Having an underperforming facility can be a disaster. You spent a lot of time and money on it, and the business will want to see the planned return on investment. Here are some of the most common causes of underperformance:
1. Traffic jams / congestion
People, machines, or goods can get in each other’s way. This elongates journey time, which can decrease throughput. If your picker is waiting for someone to get out of their way, they are not busy picking. If your storage shuttle is waiting for another shuttle to get out of the way, your picker is waiting longer for the goods to arrive before they can do their pick.
It can be very hard to correctly anticipate the impact of traffic, particularly given that traffic patterns are likely to differ throughout the day.
2. Bursty workloads
Your site may be able to handle a decent throughput, but if your workload is bursty, your overall site throughput will be lower than the site’s peak throughput.
For example, this may be due to when inbound or outbound vehicles arrive. If you have no outbound vehicles to load onto, your outbound operation comes to a halt. If suddenly all of your outbound vehicles arrive, you can still only load them at your peak rate.
3. Automation not suited to your use case
Some automation handles certain usage better than others. Things like the number of different products you handle, whether you sell many more of one product vs another (how extreme the pareto distribution is), product sizes, customer order profiles etc. can all affect how well a given automation approach will work for you.
For example, online supermarkets sell an awful lot of cucumbers compared to how much they sell a certain shade of lipstick. Cucumbers are in a very large proportion of their customers' orders. If their system had to store their cucumbers far away after each pick, it would hamper the throughput of their whole system. Pickers would frequently be waiting for cucumbers to arrive.
4. Unanticipated bottlenecks
Often, your operation contains bottlenecks that you hadn’t expected. It can be hard to identify these from the symptoms you observe such as low throughput.
It is often fairly simple things like a conveyor diverter. Many conveyor diverters have a high “straight ahead” throughput, and a lower “divert to the side” throughput. This was probably known at the design stage of your facility. You may have written it off as fine, because you assumed some distribution of “straight ahead” vs “divert to the side” operations.
However, it may be that at certain times, a larger than anticipated proportion of containers need to be diverted to the side, at which point your diverter throughput starts to constrain your system. We have another article going deeper into bottleneck analysis.
5. Insufficient training, experience and incentivisation
Even if you’re not hit by all of the above, new facilities built almost identically to previous ones underperform initially, due to staff being unfamiliar with how to get the most out of it. This is both in terms of hands-on staff not being familiar with things like the picking process and how to pick fast, and also the management team not being expert in how best to manage the site, e.g. where to direct labour when.
This is even the case with quite a lot of training in place. Training makes people a bit more effective, but not as effective as doing the job day after day and building muscle memory - practice makes perfect. Even if they can operate quickly, they aren’t necessarily motivated to do so. Incentivisation can help address this.
It can be hard to identify what your issues are. Even once you have, it can be tough to know the best way to address them.
Two common approaches to identify and alleviate performance issues are:
Guesswork based on intuition followed by trial and error in the real world, or
Building a simulated model or digital twin of your operation and using it to answer questions and experiment with improvements
The latter is cheaper, faster, less risky, and results in a more optimal operation. Here’s how simulation can be applied in order to identify issues and how best to solve them:
First, you create a simulation model or digital twin of your existing operation. It’s important to have confidence you’re making the right decision, and therefore you need to ensure that the simulation is accurate. You do this by comparing its predictions with what actually happened in the real facility.
Then, you use the simulation to explore what is constraining your operation. Here’s how that applies to each of the issues we identified above.
1. Traffic jams / congestion
You run a simulation with the model altered such that everything is able to pass through each other. This is obviously not feasible in reality, but it will very quickly and easily tell you the biggest effect you could possibly have through optimising your traffic. This technique is known as constraint relaxation.
If it makes a significant difference to your key metrics, you then run various simulations with the congestion constraint re-applied, exploring different ways of managing congestion more effectively. This would include things like different floor layouts, changing the order of visits, changing where different products are stored and handled, adding more paths and bypasses etc. Each different simulation will quantify the effect of the change that has been simulated.
2. Bursty workloads
You run a simulation where the work is spread more evenly than you are seeing in production, and see whether that affects your key metrics. If it does, explore ways that you could either smooth the workload (e.g. customer incentivisation by pricing delivery slots differently) or tolerate the bursty nature (e.g. better buffering). The latter is easy to explore in simulation. You would model various buffering approaches, and re-run the simulation to quantify what impact they have. This can be used to hone in on the best buffering solution for your use case.
3. Automation not suited to your use case
It can be highly expensive and disruptive to replace machinery. You may not get it right the first time, and have to replace it again (and again). Instead, you run simulations where you replace your machinery with different options. While it will still be expensive to make the change in the real world, you only have to do it once, and only if the return on investment predicted by the simulation justifies it.
4. Unanticipated bottlenecks
You run a number of simulations, each of which replaces a part of your system with a faster alternative. The replacement which had the biggest effect is the one to target first.
You then simulate different ways of addressing the bottleneck (using different machinery, using more machines, relocating it, adding extra traffic flow to/from it etc.) to find the most optimal approach.
5. Insufficient training, experience and incentivisation
You run simulations at your currently observed work rates (things like pick times), and your theorised work rates (e.g. from time and motion studies). If there’s a significant difference, the question is whether your theorised pick rates are realistic or not.
If you find they are reasonable, you need to look at how to get your staff up to that level. This could involve training, practice, and incentivisation. However, if you find that actually your theorised work rates aren’t realistic, you need to explore other ways of increasing throughput in this area. You then simulate things like adding workstations (because each workstation is capable of a lower throughput than you had thought).
Data driven decision making can reduce the uncertainty which comes with needing to improve performance in production and distribution centres.
Our engineers have created simulations used to design and optimise the most advanced fulfilment centres for many of the world's leading online grocers. With over a decade of experience working with operations, warehousing and logistics, we have a proven track record of helping organisations like yours achieve significant improvements in operational efficiency and cost savings.
If you would like to talk to us about how we can help optimise your operation through simulation and digital twins, get in touch.
Comments