As someone who has reviewed dozens of failures over the years, it’s clear that the times in which bad software killed a project are virtually non-existent. What’s more common is the projects were poorly run, that specifications weren’t translated in to a workable system, go-live took place before the data transfer was verified and the users trained: All of which gave the appearance that the main goal of the SI was to latch on to the client and suck as much revenue out of the project as possible, while hopefully leaving things messed up enough to guarantee the project extensions and modifications that could turn the project into an annuity for the SI.

Am I being too harsh? Even when things have gone remarkably well, and the majority of project have been successful, the cost burden of the traditional SI implementation has hampered attempts to get total cost of ownership under control and deliver a genuine return on investment. Again, this isn’t just the SI’s fault, the vendors have been colluding with them along these lines for years. To the mutual benefit of all. But if you look at the money pot that has flowed into enterprise software over the years, too much, way too much, has ended up with the SIs – certainly with respect to the value they have been able to deliver.