As more demanding requirements are placed on supply chains, new approaches must be found to reduce the impact on the operations and make them viable. Short lead times, fragmented demand, higher specialization and reduced lifespan make standard supply chain management a nightmare. That’s where predictive forecasting and dynamic MRP come in, to give you the opportunity to become more agile and better prepared for changing demand. Read on to check out what is this all about, why it matters to you and what you should do to get it to work properly.
It’s a reality and you better get used to it because it’s not getting easier. Most industrial and distribution activities are faced with significant challenges that are radically changing how business is managed. As consumer habits and marketing strategies evolve, the underlying infrastructure that makes it happen needs to transform in order to cope with the new requirements. Where you had 24-hour order lead times, you now have 6 and soon you’ll have 2. Where you had a limited SKU universe, for which you could plan with relatively certainty, now you have SKU universes that increase by 30% each quarter and soon an SKU lifespan could be as small as a month. As globalization thrives and different cultures mix, so specific selling opportunities are created, trying to target the specific consumer habits of market niches. Soon, we’ll only have market niches and no more standardized offerings.
This means that either you change the way that you plan your operations and the way you map your requirements or you’ll be faced with a problem of colossal proportions that either will drain your resources (financial and human), or put you in a non-competitive position in the market. You can no longer base your sales and operation planning in gut feelings, educated guesses or simple extrapolations, which are what the standard ERPs have to offer you. You need to engage in a serious and disruptive way of estimating what your demand is going to be in order to accommodate for the underlying diversity and variation. For this you need to engage in predictive forecasting, supported by specific technologies and using a proven methodology, in order to obtain reliable and sustainable results.
Now, this is not a miracle and it certainly is not an easy, immediate dream as very often you are told. Actually, this is useless if not integrated in a global strategy to achieve a level of accuracy and agility that produces sustainable results. So, if you just spend money on some sexy machine learning platform without assuring that you are in control of the rules that drive the models, if you are not able to change the model behavior, if you are not certain about the accuracy of the data used to feed the model and if you don’t have a dynamic planning engine to help you define requirements and optimize capacity, then you’re just throwing money out the window.
When looking into predictive forecasting strategies, you must first define precisely what is relevant for your business, since predictive forecasting can be applied in multiple areas and in multiple levels. For instance, you can apply predictive forecasting in transportation requirements, in warehouse required capacity, in sales planning and many other fields. The fundamental thing to look for is what are the areas you will engage in predictive forecasting and what is the level of detail you need to support your operation planning. For instance, it certainly looks very interesting to build a forecasting mechanism to predict SKU demand. But is this really relevant and what is the reliability of the data generated? And what about the amount of data required to make such models work properly? Maybe an SKU category demand prediction is as effective as an actual SKU prediction and it does give you some latitude in accommodating for variability in marketing campaigns. Then you need to select a platform that allows you to build the rule sets that drive the forecasting engine. You should be weary of anything that is a miracle, especially the miracle that is already a complete fit for your business. Even if you are a standard operation, your business and your organization are unique, so you need to be in control on how the model behaves. And you also need to have this degree of flexibility, because the rules behind the forecasting engine will have to change to accommodate for operational improvement and new business requirements. So, keep in mind that you need to be flexible and you need to be able to control the model.
Then you need to define the data collection, data inference and data learning process, so it is accurate and directed to your forecasting needs. Again, flexibility is key, since as rules change to accommodate for process modifications, so data collection definitions need to change. When all is in place, comes the hard part. You need to iterate the model to improve it. The chance of you getting the model definition right on your first try is, well, low.
That’s not a problem, that’s a part of the trail you need to cover. It’s by analyzing the results and matching them against reality that you are able to improve the rules that drive the engine, that you improve automated learning ability and that you can assess if your base data is accurate and properly collected. If you stick to this methodology, then you will be able to achieve a sound forecasting model that will enable you to predict in variable timespans with accuracies as high as 98%. This also means that you will be able to reduce costs by as much as 40%, cut your operating stocks by more than 50% and achieve performance indicators above 95%. Good, so I loaded up my forecasting strategy and I’m going for it. That means I’m all set, right? Nope, forecasting is just a part of the process. Besides making sure you are able to efficiently collect accurate data for the forecasting engine, which translates into mapping the complete supply chain process, you still need to use the results of the forecasting process consistently.
Traditionally, this means MRP. But the old school ways of performing MRP just don’t fit in the level of change that needs to be accounted for when looking at the global distribution landscape, nor to the responsiveness you need to achieve to really take advantage of powerful forecasting. Standard, ERP driven MRP, is a monolithic batch process that, using a snapshot of the demand, stock, master data and planning, cascades requirements and derives what needs to be produced or procured. All good, but this means that, between MRP runs, you just don’t know what is actually going on. And, because it’s a batch process, you can count on a nice timeframe to get it to run again. Sure, ERP vendors are putting on a brave face and saying they are using new database technologies, and the cloud and what not.
Sorry, that’s just not going to cut it. You need to have a dynamic approach onto MRP. This means that your MRP needs to adapt to changes as they happen, giving you an on the spot accurate vision of what is actually required. Bear with me on this one, this one is actually quite fun. Dynamic MRP means that every time anything that affects requirements happen, it is automatically accounted for. And I mean anything. Change in engineering specifications, as soon as it is approved an event is raised and the modification in components is cascaded down the entire production and procurement chain. A change in demand was identified in the forecasting engine, it is automatically processed to adjust production and procurement requirements, up or down. A new firm order is placed and scheduled, procurement and production requirements are regenerated on the spot. A production order is finished and their consumption variations and yields are automatically used to adjust requirements for production and procurement. A dispatch advice is received from a supplier and expected goods receipt quantity variations are cascaded across the landscape.
The immense power this gives you translates into having accurate information on the spot to make flawless planning. No more buying in the dark, knowing not if something is lurking in the background while the MRP runs. This strategy cuts safety stocks by more than 60% and leverages your bargaining power with your suppliers. No more producing more, just in case. This cuts raw material waste by more than 20% and increases production yields by more than 30%. And you can optimize your planning because you can account for variations on the spot, therefore adjusting to what is actually needed instead of planning for what was needed a week ago.
This disruptive way of engaging in MRP requires very specific strategies and technologies. The massive amount of data that needs to be processed and the real time nature of the required responsiveness makes this a gigantic challenge. You need to rely in distributed computing, rule-based engines and transactional load balancing to make this happen. And, as with predictive forecasting, you need to iterate the event processing setup, assuring it is based in configurable rules, until you can prove that requirements are being regenerated consistently and according to the entire landscape.
Bottom line, you need a different approach. Forecasting and dynamic behavior provide the foundations for that new approach, allowing you to face these challenges with a sustainable operation. Don’t be lured into thinking this is easy or a dream. It takes solid effort and consistent work to get it right. And you certainly need a process focus on your organization to allow the full potential of these strategies to be unleashed. All combined, this is the landscape that delivers a sustainable and efficient operation.
Over 10 years, Processware has been researching, implementing and fine tuning a state of the art software platform that delivers these and other advanced technologies, focused on providing a fully enabled, responsive, process awareness to any organization. This platform has been successfully deployed in world class customers, providing remarkable return for extreme landscapes. You can read more about these success stories and our platform here.
