-to-use baseline forecast models, if your team can support Python code. Most growth from here will involve more detailed ML modeling and more automated processes.
These techniques will give you an estimate of future demand for your products, but it can still take significant work to convert these forecasts into Purchase Orders. You’ll want to use extra consideration if you sell goods that are not shelf stable.
Intelligent warehouse allocation
If you stock inventory in multiple fulfillment centers, you must have some way of determining how much of each product to store in each of them.
Early on, this might just be an even split of inventory across all your 3PLs. But as you grow more comfortable using data to drive this decision, you can apply historical sales data. Using shipping addresses, you can check the distance of historical orders from each of your warehouses to find what percent of orders are closest to each. Then you allocate inventory storage according to those percentages.
The strategy can get more complicated as your analytics process matures. You can start using actual shipping cost tables instead of using distance as a proxy - since, ultimately, we want to minimize the average cost of shipping products. You can analyze the geographic trends of wholesale vs retail orders to make sure you can support wholesale orders without splits. Later you’ll want to incorporate the original cost of transportation to your 3PLs as well, especially if you are importing products from overseas.
With each iteration, your total cost to deliver a product will be reduced, meaning more and more margin you get to hold onto.
Inventory optimization
We’ve covered demand forecasting, which will help you estimate order quantities. But we need to go a step further, because inventory management is a “constrained optimization” problem. What’s that mean? You don’t have an unlimited budget.
The answer is adding an optimization element to your inventory management process. You use your demand forecasts as a starting point, but do some initial analysis on your return on inventory expenditure.
You’ll want to look at pricing and the total landed cost for each product you’re thinking about ordering. Compare the margins. If you face budget constraints, you’ll almost certainly want to prioritize high margin products in your purchase order.
That’s just the start, though. We also have to consider what’s even possible to order - factory specific MOQs. Then overhead costs like shipping containers. Then shipping timelines. Then risk factors like COVID supply chain shutdowns, port delays, raw material shortages, and so on.
To do this properly requires a ton of expertise in optimization. But fortunately, if you’re willing to settle for a slightly less than perfect solution, you can make big profit gains by analyzing some of this data and applying some business logic to make a decision. No probability theory necessary.
Operations analytics
There are endless ways to use your data for improving operational efficiency. If you use a 3PL for fulfillment, you may want to use their data to confirm they’re meeting SLAs. Or analyze product sales to determine warehouse layout and order-picking strategy.
If you’re growing your sales team, you might want to assess the ROI - are sales increasing as expected? How is your revenue per salesperson trending?
Have conversions increased since we changed our website theme? Is the revenue increase enough to cover our development costs?
If we decide to go with a different packaging or shipping supplier, what’s the expected impact to our total landed cost for this product collection?
As you can see, just about any question you dream up can be answered by data.
Post purchase journey
We never want to forget the last step of the fulfillment process - the post-purchase journey. Analysis of data in this area can help you ensure continued customer satisfaction, good standing as a vendor on marketplaces, and catch any slippage that incurs indirect costs.
Shipping data is a good starting point - you’ll want to monitor shipping times and costs, and the trends in these over time. If shipping times grow longer, you risk unhappy customers and de-prioritization in the Amazon buy-box, for example.
You can mine customer review data to gather quality control insights.
Look at trends in return data to detect quality issues or which marketplaces attract less satisfied customers.
All these insights can help you identify and resolve upstream issues, which may otherwise have gone undetected.
But how do I get started?
Okay, I’ve given some examples of what you can use data for, and a high level description of how to run the analysis. But I’ll admit this can still be overwhelming for someone who is unaccustomed to running their business this way.
One approach is to start small, doing spreadsheet-based data analysis to build up an organizational comfort with the process. Then, as your analytics process matures, you can learn new skills to improve the analysis - for analyzing data, you’ll likely want to learn coding in Python.
That said, many business owners don’t have the time or resources to learn these skills. Even if they did, they want to speed up the timeline to results - taking months or years to learn analytics coding is unacceptable to them.
If you’re one of these people, who are serious about using data to drive supply chain decision making and want to complete your first analysis in as little as one week, then you’re going to need support from a data professional. We at Bro Analytics provide just that kind of support to eCommerce companies like yours. If you’re interested in learning more, reach out on our Contact Page.