Total cost of ownership (TCO) is a measurement of the cost of assets that includes the cost used to run them as well as the cost of acquiring them. It is not used in accounting (recording and reporting transactions). It is an estimate used to evaluate possible purchases, most commonly of IT systems, and is primarily of interest to those involved in such decisions.
The things being purchased may not even be assets in the accounting sense at all: for example open source software may have no purchase cost, but in practical terms it may be necessary to pay the supplier for support while it is in use. The use of the word “project” is usually preferable for this reason.
There are a number of problems with TCO:
- It does not take time value of money or risk into account.
- It does not take into account the option value of more flexible systems.
- TCO numbers that claim to compare one system to another in general, independent of circumstances, should be treated as suspect.
- The number of hard to estimate or forecast numbers needed make TCO numbers intrinsically unreliable.
The first of these is a serious flaw: it puts TCO in the same class of fundamentally flawed approaches as ARR. It might be possible to incorporate these, but then we might as well calculate an NPV.
The lack of consideration of option value is less serious for TCO per se because it is unlikely that any assessment of such complex purchases will do this: it is too hard to quantify.
To see why it is important consider two products with the same TCO: one with a large upfront payment to purchase it and low maintenance costs, the other with no initial cost and a high maintenance cost. The former may be cheaper, but what happens if requirements change and the system has to be replaced halfway through its life? The cost of the latter over its lifetime is suddenly halved.
TCO numbers that compare products in general should be treated with suspicion. Firstly, such estimates are often produced with the sponsorship of vendors and their impartiality is therefore questionable. Secondly, they will vary with circumstances: for example people who know product A may be more expensive than those who know product B, but if you already have people who know A but would have to hire people who know B, it makes the comparative costs very different.
The final objection to TCO is a very general one that includes some of the above: it looks only at costs, independent of benefits. Again, an NPV is the ideal solution, but harder to do.