Yesterday my google feed sent me the below video on Chinese Mega Projects completed last year.  There are some quite astonishing achievements within this clip.  

This started me thinking - how much time and resource did these projects take to complete and were they largely completed on time and to budget?  

Our own, well published, Crossrail was due to open in 2018 but has been postponed until 2021, with an original contingency budget of £3bn since being more than doubled.

Plucking another flagship project from history, the much celebrated Sydney Opera House cost a staggering 1,400% more than originally budgeted (as an aside, I would recommend reading Bill Bryson's chapter on its origins, plans and construction in his book "Down Under")

This led me to reading up on a few articles about the subject of over spending and time delays.

Oxford Global Projects Database (OGPD) obtains data on 12,000 construction / infrastructure projects.  Of these, time and budgetary failures accounted for:

 - 61% of aerospace projects

- 39% of rail projects

 - Only 7.8% total projects were delivered on time and to budget, with 0.5% delivering all the expected benefits

When delving into the murky world of why large projects fail to deliver on time or to budget Professor Bent Flyvbjerg of OMPD highlights:

 - Such projects are exceptionally complex and so it is hard for a Project Manager to gain a handle on where to start

 - Traditionally, such projects are broken up into smaller chunks and ran through the "Monte Carlo analysis" computer simulation to work out every eventuality and thus provide investors with three estimate points.  Flyvbjerg believes this is flawed due to human biases - the uniqueness of the work (often there are past projects that teams do not learn from); "Black Span events" are devastating events that rarely happen but hugely affect the running's of the project; lack of processes to communicate failures and issues

However, there is light at the end of the tunnel (every pun intended) assuming construction firms cultivate a culture of "data trust" champions.  This would involve companies logging mistakes and compensation events on every project undertaken, thus allowing future project data analysts to review information and forecast more realistically.  

This is the advice of behavioral scientists Daniel Kahneman and Amo Tversky, coining the process of "reference class forecasting" and thus abolishing the human bias, hypothetical approach to planning projects.

Clearly (and understandably) this leads to firms becoming nervous about releasing sensitive information on project work and so work is starting to combat this.  Only time will tell if it can become a reality.