Activity-Based Costing

Goal

Activity-based costing establishes the cost of each activity in a process and of each object that consumes these activities. It neither removes nor changes any cost. It shows how the process consumes its resources.

A direct cost attaches to a product without calculation: the material in a part, the billed hours of a client project. An indirect cost serves several products: sales administration, the complaints department, the premises. Traditional cost accounting spreads it with an allocation key based on a volume, for example revenue or production hours. The CBOK states that activity-based costing traces each cost to a cost object instead of allocating it.

Robin Cooper and Robert S. Kaplan formulated the method in the late 1980s. Their 1988 article in the Harvard Business Review finds that some companies base their prices, their product range and their production technology on distorted costs.

The deliverable is a table of the activities with their cost, driver, frequency and unit cost. A second table gives the cost of each object. The method also costs the do-nothing scenario and identifies the processes that add value. Total cost of ownership, cost-benefit analysis and return on investment start from a process cost. Activity-based costing supplies that cost for the current state and the target state.

Usage

When to use it

• High indirect costs: allocation by volume distorts the cost of each service; trace the costs by activity.
• Process redesign: cost the current state, the target state and the do-nothing scenario with the same unit costs.
• Strong price competition: know the cost of each service and of each customer segment before setting a price.
• Diverse customers or product range: small orders, demanding customers and special products consume more activities than their volume suggests.
• Measuring the cost of quality: cost inspections, rework and complaints with their drivers.

When not to use it

• Almost all costs are direct: nothing to trace; add up the direct costs of each service.
• Lead times and waiting times to reduce: the method costs resources, not the customer's waiting time; map the flow with value stream mapping.

Description

The elements

Resources are what the organisation pays for: salaries with OASI and occupational pension contributions, rent, licences, depreciation. The general ledger lists them by type of expense. An activity is repeated work that can be counted: entering an order, handling a complaint, chasing an invoice. A cost object is the thing whose cost the analysis calculates: a product, a service, a customer segment, a sales channel or the whole process.

A cost driver measures one element's consumption of another. The CBOK gives this name to both links in its chain: between resources and activities, then between activities and cost objects. Roztocki et al. speak of first-stage and second-stage cost drivers. A resource driver assigns an expense to the activities, for example the share of the team's time or the square metres occupied. An activity driver assigns the cost of an activity to the cost objects, for example the number of orders, complaints or set-ups.

Tracing instead of allocating

An allocation by revenue assumes that each franc of sales consumes the same share of administration. Yet an order of CHF 200 takes the same entry work as an order of CHF 4'000. Allocation by revenue therefore loads the large orders with costs that the small ones cause. Roztocki et al. draw the rule from this: a cost that does not depend on volume distorts product costs when it is allocated in proportion to volume. Their examples of volume-independent drivers follow the work: set-up hours and the number of set-ups, ordering hours and the number of orders.

Running the analysis

Roztocki et al. describe an eight-step procedure built on Cooper's two assignments. They calculate each assignment in a matrix of proportions and check that the result equals total expenses. The analysis sequence is a synthesis: it takes their matrices and their check, then the before-and-after comparison that the CBOK assigns to the method.

1. Set the scope and the cost objects. Write down the question: the cost of a customer segment, a service or a process. The scope sets the departments included and the period, for example the financial year.
2. Collect the expenses. Take the indirect expense categories from the income statement: salaries, rent, IT, insurance. Their total is the control figure at the end.
3. Identify the activities. Roztocki et al. start from the process flow chart: each box of the process model is a candidate activity. Group homogeneous tasks. Separate the activities that a product triggers from those that a customer triggers. Each new activity needs a quantity to record every period.
4. Assign the expenses to the activities. Draw up the expense-activity matrix with one column per expense and one row per activity. First tick the cells where the activity consumes the expense. Then replace each tick with a proportion taken from the resource driver. The proportions in a column total 100%. The cost of an activity is the sum of its row.
5. Choose the driver of each activity. The driver counts the events that drive the workload of the activity: orders, complaints, reminders. It comes from an existing system (ERP, ticketing tool, accounts receivable) so that the model updates without a survey. Revenue and quantities sold reproduce the distortion that the method corrects.
6. Assign the activities to the cost objects. Draw up the activity-object matrix on the same principle. The proportion of an object is its share of the driver quantity. The calculation amounts to dividing the cost of the activity by the driver quantity, then multiplying this unit cost by the quantity that each object consumes. Order entry that costs CHF 312'000 for 12'000 orders comes to CHF 26 per order.
7. Add up per object and check. The cost of an object is the sum of its row. The sum of the objects equals the total expenses collected in step 2. A difference signals an expense that was missed or counted twice.
8. Compare and decide. Read the cost of each object against its revenue. For a redesign, cost the target state with the same unit costs and the planned frequencies.

Roztocki et al. obtain the proportions at three levels of accuracy. The first, which they call educated guess, brings management, finance and the people who perform the activities together to estimate the proportions. The second, which they call systematic appraisal, combines individual judgements, for example with the analytic hierarchy process. The third measures the work, for example by work sampling. At random moments the analyst asks a person what they are working on. Measurement is the most accurate and the most costly of the three levels. A model can combine the levels: estimates for the small activities, measurement for those that carry weight in the decision.

Comparing before and after a redesign

The unit cost of an activity multiplied by its frequency gives its consumed cost. A redesign acts on one of the two factors. It removes an activity, reduces its frequency or shortens its duration.

The calculation shows the capacity released. The released hours stay paid until the organisation reassigns them or reduces headcount. Rent and IT costs do not fall either. The comparison therefore separates the consumed cost from the unused capacity.

The portal takes half of the orders and checks the entries. Complaints fall by half. The consumed cost drops by CHF 240'000, but spending rises by CHF 40'000 until the organisation decides what happens to the released capacity. The redesign business case names this decision and its date.

The time-driven variant

The first generation of the method collected the share of time by survey. Each employee estimated how their hours split across the activities. Kaplan describes these surveys as time-consuming, costly and error-prone because they rested on subjective estimates. They irritated the employees who filled in the forms. Kaplan and Anderson note that many executives abandoned the method as its costs rose and the staff grew irritated.

Kaplan also acknowledges that the proponents of the method misunderstood the role of capacity in calculating unit costs. In his view, the unit cost is calculated on the capacity supplied: the time that the organisation pays for. A unit cost calculated on the time used alone charges the unused capacity to the products. The wholesaler's rate of CHF 26 per order assumes a fully occupied team. The before-and-after comparison of a redesign therefore keeps this rate. Recalculated on the remaining 6'000 orders, it would double and hide the released capacity.

Time-driven activity-based costing (TDABC) replaces the survey with two parameters. The first is the cost of one unit of capacity, for example one minute of the team's work. The second is the time that one unit of activity requires, for example entering one order. The managers estimate these times. The cost of an entered order becomes the product of the two parameters. The capacity that the activities do not consume appears as unused capacity.

Pitfalls

Taking the assigned cost for an avoidable cost. The cost that a customer segment carries in the model is not what dropping the segment would save. The premises and the licences remain. Before a decision to drop a segment, separate the costs that would disappear from those that would remain.

Building a model that costs too much to feed. Roztocki et al. name the lack of data as the main obstacle for small companies. Kaplan observes that the early models were expensive to update and that many were never updated again. A model of fifteen activities kept current at every close gives better information than a model of two hundred activities frozen for two years. An educated guess is enough for a first model.

Letting staff believe they are under surveillance. A team that believes its individual productivity is being measured under-reports the tasks it thinks are frowned upon. The analyst announces that the model aggregates times by team and shows the result to the people who supplied the data.

AI considerations

A language model proposes a draft mapping of accounts and expense labels to activities. From process descriptions and ticket categories it proposes an activity dictionary that the team corrects in a workshop. A spreadsheet or a script recalculates the model at every close.

These proposals miss the work that no system records: a phone call with a customer, help for a colleague, the correction of an error that came from upstream. The analyst checks the proposals against the activities and the time shares that the managers report. The choice of a driver affects prices and team budgets. Management accounting settles it with the activity managers. Salaries listed by name are personal data under the Federal Act on Data Protection (FADP). The analyst aggregates them by team before passing them to an external AI service.

Examples

This illustrative example shows what the activity driver changes compared with an allocation by revenue. The sales administration of a wholesaler costs CHF 600'000 a year. The cost of each activity comes from the expense-activity matrix.

Allocation by revenue makes the key accounts carry CHF 390'000 that the small customers consume. An order from a small customer brings in CHF 200 of revenue on average and costs CHF 26 to enter. The small customers also receive more reminders per order than the key accounts: 22 per 100 orders against 10.

Visualisations

The CBOK draws the method as a chain: resources, activities, cost objects. A cost driver labels each link. The diagram of the two assignments follows this chain with the amounts of the example. The width of each flow follows the amount it carries. The two assignments are calculated in matrices of proportions. The table of costs by segment is the deliverable that management reads. The before-and-after table separates the consumed cost, the unused capacity and the total spend.

Cost

Tools

A spreadsheet is enough for one department or one process: one sheet per assignment, one row per activity. ERP systems include a management accounting module. SAP CO offers activity-based costing linked to cost centres. Process mining tools such as Celonis derive the frequencies and durations of each activity from the event logs of the ERP. They supply the driver quantities and the times of the time-driven variant without a survey. Time-recording software supplies the share of time per activity when the team enters activity codes in it.

Sources

• ABPMP, Guide to the Business Process Management Common Body of Knowledge (BPM CBOK), version 3.0, §6.11.3 Activity-Based Costing: the definition, the chain that links resources, activities and cost objects through drivers, the tracing of indirect costs, the comparison before and after an improvement, the do-nothing scenario and the conditions of use.
• Robin Cooper and Robert S. Kaplan, Measure Costs Right: Make the Right Decisions, Harvard Business Review, 66(5), 1988: the origin of the method and the pricing, product-range and technology decisions taken on distorted costs.
• Robert S. Kaplan and Steven R. Anderson, Time-Driven Activity-Based Costing, Harvard Business Review, November 2004: the abandonment of the method by executives and the time-driven variant built on the estimates of managers.
• Robert S. Kaplan in Harvard Business School Working Knowledge, Adding Time to Activity-Based Costing: the flaws of employee surveys, the role of capacity and the two parameters of the time-driven variant.
• Narcyz Roztocki, Jorge F. Valenzuela, José D. Porter, Robin M. Monk and Kim LaScola Needy, A Procedure for Smooth Implementation of Activity Based Costing in Small Companies, conference paper, 1999: Cooper's two assignments and their drivers, the distortion caused by volume-based allocations, the eight-step procedure with its matrices of proportions, the three levels of data accuracy and the lack of data in small companies.