Chris Matts recently wrote about the impact of discounting on investments in software development. He showed a comparison between a few different investment options for a software project in order to demonstrate the negligible effect of discounting. Whilst I agree with the conclusion – that discounting is mostly pointless for software — I wanted to add a few thoughts and observations.
- The appraisal period matters. When considering investment periods that are short (i.e. only 6 months) it should be obvious that discounting has almost no effect. Of course, if the investment period is longer, the effect of discounting can be much greater. It is spurious to compare various short appraisal periods claiming to show the effect of discounting.
- Please don’t forget the benefits. Not only is this the whole point, but when doing investment appraisals it is the benefits side that is typically more affected by discounting. Why? Well, the benefits typically last a lot longer than the initial investment. In some cases, the benefits accrue in perpetuity, so adding up all of the discounted future benefits into today’s money can be important. In particular, if the lion’s share of the benefits are projected to occur a long way into the future, ignoring the effects of discounting can lead to really poor investment decisions.
- Always compare options with a “do nothing” scenario. The “do-nothing” scenario is the base-case (or counterfactual) against which all other options are compared. Sometimes “do-nothing” involves no costs and no benefits – but for software investments, there are often existing manual alternatives or some economic risks to consider in the counterfactual.
- The rate reflects the riskiness of the investment. The relatively low interest rate on your mortgage is a reflection of the relative safety that comes from the bank being able to sell the property if you stop paying. Given the risks (and alternative investments that are less risky), we should only invest in more risky options if they are likely to generate higher returns. The most commonly used measure of riskiness for a company is the Weighted Average Cost of Capital (or WACC). None of this is nasty or tricky business though – just ask someone in your Finance department what discount rate to use for your organisation.
When discounting makes sense
So when should you use discounting? There are lots of situations where discounting future costs and benefits makes perfect sense. When the effects of an investment are long and the timing and profile of cost and benefits are very different there is really no other way to fairly compare options. For any large and irreversible capital investment with lots of options, you absolutely should compare options thoroughly. Discounting future cashflows as part of a cost-benefit analysis will help you make a better decision than your gut-feel may suggest to you.
When discounting is mostly nonsense
The sort of decisions described above should be very rare in product development. I’ve seen very few examples where I would recommend applying discounting, for the following reasons:
- The horizon of commitment should be fairly short (measured in months, not years). If it’s not, you’re either doing too much analysis up-front (and fooling yourself with false certainty), or you’re knowingly flying blind for way too long. This is not to say that you can’t or shouldn’t invest over many years, but the decision to continue investing should be made on a frequent enough basis such that discounting is pointless.
- The precision of the estimated benefits is very low. Applying discounting to benefit figures would imply false certainty, when the benefits can easily range by +/- 50%. This is even more true when the benefits are further into the future. Benefits estimates don’t tend to suffer from the same optimism bias as cost estimates, but the spread is typically worse. Discounting is like putting lipstick on a pig – it wastes your time and it just confuses the poor pig.
- The confusion it causes is unnecessary. Why bother explaining to people the whys and wherefores of discounting when the real cause for concern should be the cost of delay you are incurring. In the worst cases, I’ve seen organisations waste valuable time performing week’s worth of detailed analysis in the fuzzy-front-end — when the improved precision is miniscule compared to just one day’s worth of the estimated Cost of Delay.
Chris’ post was titled “Duration and the Time Value of Money“. This nicely points out how badly we’ve missed the point, focusing on all the wrong things:
- By focusing on working out the time value of money we ignored the far more important value of time.
- By focusing on the duration of the investment we ignored the far more important duration that the pipeline will be blocked.
The cost that really matters is the cost of delay. The benefits that really matter are the benefits you get from shortening lead time. Don’t bother with discounting.