STARWEST 2011 - Software Testing Conference

One of the challenges associated with testing in agile projects is selecting test techniques that “fit” the dynamic nature of agile practices. How much functional and non-functional testing should you do? What is the appropriate mix of unit, integration, regression, and system testing? And how do you balance these decisions in an environment that fosters continuous change and shifting priorities?

In most organizations cost reduction is still the number one motivation for test process improvement. Although several formal improvement models are popular, they require formal assessments, process change working groups, extensive implementation programs, and new organizational structures. Instead, you can quickly implement measures that improve your testing process incrementally within your day-to-day activities.

Acceptance Test-Driven Development (ATDD) uses specification by example to define expressive automated tests that facilitate project communication and drive product design. Acceptance testing tools like Cucumber, FitNesse, and Concordion are powerful tools for building a common, ubiquitous language connecting business people, developers, and testers. A common language leads to common understanding and ultimately to better software that meets business needs and delights customers.

Many of us have struggled with test estimation. We have tried simple, heuristic models to craft a best guess-often without much success. We have also tried using a variety of complex, scientific models to calculate an accurate number. The problem is, we are usually fooled by the models-both simple and complex ones-and either overestimate testing needs or are lulled into impossible commitments. Lynn McKee and Nancy Kelln explore the realities of test estimation and propose a new mindset for handling estimation requests.