How do you specify costs, time windows, and periods?


Properly specifying the scope and structure of cost estimation can greatly help in subsequent steps. Such a specification requires being precise about the costs to model, the time window (from which date to which date?) and the period (hour, day, week, month, quarter or year?) over which costs are to be modeled. These parameters have a huge influence on the cost estimation technique you can use, the reliability of estimates, and their meaning.


Which costs?

Which costs have to be modeled is fundamentally determined by the manager’s needs. A business unit manager may be interested in all operating costs and an operation manager may want a better understanding of quality costs only. The question the management accountant must address to give them the information they need is not so much the scope of the cost that managers decide, but the level of aggregation of these costs: should the accountant model all operating (quality) costs at once? Or should she split these total costs in sub-groups and model them separately, repeating the estimation process for each subgroup?

The answer to this question is the same as the one we provided for the definition of cost pools in Chapter 2, for the same reasons. Ideally, cost estimation is better applied on homogeneous costs, i.e. costs which have the same underlying cost driver(s). Some cost estimation techniques can work with multiple cost drivers, but this makes them more demanding in terms of available data and makes estimates less reliable.




Which time window?

The question of the time window and period are closely related. In general, we want the largest time windows possible because the more periods you have, the more cost estimation techniques you can used and the more reliable your estimates. Indeed, some cost estimation techniques do not require historical data (engineering approach), some require at least one period (account analysis), some require at least two periods (high-low method), and others require ideally several periods per cost driver included in the model (regression). More periods allow you to use more techniques which can then either corroborate each other or on the contrary call for caution about estimates.

Unfortunately, there are limits to the time window. When a company starts off when it changes its technology (e.g. old machines are replaced by new ones, more efficient, or part of the production process is automated), we must ignore all observations based on the old technology. This drastically reduces the time window and thus the number of periods management accounts can use, reducing the reliability of estimates.

Note that this is true whichever period is selected. It is tempting to believe that using shorter periods (e.g. months instead of quarters) can improve the estimates by mechanically increasing the number of periods (e.g. 12 months instead of 4 quarters). However, you should not compare estimates derived from different periods because they do not have the same qualities and the same meaning.


Which period?

Using different periods results in different estimates which are difficult to reconcile. When you rely on monthly observations, estimation returns monthly fixed costs; when you rely on yearly observations, it returns yearly fixed costs. Unfortunately, the latter is rarely 12 times the former because some costs which are fixed on the short-term become variable on the long-term. For instance, a lease with a three-month notice period will appear fixed between two months but variable between two quarters.

Dis-aggregating periods is also likely to reveal different ranges of activity which should be modeled separately, as we will see a bit later when I introduce the concept of relevant range. Moreover, shorter periods are characterized by a greater likelihood to exhibit more cut-off errors: when changes in costs are recorded on a period different from the period of the changes in cost driver which caused them (for instance because an invoice is recorded one month after the resource consumption and not thus properly matched). These errors tend to cancel each other out on longer periods, but they come back when periods are dis-aggregated.

What will determine the relevant period is ultimately the managers’ time horizon and whether they wants to manage costs on the short term or on the long term.




At that point, you probably start to understand why financial accountants are so wary of estimations. All this calls for a lot of skepticism about any isolated approach. You can only have some confidence in your estimates when different techniques applied to the same costs over the same period and time window yield similar results and thus corroborate each other. However, you should not compare estimates obtained on different costs, periods, or time windows.


Please indicate how clear and understandable this page was for you: