Category Archives: Software measurement

What comes first - estimates or requirements?

We’ve seen many advancements of late in process improvement, better software estimating models, introducing flexibility and agility into software development, formal project management, limiting work-in-progress, etc. – all which incrementally advance projects forward and promise to reduce costly rework.  However, none of these methods addresses one of the most fundamental questions in software development:  what comes first estimates or requirements?

Chicken or Egg?If estimates come first… It seems like putting the cart before the horse if estimate precede requirements, but that is precisely how a good number of software projects go.  To begin with, someone has an idea or a business problem that software can solve but before any work even one dollar can be spent, the work has to go into next year’s budget and get funded.  This means figuring out some sort of estimate of work that has yet to be scoped out.  “Is it going to be bigger than a bread box and smaller than a football field?” is one way of saying – we need a “rough ball park” (guesstimate) that we can use in the budgeting process. Unfortunately this guesstimate process is clearly flawed because it is based on invisible and ether-like requirements. As such, the guesstimate is prone to + or – 500% or more variance once the real requirements are known.

This is like saying – how much will it cost to build a house for my family, just give me a rough estimate so I can go to the bank and arrange a mortgage.  This would be an absurd behavior – especially when one usually doesn’t get a mortgage in advance, and especially because the cost will vary depending on where, how big, how custom, and how the house will be built.  If  one secures a $500K mortgage amount – it gives an upper limit but doesn’t guarantee that a suitable house can actually be done for that amount.  Yet, we engage in this behavior in IT all the time – we guess(timate) for the budget cycle, the amount gets slashed during meetings, and ultimately the fixed price (based on little information) becomes the project budget!

If requirements come first… then in many companies nothing will ever get built and problems will remain the same forever.  Analysis paralysis is common (especially with shifting new business requirements) which gives rise to the support of agile and extreme programming approaches to requirements.  Many companies shifted their support from the arduous front end heavy “waterfall” methods of software development in favor of “progressive requirements elaboration” whereby requirements are discovered along the way. As such, requirements are always evolving with new user stories emerging only after the earlier ones are delivered in software.  So what happens when requirements are needed to build a better estimate (and thereby make sure the project has enough budget) yet an estimate is required before one can begin to scope out the requirements that will fit the budget?  It is a circular situation akin to the chicken and egg conundrum that has plagued humankind for years.

Pathway to cooperative results…One method that proves to work well with this dilemma is scope management – whereby a business “project” (more likely a program of work) is divided into sub-projects, scoped out at the highest level, quality requirements are thought about, and traceable estimates ensue.  More to come on this topic in the next post…

To your successful projects!


Carol Dekkers

Carol Dekkers provides realistic, honest, and transparent approaches to software measurement, software estimating, process improvement and scope management.  Call her office (727 393 6048) or email her ( for a free initial consultation on how to get started to solve your IT project management and development issues.

For more information on northernSCOPE(TM) visit (in English pages) and for upcoming training in Tampa, Florida  — April 26-30, 2010, visit

=======Copyright 2010, Carol Dekkers ALL RIGHTS RESERVED =======