Features Research & Innovation
Solving the Supply Chain Cost Riddle

As this magazine’s readership is aware, distribution can be more complex than many manufacturing processes and is often a major piece of a company’s cost structure. In spite of such significance and complexity, the accounting profession offers little guidance in terms of a model and approach to explain the transportation component of a supply chain. At Catalyst Paper, we recognized such a need and spent several months formulating and developing an appropriate model for our $250 million annual distribution spend. The eventual model quantified a host of “variances” (differences between actual and budgeted costs) to provide a somewhat automated explanation of distribution costs at the company-wide level as well as at the level of individual customer accounts. The Canadian Academic Accounting Association peer reviewed those results and published them in their academic journal, Accounting Perspectives, in May of this year. What follows is a less-academic, less-mathematical synopsis of those results.

At Catalyst Paper, as at many other organizations, the number of customer ship-to locations and combinations of routes, modes and carriers available to service those locations is many times greater than the organization’s number of products and variants of those products. As well, Catalyst’s costs of distributing product to customers typically range between 5% and 30% of the eventual selling price, which is not unusual for manufacturers. Catalyst has four source manufacturing plants, several means of transport to and from many warehouses and multiple carrier options. Taking into account the fact that the company typically has 500 or more customers to service in a single month, there are thousands of possible combinations of plant, warehouse, mode, carrier, and currency. Catalyst programmed the logic for an automated cost variance analysis tool that quantifies fully the differences between actual and budgeted costs.

The tool incorporates the key drivers of costs, such as the destination, source plants, warehouses, modes of transport, carriers, freight rates, surcharges, exchange rates, yields and other factors and quantifies how each of these items contributes to cost overruns or under-runs against a budget or against a prior period. The tool identifies which factors in the supply chain are failing to be at expected levels.

Accountants will recognize at this point the applicability of the variance analysis tools that they may have learned as students. At Catalyst, we started with that model, but found in short order that what looked good in theory in textbook analyses of manufacturing costs left much to be desired when applied to real cost data for complex distribution systems. Textbook examples didn’t have the complexity of interdependent variables that supply chains typically exhibit.

While we will not discuss data compilation and data cleansing in any detail, we would be remiss to not mention that the issues around data integrity and completeness can be daunting. Seemingly insignificant flaws in the source data can destroy the integrity of the entire analysis if not treated properly. Do not underestimate the importance of getting the data complete, consistent and correct, as it needs to fit neatly into a somewhat unforgiving mathematical model.

Without delving into details of the myriad of complexities inherent in logistics, systems and data, the following provides an outline of the concepts and output that Catalyst settled on. Other particular situation may not be the same as that of Catalyst, but the concepts are generally portable.

Examining the data

At the highest level, distribution costs in total will be more than or less than budgeted. Take a look at the example in Exhibit 1. You can readily see that the company spent $36,650 less than budget. Scan through the numbers and try to identify the reasons for those savings. In this simple example, one can discern that transportation costs were lower than plan as a result of reduced sales volume; only 3,360 units were sold versus a plan of 3,685. Now try to identify the reasons that costs were $2.09 per unit less than the budget of $91.17. The reasons for that difference are not readily apparent but the major factors can be identified with a few minutes of eyeballing the data. Now imagine having to do this for dozens or hundreds or thousands of customers. The task at hand is to break down the total variance into a set of numbers that identify all of the reasons that costs aren’t as planned – whether at a corporate level or by customer.

Scanning through the numbers in the example, you’ll notice cost impacts, including, for example:

* 3,360 units were sold in the period versus a plan to sell 3,685;

* the proportion of sales to each customer was not as planned;

* XYZ Trucking hadn’t been considered in the budget but took 30 truckloads to ABC Inc;

* DEF Rail’s rates were higher than planned;

* DEF Rail was budgeted to carry 85 units per railcar but only took 72 (360 units in five railcars).

There are many other reasons for costs to vary from plan in this example. The trick is to quantify each of them appropriately in a solution that adds to $36,650. We use the following categories of variances to arrive at the solution: volume, customer mix, distribution mix and carrier charges.

Volume variance

Simply put, the volume variance calculates the overall impact of selling more or less than budgeted. Even if all cost factors are on plan and the customer mix is precisely as planned, a change in volume will change total distribution costs. The calculation of the impact of volume changes is simple and straightforward: the overall difference between actual and budgeted volume times the total budgeted cost per tonne. Exhibit 3 shows the calculation of a $29,629 favourable volume variance for our example. Distribution costs are lower than planned, partly because the company sold less than planned.

Once the volume impact has been explained, the remaining variance calculations deal with explaining why the per-unit costs are higher or lower than planned. In our example, the planned distribution costs were $91.17 per unit while the actual costs were $89.08 per unit. The $2.09/unit explanation will involve customer mix, distribution mix and carrier charges.

The customer mix and distribution mix variances identify changes in the proportions of customers, routes, modes and such that impact overall costs. Exhibit 2 illustrates these choices and how they can create complexity in supply chain cost analysis.

Customer mix variance

The customer mix, or destination mix, variance determines the impact of shipping to different customers. For those customers who don’t ship directly to their customers’ homes or places of business, the equivalent could be shipping to their retail outlets or distribution centres, but the general concept is the same. Rarely does a company manage to ship to all of the customers they planned to and in exactly those planned proportions. A calculation of the mix impact is done for each customer, some will be favourable (i.e. more product shipped to low-cost locations) and some will be unfavourable, and the results are summed to arrive at the total cost variance attributable to customer mix. The customer mix variance multiplies the planned cost per unit for each destination by the difference between the actual volume shipped and the volume that would have been shipped if this customer received the planned proportion of total sales.

In Exhibit 3, the destination mix calculation shows that the company sold more product to ABC than planned and less to K&V; and, because K&V was a more expensive ship-to location than ABC, the net result was favourable, a savings of $561.

While the first two variances are fairly simple to calculate and represent high-level impacts, the remaining variances are mo
re complex and involve calculations at the level of groupings of actual deliveries.

Distribution mix variances

Distribution mix variances deal with the logistical complexities involved in routing product from source to customer. The selection from many options of individual manufacturing plants, warehouses, modes, and carriers may seem simple enough to deal with in day-to-day operations, but these combinations complicate the mathematical analysis of costs. Supply Chains typically involve choices that depend on other choices, which is why the mathematics for supply chain cost variance analysis can be much more complex than manufacturing cost variance analysis. For example, the choice of one manufacturing plant over another can result in a different set of warehouse options; the warehouse chosen to route the product through may have different outbound mode options than another warehouse, and so on. We won’t go into the specific explanations and proofs of the mathematics in this brief article, but suffice to say that the combinations of routes and modes and carriers complicate the variance analysis math at each turn.

Catalyst breaks their distribution mix variances into the following:

* First-leg route (or source plant)

* Last-leg route (or warehouse)

* Mode

* Carrier

The first-leg route variance and the last-leg route variance quantify the impact of the end-to-end routing decision. If product can be sourced from plants at different geographic locations and shipped to one or more warehouses, the choices made can result in significantly different costs. Capacity constraints, production problems, strikes, weather, and a host of other factors can result in sourcing and routing product through suboptimal channels. Isolating the impact of routing decisions can go far to optimizing a supply chain or identifying bottlenecks and shortcomings. Mathematically, the calculations of the route variances are relatively complex, requiring a specific data record structure and formulae to avoid double-counting and overlap with other variance amounts. For the mathematically-inclined, the appendix to this article provides a look at the formulae. Accountants will recognize the general idea of mix variances from their management accountant training, but will find that the routing mix variance calculations here go beyond textbook examples because of the interdependency of the first leg mix and last leg mix.

In Exhibit 3, the example points to effective routing saving the company $0.70 per unit; a loss of $0.33 on the first leg and a saving of $1.03 on the final leg to the customer by routing a higher proportion of product through Warehouse 2, which has lower planned last-leg costs to ABC Inc than Warehouse 1.

The mode mix variance recognizes the impact on costs of selecting various modes of transport (for instance, rail, truck, intermodal, container or break-bulk) between which there can be large differences in cost. Customer preference, late production, late orders or adverse weather conditions are examples of what can create deviations from a planned mode mix and can be the primary determinant of average freight cost. The mode mix variance reflects whether the actual mode mix (at benchmark costs) was more/less expensive than the benchmark mode mix. The mode mix variance calculates the impact of shipping via a different mix of modes than planned and can point out the impact of bottlenecks or equipment shortages.

In the example, we find that the company managed to save $1.55 per unit in distribution costs due to an effective mode mix. The company shipped a higher proportion of product by rail than planned. Rail had an $80 planned cost per unit to ABC that was less than the average $90 per unit for trucking.

The carrier mix can be influenced by carriers’ relative rates, equipment shortages, strikes, core carrier programs, contractual agreements on minimum or maximum volumes and a host of other factors. There can be a very significant variation in rates from one carrier to the next and some carriers may even have differing vehicle capacities that can influence average yield. The carrier mix variance reflects whether the actual carrier mix for a given mode was more/less expensive than the benchmark or budgeted carrier mix.

In our example, a judicious selection of carriers resulted in a savings on overall freight costs of $0.74 per unit.

Carrier Charge Variances

Once the route, mode and carrier have been selected and costed, the analyst needs to determine the impact of the rates charged and the efficiencies achieved. Mathematically and intuitively, these price and efficiency variances are easier to determine and to grasp than the distribution mix variances, but they nonetheless require much attention to detail when dealing with real-life data from today’s complex computer systems. Catalyst breaks their carrier charge variances into two types: rates and efficiency.

Rate variances include, for instance, the difference between the actual and planned freight rate, fuel surcharges, accessorials, border charges, brokerage fees, port charges, etc. Essentially, each component of the actual cost (in dollars per load or dollars per unit as the case may be) is compared to the budget and the difference in cost is applied to the actual volume shipped to arrive at the dollar impact. In Catalyst’s case, transportation is transacted in multiple currencies, so the exchange impact has to be isolated from the other factors.

In the example, rates charged by DEF Rail were slightly higher than planned, adding $0.10 per unit to average freight costs.

Efficiency, in Catalyst’s case, is limited to the difference between the actual and planned yields (volume per truckload, railcar or container). The difference detemines, for instance, the number of extra truckloads or carloads that were required due to low yields (which may have been the result of poorly planned stows or suboptimal order sizes). The change in the number of trips is applied to the planned cost per trip to determine the dollar impact of the efficiency loss or gain.

In our example, we planned on shipping 85 units per railcar but only managed 72, adding $3,250 or $0.97 per unit to distribution costs.

Reporting the Results

The rubber hits the road in any cost analysis when the results are finally presented. The variances need to be summarized in a format that can be understood and, of course, acted upon. You may want to consider utilizing one or more of the following methods of presentation:

* Graphically, as shown in Exhibit 4 for the overall company or for individual customers.

* Top-ten lists for each type of variance, showing the ship-to destinations with the highest impact. For example:

1. carriers or warehouses providing the worst/best yields;

2. warehouses arranging the worst/best carrier mix;

3. carriers whose rates have increased significantly beyond budget or more than their competitors;

4. modes or carriers most impacted by increasing fuel surcharges;

These report examples can provide immediate results in terms of identifying business process and/or logistical issues that are resulting in increasing costs or are keeping your company from achieving planned efficiencies or cost levels.

As the significance of distribution grows with each passing year, it becomes increasingly important to have immediate and complete analysis tools for those costs. While we readily acknowledge that we have not covered all of the attributes of transportation costs and their efficiencies and that our own experience is not directly applicable to most other organizations’ circumstances, we hope that we have illustrated a general approach that you can apply to your specific circumstances.

Kevin Gaffney, kevingaffney@shaw.ca and Valeri Gladkikh, valeri.gladkikh@telus.net