A Management Accounting Framework: A Taxonomy
10. Shortened Time Interval Between ABC/M Recalculations
In the early ABC/M period, many organizations recalculated their models only on an annual basis. As organizations lowered their administrative costs to refresh and update their systems, and as end-users requested more frequent reporting of more recent results, the frequency of calculating and reporting ABC/M data dur- ing a year substantially increased.
In the early 1990s relatively few organizations recalculated their ABC/M data more frequently than once a quarter, if even that often. Some simply re- calculated their systems annually, similar to the standard cost setting habits of manufacturers.
Now ABC/M systems are often being recalculated monthly, in sync with the financial period-end accounting close. Not all the driver quantities are necessar- ily updated; those metrics from the prior period’s recalculation are simply reused if it is felt there was not a substantial shift in the driver distribution. (Activity- based costing driver assignments normalize to 100 percent, so it takes a substan- tial shift to adversely affect cost accuracy.)
Applying ABC/M: Not Just Calculating It
As widespread acceptance of ABC/M data emerges, it is apparent that there is a shift in the focus for implementing ABC/M—from excessive time wasted on de- signing and collecting data for the system to time spent anticipating the uses of the new data.
As organizations apply ABC/M data in more types of decisions, there will be broader recognition of the interdependencies that exist, which will promote more integration of data systems and processes. Cost data are rarely used in isolation.
After an organization’s strategy and mission are defined, the organization’s core business processes take over as the mechanism that delivers value. Time, service level, quality, flexibility, and costs are all derivatives of a process and should rarely be viewed in isolation from one another. The need to better integrate cost data with the suite of metrics and performance improvement initiatives is growing.
Advanced and mature users of ABC/M will continue to grow in numbers and will demonstrate how ABC/M data are linked to the diverse yet interconnected aspects of managing for continuously better results.
ABC/M’S ACHILLES HEEL: THE LEVELING PROBLEM
In ABC/M, poor model design leads to poor results. When ABC/M systems fall short of their expectations, often the system was overdesigned in size and detail, well beyond diminishing returns in accuracy for extra increments of effort to col- lect and apply the data. The illusion that more detailed and granular data provide higher accuracy is part of the explanation for this behavior. But the other expla- nation for oversized ABC/M models is that at the outset of an ABC/M project it is nearly impossible to determine what levels of detail to go to.
There are so many interdependencies in an ABC/M model that, as a result, it presents a problem. It is almost impossible to perform one of the earlier worksteps of the traditional information technology (IT) function’s systems development project plan, “data requirements definition.” Why is this workstep so difficult to apply when designing an ABC/M system? Why do implementation teams fall into the trap of building excessively large models?
As discussed earlier in this chapter, ABC/M models often are initially over- engineered in size and detail. It would be easy to recommend following the keep- it-simple-stupid (KISS) rule for building ABC/M systems, but that rule is not applicable. ABC/M model designs are not simple, but they are logical. The key is to understand the properties of the Cost Assignment Network.
One of my ambitions is to take ABC/M from being a loose art form to a craft.
It is true that all ABC/M models are stylized, and two separate ABC/M teams will design somewhat different models for the same organization based on their varying assumptions. But in the end ABC/M model design is a craft, and we should understand it as a craft. That is the purpose of this section.
Death by Details
One of the problems plaguing ABC/M is leveling. No one can ever know be- forehand how large or detailed the ABC/M system should be. Unfortunately, in the absence of any guidance, the ABC/M project team’s initial design of the
ABC/M’s Achilles Heel: The Leveling Problem 75 ABC/M system tends to err on the side of too much detail, and this invariably spells trouble. Revisit Figure 2.16, which depicts a three-dimensional view of the ABC/M Cross. The added dimension, the depth of the three ABC/M modules, is where the leveling problem lies. Overdesigned ABC/M systems, which are too deep relative to their intended use, are noted at the bottom of the pyramid.
One solution to finding a reasonable level of detail for the ABC/M system is to work backward with the end in mind, as opposed to assuming in advance how detailed and accurate the ABC/M data must be. However, there is a double- meaning in the phrase “working backward with the end in mind”; it does not only mean considering the business problem to be solved, such as attempting to mea- sure and understand customer profit levels. It also literally means knowing the key characteristics about the final cost objects—those items targeted by ABC/M to be costed by assigning to them the actual costs consumed from their work activities.
Using an analogy, by knowing in advance the thickness of a glass sheet, one can sense how hard one must strike it to break it. The glass determines the effort.
This analogy applies to sizing the number of activities with ABC/M, taking the breadth of diversity of the cost objects for the thickness of the glass in the anal- ogy. We need to explore the meaning of diversityand why it is so important to ABC/M model design and architecture. We will discover that diversity, variety, and variation of cost objects govern the size and depth of ABC/M systems.
In conventional IT systems development, one of the initial project work- steps is the data requirements definition phase. Although that phase may be ap- propriate for developing a typical information system, such as an invoice and billing system, in ABC/M it is very difficult, perhaps nearly impossible, to pre- define the system because there are too many interacting variables, including the work activities, their drivers, and all of their outputs to be costed. In contrast to conventional systems development, defining and developing ABC/M systems is more effectively accomplished using Rapid Prototyping, a technique in which the information system is quickly mapped and then constantly and iteratively ad- justed to meet the users’ requirements. Similar to a military ballistic cannon crew, who repeatedly shoot and re-aim to get closer to their target rather than try to aim precisely once, ABC/M Rapid Prototyping achieves quicker results than trying to get it perfect from the outset. (Chapter 9 discusses ABC/M Rapid Prototyping.)
ABC/M Rapid Prototyping works because ABC/M models are scaleable.
The implication is that one can quickly initially build a model with activities and product or customer families at a more aggregated level. The input data can orig- inate from estimates provided by a few knowledgeable employees rather than from transaction-based databases. The results will yield a good first cut.
Given this first-cut ABC/M model, it is substantially easier to iteratively modify and adjust the level of detail of the successive ABC/M models based on relevance to the uses of the data and the accuracy requirements of those uses.
Each prior model can be examined to determine where it is more or less sensitive
to error. In short, ABC/M Rapid Prototyping is a highly managed “trial and error”
technique. This quick alternative for implementing ABC/M works largely be- cause of the ABC/M model design principles discussed in the next section.
Solving the ABC/M Leveling Problem
The ABC/M leveling problem can be solved not just through trial and error but also through better thinking. As the designers construct their ABC/M information system, they usually suffer from a terrible case of lack of depth-perception. There is no perspective from which they can judge how high or low or summarized or detailed they are. Because the implementation of ABC/M systems is usually in- fluenced by accountants, human nature is one of a “lowest denominator mental- ity.” Accountants usually assume a detailed and comprehensive level of data collection based on the premise that if you collect a great amount of detail every- where, and from everybody, and about everything they do, you can then always
“roll it up” and summarize anywhere. This is a “just in case” approach in antici- pation of any future remote questions. As a result, ABC/M models tend to be- come excessively large. Ultimately they may become unmaintainable and not sustainable. Eventually the ABC/M system does not appear to be worth the effort.
This outcome is unfortunate because ABC/M systems do not need to be ex- cessively detailed to be useful for decision making. What is not well understood is that ABC/M systems are scaleable in detail without much distortion. (Scientists call this expansion property “re-normalizable.”) This is very important because it means that with ABC/M you will see the same things that you can from a 50,000- foot view, only with better resolution, from a 20,000-foot view. The elements and components of an ABC/M system can be continuously subdivided and decom- posed, yet all costs will remain reasonably constant in their proportions relative to each other for a given time period.
I am not criticizing or attacking accountants. Years of training reinforce their high need for precision.
Diminishing Returns in Improved Accuracy
When people who are first exposed to ABC/M hear the phrase, “It is better to be approximately correct than precisely inaccurate,” they smile because they know exactly what that means in their organization. But they usually do not know what causes ABC/M to produce substantially better accuracy relative to their existing legacy cost system despite its abundant use of estimates and approximations. To design effective ABC/M systems, it is crucial to understand why, how, and where an ABC/M model can produce greater or lesser cost accuracy. Knowing ABC/M model properties is important because a reasonable level of accuracy of product and other types of costs can be economically achieved simply with good ABC/M model design and less reliance on having perfect input data. This seems impos- sible, and certainly counterintuitive, but it’s true.
ABC/M’s Achilles Heel: The Leveling Problem 77 Just as chiropractors do, ABC/M designers benefit from knowing where the error-sensitive pressure points are. That is, it is helpful to know roughly where an ABC/M model is more or less sensitive to error. The error and accuracy of costs are not evenly or consistently affected throughout an ABC/M Cost Assignment Network.
With ABC/M, imprecise inputs do not automatically result in inaccurate out- puts. That is, precision is not synonymous with accuracy. In ABC/M’s cost as- signment view, estimating error does not compound, it dampens out. After all, what is the real consequence of error when reassigning a source cost to its cost objects? The result is that some destination cost objects may be over-costed while the remainder must be under-costed. Cost allocations are a zero-sum-error game, and in the end 100 percent of the source costs are always completely assigned for each and every source assignment, no more and no less. To an enterprise-wide ABC/M system, this means that ABC/M is a closed system. The total resource costsmustequal the total activity costs, which in turn mustequal the total final cost object costs. (As previously stated, final cost objects are defined as the end- destination or output of an activity cost.)
Figure 2.22 shows a simple 3 ×4×3 cost assignment (i.e., three types of re- source costs that produce four activities and are used by three products). The dis- persion of estimating error for determining any single activity cost may be unacceptable in isolation. However, as multiples of different activity costs pile up
The “Dispersion of Error” contained in upstream assignment offsets as each downstream path reaggregates into each
final cost object.
Assignment error has a “ zero-sum” property:
Over- costed path $s (+)
Under- costed path $s (–)
=
+ -
+ -
Resource
Activities
Final Cost Objects
Assignment View
+ + – Contribution View Many-to-
One One-to-
Many
The Two Path Views
With ABC, it is counterintuitive that error does not compound. It dampens out.
FIGURE 2.22 ABC Error Has “Offsetting” Properties
into a single product’s cost, any estimating error begins to offset itself. As a re- sult, the dispersion of error contracts for each individual product. Consequently, the cost error level (relative to perfection) is reduced to potentially well within the comfort level of any decision maker who may use the data.
How does the estimating and data collection error dampen out? With ABC/M, the recorded general ledger expenses (which start out being error free) are first segmented and traced to their activity costs. These activity costs are then further traced proportionately to reflect the diversity of the consumption effect that each of their cost objects (e.g., products) is placing in the form of demands on the activities. That is, allthe activity costs are reassigned back into the cost ob- jects consuming the activity costs after the activities have been segmented. Each activity’s assignment to all of its cost objects may incur some modest error. But from the cost contribution point of view for an individual destination cost object (e.g., an output, product, standard service line, or customer), the costs are accu- rate. All the cost assignments are aimed at the individual cost object as if the cost object is a bull’s-eye target; some of the assignment errors will be plus and some will be minus. But collectively the accuracy of the cost objects’ assigned costs will be fairly reasonable.
As the activity costs recombine back into each cost object, any earlier and upstream error tends to offset and partially cancels out. In effect, the “law of off- setting errors” has kicked in. Any upstream estimating error that produces the under- or over-costing has a canceling effect; hence error dampens out. These are properties of statistics found in equilibrium networks (i.e., the amount of costs re- mains constant). And ABC/M is a cost reassignment network, as illustrated in Figure 2.9, much more than it is an accounting system.
This property of a Cost Assignment Network is very relevant to ABC/M. By understanding this property, the excessive administrative effort to collect and re- port worker timesheet data can be tremendously reduced and the data collection will be much less invasive for employees. Generally employees do not enjoy completing daily or weekly timesheets. In ABC/M, if the “vital few” employees estimate on behalf of all the employees, the cost objects may not be materially different than had everyone completed the timesheets. If a portion of that saving is applied to better tracing of the activity-to-cost objects, the entire ABC/M model becomes more accurate with much less overall effort.
Effective Right-Sizing of the ABC/M System
Figure 2.23 shows several curves that all have the same destination: 100 percent perfectly accurate cost results, where the accuracy level is represented by the ver- tical axis. The horizontal axis represents “the level of effort.” As explained previ- ously, for each incremental level of effort to collect more and better data, there is proportionately less improvement in accuracy. So the phrase, “Is the climb worth the view?” is truly applicable to ABC/M. Figure 2.23 also draws attention to effi- cient and inefficient performance levels exhibited by ABC/M project teams to find
ABC/M’s Achilles Heel: The Leveling Problem 79
the right combination of accuracy and ABC/M administrative effort. There will al- ways be a balanced trade-off of more data for higher accuracy. But an appropriate question being raised here is, “Whichdata, and whatis the effort to collect that particular information?”
Unfortunately, most ABC/M project teams perform too far to the right on Figure 2.23 and usually on a much lower “frontier curve.” That is, their intersec- tion of both axes in the figure means they have put in a much greater effort than was needed and they received less accuracy in costing than they could have achieved if they had been more clever. The challenge for today’s ABC/M teams is to determine how to right-size their ABC/M models, and do so economically.
Few organizations can afford excesses. Excess ABC/M model structure—such as number of activities and drivers—saps the strength of ABC/M in the initial stages.
Determinants of Accuracy
What is missing in most ABC/M implementations is a good understanding of what factors actually determine the accuracy of the ABC/M-calculated outputs.
Whenever there is uncertainty, it is human nature to collect more than what may be needed. But in ABC/M the producers of information must take into account the accuracy requirements of the users of the data. Accountants and engineers tend to be driven by a desire for precision. However, there are trade-offs among accuracy, relevance, and effort.
When the goal of the initial ABC system is profitability reporting, more poorly designed and misleveled ABC systems will yield less accuracy despite greater effort!
Accuracy of Final Cost
Objects 100%
0%
World Class ABC System Design
Little
Level of Data Collection Effort
Great Modest
ABC Project Team A is achieving higher accuracy with much less effort than ABC Project Team B.
A B
FIGURE 2.23 Balancing Levels of Accuracy with Effort
Six determinants help properly level and improve the accuracy of an ABC/M model’s results, and their impact is roughly in the following order ranked by de- creasing impact on achieving higher cost accuracy:
1. The breadth of diversity, variety, and variation of the final cost objects.
2. The level of disaggregation of the activities (this is where cost materiality is considered).
3. The assignment relationship between the source cost and its destination cost objects (i.e., which objects are consuming).
4. The correlation of the activity driver to its activity (how linearly variable it is).
5. The accuracy of both the resource and the activity driver quantities.
6. The accuracy of the ledger expense data (which are usually error free).
Ironically most ABC/M project teams start constructing their ABC/M model at the wrong end of the Cost Assignment Network. They like to begin with informa- tion that they already know and are comfortable with: the expenses that have al- ready been accumulated in their accounting ledgers. By starting with these expenses as a source (i.e., in the resource module of ABC/M), the ABC/M project teams initially define and compute the costs of activities. Although work activities are the main players of ABC/M and are central to what an organization does, a highly desirable objective is to construct an ABC/M system of the right size. (I wear a size 40 suit, but I do not buy a size 42 or 44 suit so that I can get more for my money.) There is a better and more pressure-sensitive place to begin determin- ing the ABC/M model design.
Although it is not obvious at first glance, ABC/M models are extremely sen- sitive to the diversity and variation of the eventual destination of costs: the final cost objects. An ABC/M project team should begin understanding its “leveling”
requirements at the final cost object, not at the resources located at the opposite end of the ABC/M network. Accountants may be more comfortable with the ledger costs that they are already familiar with, but traditional cost ledgers sim- ply give the accountant a perfect view of where the organization has already recorded expense data. They are looking at the wrong end of where and how costs flow. The idea is to understand the costs of outputs as seen through the costs of work activities. Costs originate with final cost objects that place demands on work activities of people and equipment.
Following is a discussion of the six determinants of ABC/M accuracy, ranked by decreasing impact on achieving higher cost accuracy.
Breadth of Diversity, Variety, and Variation of Final Cost Objects The final cost objects are the least understood part of an ABC/M system, but they are probably the most important of the three ABC/M modules. The final cost object module that houses the final cost objects consists of (1) the final cost objects them- selves and (2) the interrelationships among the final cost objects. (Final cost objects are defined as the end-point destination of the consumption of activity costs.)