The dairy management wished to quantify improvements in production planning and sales forecasting and see the benefits in terms of a reduction in finished product wastage and improved service levels to its customers.

The latest short shelf life model is a simplification of its dairy orientated predecessor, as much of the complexity and functionality has been removed to offer a more generic tool with a number of potential applications.

The model requires the usual inputs of weekly sales forecast, daily profile, sales forecast error, lead time, minimum batch quantity and supply interval leading to the calculation of production quantities, daily demand quantities and inventory levels.

However the main difference between the short shelf life model and a standard inventory model is the tracking of date codes and remaining shelf life within the total manufacturing and total distribution shelf life constraints, and also that service level is a model output.

This means that although total stock quantity will cover the forecast period, the various date codes and distribution shelf life of the SKUs will mean that certain elements of stock cannot be used to meet the demand, as the maximum shelf life has been exceeded.

It is particularly useful to understand the sensitivity of the inputs described above, as this dynamic leads to both stock write offs or lost sales or both.

The model works stochastically using multiple replications as prescribed by the end user as the demand is varied around the forecast by means of the forecast error variable.

The results are smoothed keeping track of all key stats such as service level from the multiple iterations.

These handling units are very expensive but necessary assets, and control of these is vital given the various routes to market that the company may be involved in.

Model Logic have developed models for the brewing and dairy industries enabling the calculation of the total trolley or keg pool required given the production, inventory and distribution requirements through the various routes to market.

Allowances are factored in for scrappage, mis-use and loss, giving the total equipment pool required to keep the production line going whilst also reflecting the demand profile, including seasonal variations.

This allows the management to be able to plan better, and purchase the optimum number of trolleys or kegs to maintain the correct pool level. The same modelling techniques can be used in any high street retail sectors such as clothing, where equipment is used to facilitate handling.

The Model Logic systems are tailored to the specific transport problems. If your business is facing any of the following issues, then a Vehicle Routing System would be a very useful tool to quantify cost implications of alternative strategies.

1 Investigate the trade- off between economies of scale on production and /or warehousing, and the difference in transport costs to the customer base. In other words when production or warehousing capacity is being re-aligned, then primary and secondary distribution will be affected, and the cost differences tracked.

2 In a multi depot problem – allocate customers to the closest depot or within capacity constraints.

3 Identify resource levels (eg men and vehicles) to deliver to or collect from a set of customers.

4 Quantify savings/on costs of removing or adding depots following re allocation.

5 Quantify savings/on costs of operational changes. (eg Delivery Time Window changes)

6 Bench marking of the operation for meaningful comparisons between depots.

7 Gain a better understanding of the distribution problem being solved with key statistics such as inter call distances, driving time, duty time, load factor achieved, journey times, distances and costs by journey being reported.

8 Identify potential savings of clustering by days of the week and better territory management.

9 Quantify savings/on costs of increasing or reducing delivery frequency.

10 VRP systems can also be used as a work study tool for analysing the driver working day in more detail.

11 Merging two or more distribution networks and identifying coincident delivery points. The system will quickly identify the level of savings which may occur from a known base case.

It is always useful to understand very early on whether the problem is time constrained OR capacity constrained, and an initial constraint free run will give an indication of this.

For example, in a time constrained problem, increasing vehicle capacity will be of no benefit to the operation, as vehicle fill will be even more inefficient, and the costs will be greater.

The Vehicle Routing Problem (VRP) is also sometimes known as the Travelling Salesman Problem (TSP). The only difference is that TSP usually requires a 1:1 customer to sales person relationship, so that the same sales person visits the same customers all the time. Other than this, the data input structure and core algorithms are very similar.

The purpose of an inventory model is to understand the interplay between lead time, supply interval, minimum order quantity, desired service level and forecast error. This leads in the first instance to the setting of safety stock levels and re order point stock level in order to minimise the risk of a stock out or lost sales. At the same time we wish to avoid overstocking of product as working capital is unnecessarily tied up as well as valuable warehouse space.

We have used the model in a variety of sectors from construction to food, where our clients wished to identify and quantify the inventory savings of stocking SKUs at one central distribution centre versus several regional distribution centres. This has enabled them to classify the stock and identify which products would better be held centrally as opposed to regionally.

It is worth exploring the variables that make up an inventory model in more detail, as most organisations will have data that can be manipulated quite quickly in order to start generating some meaningful results.

Model Inputs

1 Lead Time
The period of time between the placement of an order and the visible availability of product for despatch in the warehouse. This is an important variable to consider when sourcing products from the Far East, with a possible sixteen week lead time.

2 Supply Interval
The period of time between the receipts of orders arriving in the warehouse.

3 Minimum Order Quantity
The smallest agreed order quantity that can be placed on the supplier.

4 Desired Service Level
The actual cases supplied divided by cases ordered expressed as a percentage. For example if the Desired Service Level is set at 95%, then for every 100 units ordered, 95 cases supplied would be acceptable.

5 Forecast Error
This is perhaps the most difficult input variable to ascertain.

Model Logic has developed a sub module attached to the inventory model which calculates the Forecast Error also known as Mean Absolute Deviation (MAD)% from a sales history dataset.

We would suggest that, where available, at least a two year sample should be used.

Alternatively, the forecast error is often calculated within a typical sales forecasting system and extreme data or outliers should be filtered out of the analysis.

Model Outputs

1 Safety Stock
Also known as buffer stock. This provides the contingency in case the sales forecast underestimates the actual sales.

2 Re-Order Level
This is the stock level at which an order should be placed. Delaying the order any longer will risk depleting the safety stock level.

The models using the above framework work well in industries where shelf life is not an issue. However in the food industry, for example, we often face the added problem of stock write offs – where SKUs have exceeded their sell-by date. So even with efficient stock rotation, parts of the stock have to be thrown away as the actual demand is less than the sales forecast.

This leads to a trade-off between stock write-offs and lost sales, so Model Logic has also developed a Short Shelf Life Model for these applications.

The technique employed is usually known as simulation, and is a very useful in gaining understanding of the dynamics of an operation and tracking the behaviour when various demands are placed upon it.

The common four variables in all of these situations are:

1.Arrival rates of objects and pattern
2.Process or transaction times at each workstation.
3.Resource level (number of operatives or machines at each workstation)
4.Critical path with decision points along which the objects must travel.

Once the above framework is established the queue abandonment rules consider alternative behaviours when confronted by the queue length and rate of depletion.

The interplay of all the above variables and rules determine queuing levels, work in progress, resource utilisation and service levels at different times of day.

Here are four examples of how we have used queuing theory or simulation techniques to solve real life commercial problems.

1.A defence industry contractor wished to tender for a programme of armoured vehicle manufacturing. The requirement was to establish a timetable for delivery given a complex critical path and limited resource capacity. A system was developed by Model Logic using Microsoft development tools highlighting pressure points where bottlenecks were occurring allowing extra resources to be brought in/contracted out to meet the various delivery deadlines.

2.Alternative car park layouts were evaluated on behalf of a major retailer in order to find the best layout with regard to throughput and access points. Each store was considered on its own merit as there was no one cap fits all solution.

3.When a leading high street retailer was faced with the prospect of replacing all the cash tills as they were no longer being supported by the manufacturer, they needed to understand how many tills were actually needed by each department, would till banks be appropriate with a single queue, and what would be the impact on the customer experience in the form of queuing levels. A system was developed by Model Logic using Microsoft development tools considering alternative queuing strategies between a single queue multiple till bank and multiple tills each with their own queue. The system reported queuing levels and instances at a minute by minute time grain throughout the store opening hours, given various footfall, till bank formation and transaction time parameters. Using this data, the retailer determined that a single queue with multiple tills offered the most efficient system.

4.A leading retailer wished to benchmark its call centre performance and standardise the target service levels across the group. The central service level KPI to be achieved would be measured as x % of calls to be answered within y seconds. The system developed by Model Logic using the standard Microsoft tools reported the operator numbers by hour of the day required to achieve the KPI standard, given the number of incoming calls, their length, and abandonment rate. The model uses simulation techniques, but also tracks in parallel a formulaic approach developed by AK Erlang in the early 20th century. The two approaches can then be compared to double check the findings.

In each of the cases highlighted above, the throughput or objects being processed are randomly generated by the model at the specified rate. Alternatively, they could just as easily be generated from a user input schedule – depending how much control the end user wishes to place on the model.

The system enables appraisal of existing or proposed vehicle routes, reporting on arrival times, departure times, inter – call drive times, kilometres and costs as well as time adherence variances.

Through Model Logic, we were able to fully understand the capability of our distribution delivery service. The highly detailed system allowed our distribution department to quantify the route detail such as distance travelled, estimated time of delivery, total drive and duty time. This paved the way for improvements in our distribution service, which most importantly is passed on to our customers” says Marcus Foley, Warehouse Engineer at Uniphar

Model Logic has previously developed tailored vehicle routing models for the food and drink and construction industries.

The system enables General Mills UK, which is part of the world’s sixth largest food company, to make accurate identification of costs from the international sourcing of products right through to final delivery to the retailer by apportioning shipping, transport and warehouse activity costs at case level.

This is a very comprehensive, detailed system that has enabled us to gain a better understanding of our cost to deliver to customers and drive improvements!’ says Shaun Campbell, Distribution Planner at General Mills

The system is a bespoke development tailored specifically to General Mills UK’s operation, and is run on a monthly basis to enable the monitoring of operational costs. The system also has the capability to be run ‘off line’ for the comparison of more strategic scenarios.

General Mills UK & Ireland is one of the fastest growing food companies in the market and was recently ranked for the 4^th year running in the Top 50 Best Workplaces in the UK and is also proud to be consistently named in the Sunday Times Top 100 Companies to Work For.

Glanbia, one of Ireland’s largest food companies, has been awarded the Chartered Institute of Logistics and Transport’s Environmental Award for its carbon reducing initiatives. Model Logic worked closely with Glanbia and Paragon Software Systems to reduce Glanbia’s environmental impact, while improving the efficiency of its delivery operation.

Model Logic created datasets for a wide variety of delivery options for Glanbia to input into the Paragon system, then prepared detailed costs for the alternative scenarios.  This clarified the savings to be made from different delivery windows and frequencies, shift patterns, and use of central distribution centres.

Glanbia is one of Ireland’s largest food companies and is the supplier of Yoplait, Avonmore and Kilmeaden dairy produce. The company’s Consumer Foods division is the largest branded food supplier into the Irish grocery sector with more food brands in The Top 100 than any other supplier. Glanbia makes over 2,500 deliveries a week to a wide range of customers, including most supermarkets and convenience stores operating throughout Ireland’s 26 counties.