Intelligent Enterprise

My recent posts on simulation analysis and BPMN naturally lead into simulation use case three, which deals with Activity Based Costing (ABC). A number of users have asked me if simulation provided activity based costing, and I always said yes, since I assumed it could. But it's not built into the tools at all. This turned out to be a really interesting part of the training to develop. All completely original, since the modeling tool vendors don't really talk about it (or at least correctly), and the ABC literature doesn't mention simulation, either.

Activity Based Costing is not just determining the cost of each process activity from its active labor costs, or even its total direct costs. Yes you get that directly from the simulation model, as in use case 2. ABC is about allocating the indirect costs, both labor -- management/supervisory, plus ancillary functions -- and fixed overhead, like plant and equipment. You can't get that from the simulation model!

Or can you?

Turns out you really can, by making some simplifying assumptions. Robert Kaplan of Harvard Business School, one of the original inventors of ABC back in the 1980s (and also known for Balanced Scorecard), came up with a streamlined form of ABC a couple years ago called Time-Driven ABC. The original ABC used management interviews to allocate the indirect costs., but too hard, it turns out, to scale that to the enterprise and maintain over time. So time-driven ABC makes it simpler.

Time-driven ABC says just allocate the indirect costs by the relative fraction of active labor time for each activity, and you can get that directly from the simulation model. In the training we used simulation and time-driven ABC to provide the "true" cost of various types of work, i.e. types of process instances running through the same top-level process… even though each type of work — which could represent, for instance, a different product or service provided by the organization — typically follows a different path through the common end-to-end process. ( A common process is needed in Process Modeler to allow for resource contention across types of work.)

Time-driven ABC calculates what it calls the activity cost driver rate as the unit cost x the unit time for each activity (or type of work). The idea is that the organization as a whole invests its assets to create capacity to produce. That capacity is measured in man-hrs (e.g. per year or per quarter). The organization's total costs (direct + indirect) to produce a set of products and services, measured in dollars per year (or quarter), divided by the total effective capacity per year (or quarter), gives the unit cost, in dollars per man-hr of effective capacity. The "effective" part is an 80% fudge factor that says that's the highest utilization rate you can really get. So for a FTE working 2000 hrs a year, the effective capacity is 1600 man-hrs/year. The organization's effective capacity is N x 1600 man-hr/year, if N FTEs are deployed to do the work.

The unit cost for the organization as a whole is modeled as the organization's total cost (direct + indirect) divided by effective capacity, measured in dollars per man-hr.

All of the activity-specific (or work type-specific) breakdown is in the unit time, which represents the man-hrs of capacity required to produce one unit of work (one instance, in our model). That comes from the active time to perform each activity (or process instance for a particular type of work). This method allocates the indirect costs, as well as the direct costs, to specific activities or classes of work. But when you multiply the unit cost x the unit time x the number of instances of each type produced, you get less than the total cost. In Kaplan's terms, what that gives you is the cost of the used capacity; what's left over is unused capacity. Meaning, you could get more work out of the available resources. Let's not debate the economic meaning of all this. The point is that these numbers can be easily calculated from the simulation model, assuming you know (on an annual or quarterly basis) your total indirect costs to be allocated.

From a simulation standpoint, it turned out that the key thing to have was the simulation instance data, down to the activity instance level. Using Excel array formulas, I could then separate the different types of work. After that, the time-driven ABC is a simple formula. Really cool! For some of it I could have used more than Excel's 65,000 rows per sheet, but there is a way to direct the simulation output to Access or SQL Server if you have lots of data. Some day I think I'll write this up in a more understandable way. Maybe work with ITP-Commerce to make it out-of-the-box!