Cost Accounting for Software Product Development

Cost accounting seeks to attribute value from parts of the product sold against activities involved in creating it. The product must be divided into piece parts and an attempt made to value the contribution of those piece parts to total revenue. The value generated from the piece parts is then assigned against activities that may have contributed to the creation of the piece parts, and some derivation is made as to whether or not those activities added value or not.

Imagine that the new version is ready. It took 1 quarter to produce and will sell for 1 quarter before it is replaced with the next version. The sales forecast must try to predict how many of the sales are attributable to the features in the new version. It is necessary to predict how many sales would have been lost to the competition if the features in the new version were not made available. That is to say, new features have both a positive additional effect and a prevention of potential negative effect.

The sales mix of new sales against upgrades must be considered. Upgrades can be apportioned directly to the features in the new version, whilst new sales must be cost assigned across the new version and each of the previous versions. When an upgrade is sold, it is safe to assume that the purchaser is buying the new features in the current version. However, when a sale is made to a new customer, this customer may be buying the product for all sorts of reasons. Some of the older features may be attractive, but so might some of the newer ones. Trying to determine an attribution for this gets ugly very quickly. Calculating all of it is possible, but error prone and problematic.

Is there an alternative? A system that takes a continually cumulative view of the product? In this scenario, sales and operating expense for the entire life of the product to date would be accumulated. This would certainly tell us whether the product has added value or not, but it would not tell us whether the business is currently adding value or not. It is effectively using the end-of-life equations above for the total return from the product in an on-going fashion. Such metrics would tell us whether the product produced a return on investment or not, but it would not indicate whether the latest version produced a return on investment. The problem is that the contribution of individual versions is lost in a calculation aggregated over the entire life of the product.