ROI is a great measure for trying to gauge whether an investment is worth what you will be putting into it. A large, positive ROI implies that you’ll quickly and easily make your money back. Getting more out than you put in is the name of the game.
However, some investments are very difficult to measure return on investment. Putting in an enterprise grade finance system is more about avoiding costs associated with getting it wrong than on saving money tomorrow. Similarly, moving from a spreadsheet environment to an automated environment also is difficult to measure unless you can say for certain that there would be people cuts associated with the purchase.
New technology is often hard to calculate an ROI for because many of the benefits are expected to be productivity increases where people can spend less time on reports, checking numbers, inputting data, sharing spreadsheets, and explaining the process. Decreasing time spent on these non-value added functions give your team more time to focus on the part of the business that contributes increased value. The benefits are a step removed from the investment.
You’ll often see software companies touting the fact that their software pays for itself in 4 months or 12 months or 18 months. When you press them they will almost invariably fall back on the “productivity savings” theory and talk about how much less time people have to spend on data and process. It’s a false equivalency though because just as the benefits are a step removed, there often new support processes needed. New technology means better data and reports but that also means more time needed for delivering the reports and creating and socializing the new metrics. In the same way, getting a team to use new tools can take quite a bit of time investment as well.
Measurements require fixed starting and ending points to be accurate. Just as you can’t measure the length of the wind, you often can’t measure the ROI on a technology investment. Having a set of corporate guiding principles for when and why to invest can be a better method.