How to manage uncertainty in the supply chain

In order for consumers to get their next-day delivery…

• Products have to be manufactured or ordered from suppliers months in advance
• They have to be shipped and stored amongst thousands of other products at a local storage hub to ensure availability
• A team of warehouse operatives have to pick, sort, and pack each product, ready to be shipped
• Each product has to be allocated to a courier, loaded onto a truck, and delivered on a route with hundreds of other parcels

And this is just a simple example! With rising sales, growing product ranges, and increasing customer expectations, supply chains are getting more and more complex. It falls on supply and demand planning teams to manage these complicated processes, and there are catastrophic consequences when they get it wrong… 

A reliable view of what is likely to happen in the future is key to planning, and as a result, businesses tend to look for more-sophisticated forecasting techniques to help. But this is not always the best approach.

Forecasts are never perfect. There’s always going to be some uncertainty associated with them, but downstream in a supply chain, there are a few extra effects that exacerbate this:

• The bullwhip effect amplifies small upstream changes and causes much more uncertainty closer to the customer.
• Unconsidered upstream constraints, such as limits on production capacity or inventory availability, add to the bullwhip effect and also increase uncertainty.
• The data required at each step in the supply chain is different. Trying to predict how many of a specific t-shirt a retailer should buy for the season at a national level is relatively easy, as you’re trying to estimate the behavior of a large population. Trying to predict how many of the same t-shirt will need to be packed in the warehouse in the next hour, though, is much more difficult – as you’re trying to estimate the behavior of a few individual customers.

Uncertainty in supply chains is unavoidable. This is particularly true downstream, where uncertainty can be the driving factor in decision making.

The best way to deal with uncertainty is by adding buffers – extra people or extra stock to make sure that any unexpected demand can be accounted for. The size of these buffers is important. If the buffer is too lean, customer service levels will be affected, and any excess buffer is a direct cost.

Despite this, tasks like calculating safety stock targets and resource planning are often left to gut-feel and combing through spreadsheets, when data-driven approaches can make a big difference.

The relationship between customer service levels and extra resources is a complicated one; holding an extra pallet of stock might dramatically increase customer service level if it’s currently only 20%, but won’t make much difference at all if your customer service level is 99.9%. For this reason, simulations are powerful tools for planning buffers, as they allow you to explore this trade-off in detail and set the buffer at the correct level for your particular business. They are also robust, repeatable, and automatable!

Here are five extra benefits of automating these kinds of buffer decisions which also reduce uncertainty:

1. Plans can be generated more regularly

Planning can be a boring, time-consuming, and laborious task. It often takes days to collate all of the data required, analyze it, and decide what the plan should be. This means that processes that are planned by people often don’t get reviewed regularly, which can cause problems for the business. The example below shows sales of a product (black line), and a series of target stock levels based on a forecast plus a simple safety stock level.

For the red line, the stock target is reviewed yearly. For the green line, the stock target is reviewed quarterly, and for the blue line, the stock target is reviewed weekly. The yearly review, because it has to account for a longer time period, contains a much larger buffer of safety stock. By reviewing more regularly, in this case inventory levels could be reduced by 37%.

2. Plans will be more consistent

In a team of planners, each planner may all have a slightly different way of doing the job. Some team members might be better at planning for certain products, or at certain times of the year. By automating the process, you can take the best qualities from everyone in the team and make sure that all planning is done in that same way. This improves the planning in general, but also generates more predictability for downstream processes.

3. Plans can include extra data sources

Automating a process allows you to bring extra sources of data together more easily. This can make a massive difference to the quality of planning. Joining inventory levels and sales forecasts, for example, ensures that you have enough stock to meet your sales targets. Joining raw material levels with production plans means you plan to produce things that you have the raw material for, and and joining warehouse throughput to freight plans means that you can time the pick-ups to maximize trailer fill and customer service levels.

When these processes are all kept on spreadsheets, it can be difficult to join these plans together – but with an automated process it can be done easily.

4. Plans can be made more visible across the supply chain

The bullwhip effect, mentioned earlier, is a problem because each process in the supply chain has a knock-on effect for everything downstream of it. By automating planning processes, it’s easier to distribute the results and share plans more regularly across the supply chain. The improved consistency and visibility increases the time downstream processes will have to react to any sudden changes, improving the robustness and resilience of the supply chain as a whole.

5. Plans can be examined by looking at the supply chain holistically to test what’s working

Automating a planning process requires businesses to think about what KPIs are the most important to them. These KPIs can be tracked. If this is done in a systematic way, it means process changes can be quantified and assessed to see whether they are having the desired impact and working as intended. For example, you could systematically increase the number of warehouse operatives and measure the impact on customer service level.

Final thoughts

With more complicated supply chains, more personalized products, and more demanding customers, uncertainty in planning is likely to increase. As a result, handling this uncertainty is likely to become more difficult, more important, and more expensive. The way to solve this problem, in my opinion, is through automation to improve the consistency, visibility, and regularity of planning, and simulations to improve and quantify the trade-off between buffer stock and customer service levels.