I have previously used cocomo and cocomo 2 for estimating testing, where I have access to the appropriate design artefacts and some historical data that can be used to support the estimating process.
It can be effective however does not suit all test approaches and development methodologies. I was finding on the projects, which were at the time waterfall and object orientated, but traditionally coded, we were coming in with around a 5% deviation between estimate and plan.
It should be noted we also used to cross reference these estimates with other methods of caculation such as Function Point Analysis, to check for "process blind spots" or Anchoring Biases.
In the context of Agile projects I do not find it that useful or accurate and I am working on some agile estimating heuristics with other areas on the projects I am involved.
Agile Testers of the World UNIT!
As i aware that cocomo is more of a theoritical model which can give more accuracy when we adopt the heuristic approach. Lots of external factors involved here to model perfectly. What will happen if i am at the early stage of the SDLC. Because i dont have the previous data. In such scenario will be the Functional Point Analysis is the better bet, rather then cocomo.