Forecasting has long been considered a perfect science! This fallacious statement is, of course, hopelessly wrong. Companies have often prepared forecasts based on historical data only to miss their revenue goals by miles. Then along came spreadsheets which accelerated the process and allowed analysts to modulate the forecasts by making some selective adjustments based on ‘market’ trends. Even this was not perfect. Businesses continued to miss forecasts, and then wondered why.
For logistics companies, the challenge is even harder because they are usually dependent on the information provided by their clients. My experience has been that some do very well at this, others not so much. When there is a collaborative approach for instance between the 3PL and their customer, things get slightly better, building a level of trust over time. In an extended supply chain, with multiple players across the globe, the knock-on effort downstream compounds itself if you get it wrong. Some logistics companies in the past, because of historical misses by partners, would build in ‘buffers’ or ‘contingencies’ in their forecast out of fear. In an era of tighter and tighter margins, this is very inefficient and expensive.
The principle reason for this was that all such forecasts were looking backwards – a classic oxymoron. Luckily, we now can look forward using the tools we call predictive analytics.
UNDERSTANDING PREDICTIVE ANALYTICS
Predictive analytics is an application of machine learning and AI, encompassing a variety of statistical techniques from data mining to predictive modelling, that analyses both current and historical facts to make predictions about future outcomes. For end users, predictive analytics can give them insights and suggest actions that directly impact operations, revenue, and risk assessment. In fact, it applies to business applications for a wide range of use cases across various industries. Predictive analytics allows companies to play multiple “what if” scenarios well in advance of events happening. This then trickles down to the production and logistics divisions. Why is this so important? Logistics is not simply a question of packages. Take planning for Black Friday and Cyber Monday and the run-up to the holidays, for instance. Planning begins in May and June, and sometimes before. Hiring the right number of associates and on-boarding them requires good advanced forecasting. The need to hire, on-board, and train for when volumes hit, is primordial. Just as you don’t want to have excess or too little inventory, head-count planning avoids the extra costs of either over-hiring or paying overtime in case of a shortfall.
Understanding and forecasting demand accurately remains a key challenge for organisations. Demand is never linear and is affected by numerous variables, some of which are outside the organisation's control. Predictive analytics allows organisations to improve demand forecasting through analysing past and current trends, and together with market intelligence and economic forecasts, to more closely forecast demand. Shipping and transport costs often account for a significant percentage of the final product price. By using predictive analytics, it is possible to determine optimal shipping frequency and quantity to meet demand while minimising costs. Predictive route-planning can determine the fastest routes considering traffic congestion, distance, weather, delivery points, and time to market.
APPLYING PREDICTIVE ANALYTICS IN THE SUPPLY CHAIN AND LOGISTICS SPACE
Understanding what occurred in the past (other than for root cause analysis from a process point of view) is not nearly as important as knowing what should happen in the future from a volume perspective. Using this information, supply chain leaders can now address supply chain challenges, reduce and fine-tune costs, and at the same time, improve customer service levels. Predictive analytics allows organisations to identify patterns hidden in their data to understand market trends, identify demand, and establish production forecasts. They must not forget though that PA is not simply pressing a button to run an application. It is a collaborative effort that combines brain and machine learning.
Here are two areas of supply chain that are worthy of further examination:
- Firstly, let’s focus on the demand generation and order management. These are the forecasts for companies needing to move product into stores and/or to their B2B customers, and who need to be able to predict/forecast the sales volumes that they expect so as to place orders into production in a timely fashion. It is a closed loop process – B2B2B2C, with logistics providers right and centre of it.
- Secondly, it’s important to determine how the logistics providers can improve their planning processes so that these goods arrive on time and that the best use is made of available resources.
Logistics providers comprise shipping companies, freight forwarders, 3PLs, and delivery companies. Here’s how predictive analytics can be applied within their domains:
- Freight forwarders: For freight forwarders, it allows them to predict the need to switch delivery methods across the ocean between container shipping and air freight. Forwarders are the biggest consumers of space on container fleets. They must not only contend with load availability to book space on ships, but also incorporate “what if” scenarios, for such factors as weather, port slots, and political issues.
- 3PLs: Predictive analytics allows 3PLs to determine optimal inventory levels to satisfy demand while minimising stock. Using sophisticated models, predictive analytics allows supply chain managers to determine detailed inventory requirements by region, location, and usage. In this way, safety stock levels can be reduced, and inventory can be placed where required. This ability is particularly useful for the 3PL partners when their customers have multiple distribution points, as it helps supply chain managers determine whether stock should be held centrally or at regional facilities and pulse product through the supply chain to the stores.
Predictive technologies such as machine learning and cloud-based inventory management solutions eliminate overstocking and enable warehouses to work with each other – as opposed to each operating in individual silos – to meet demand. The ultimate results are high service-parts fill rates, high levels of product uptime with minimal risk, and increased customer service levels.
On a more strategic level, with the use of predictive forecasts from the customer, 3PLs are able to plan ahead for warehouse space needs and ensure resources in terms of automation and personnel can adequately deliver through the supply chain. Efficient demand forecasting, which predicts future demand for products and parts based on past events and prevailing trends, is a key component of after-sales service success. With an accurate picture of demand, manufacturers can improve service after the initial sale of a product without having to raise costs.
- Domestic carriers: As part of the planning process for retail sales, it is essential that 3PLs inform domestic carriers (TL, LTL, and express) of expected volumes for their own planning. They depend on transportation management systems (TMS) to track and manage shipments and lead times. With predictive analytics, many transportation management systems can now predict future disruptions before they happen and help logistics companies manage their operations proactively, rather than reactively. Predictive analytics can also create new visibility into seasonal buying patterns and forecasts to help suppliers make more informed decisions. Failure to do so means that product is not in the right place at the right time and retail stores, who depend on on-time arrival though the holiday period, miss sales targets.
IMPLEMENTING PREDICTIVE DEMAND PLANNING
This is where predictive demand planning plays such a big part. It begins with telling the buyer what the consumer will buy as opposed to what they did buy, since the latter only drives a re-order. One of the critical factors in customer service is the notion of available to promise (ATP), which is information about when replenishment orders will arrive. Predictive analytics adds a level of accuracy to when product will arrive for a client. There’s also returns management, after-sales service, and the availability of spare parts. Service parts inventory optimisation helps companies across multiple industries explore opportunities to increase the financial value from their service organisations. Predictive analytics is key to after-sales service organisations' success.
It all comes back to the impact that the failure to use predictive analytics can have on their end customers. Inadequate planning for sales volumes, etc., means late deliveries and missed targets. Clearly, many companies are still somewhat stuck in the past. Excel as a technology, even with complex macro models, continues to be widely used for forecasting due to historical perception of challenges with predictive analytics. Techniques have not changed, or companies are reluctant to change, and thus operate in the same way they have been for years. They remain in their comfort zone because they are fearful, and they are failing to progress to the next level and realise the full potential of predictive analytics. Most businesses are naturally risk-averse, but only by embracing change and perhaps even by creating a company culture that encourages and withstands certain failure will firms begin utilising and learning from all that predictive analytics can offer.
HOW TO GET STARTED USING PREDICTIVE ANALYTICS
Buy-in from peers and internal constituencies is critical to begin this journey. Producing results which convince these groups that predictive analytics works is the path forward. Firstly, find a promising case. Gather a core group of your top users across a variety of teams or departments and take care to pick the right combination of people so you get a diverse range of feedback. Brainstorm the top three supply chain pain points and pick the one that everyone agrees is the highest priority. Examples might be:
- Excessive customs fines due to poor paperwork compliance from freight-forwarding
- Poor on-time delivery of next-day customer shipments
- Failure to see early warning signs of missed forecasts for Black Friday
Do not forget that we are using history to correct the future. Predictive analytics can become an integral part of the Root Cause Analysis process in Six Sigma, for instance. The most valuable predictive analytics solutions are integrated in existing workflows, processes, and decision-making steps. Users can get future insights in context of the applications they already use – and they can act on those insights without jumping into another system.
As more users benefit from predictions, even more users will want to adopt the application. Partner with stakeholders at every step of the journey. Maintain constant communication with end users and continuously respond to their needs to keep your project on the right path. You need them to succeed, and they need you to achieve successful outcomes as well.