Sunday, September 14, 2008

Soft Part of Software Requirements ...

Michael Bolton in response to a discussion on software requirements mentioned this ...

"... There are many requirements that are matters of opinion, aesthetics, value, usability, compatibility that can't be subjected to a formula, can't be anticipated in advance, and which change over time "

Really true ... what we typically ignore about software requirements is that --

1. Requirements evolver over time consuming, human interactions and communications

2. Formal languages and domain vocabularies have a place in eliciting requirements but we should not confuse them to provide completely unambiguous, clear, testable requirements.

3. Sometimes these formal languages and domain vocabularies are costly, time consuming or simply not feasible.

4. A software specification should cater not only to software program that is being developed and also for the usage and all human interaction related that the program.

What do you say?


Friday, September 12, 2008

10 Commandments for Test Automation Outsourcing

Most of my test automation experience and learning comes from working with IT groups and GUI based automation using COTS tools. Outsourcing of testing is a top agenda on today’s IT manager and automation happens to be one of the popular and “most frequently discussed” item.

I am regularly asked to formulate, present and consult automation projects in IT space. Based on this experience, I am attempting to formulate commandments - about 10 of them. I believe these are few considerations that an IT manager who is thinking about outsourcing test automation work.

Automation is not an answer to your testing problems such as limited testing bandwidth, limited time to test, poor application quality etc. These are true testing problems – one way to address them is to set your “manual testing” right then think of Automation. If your manual testing is poor, throwing automation in that will only cause that “poor” testing be completed quickly.

Automation is White Elephant While automation has clear benefits only when it is treated carefully. Automation is not a turnkey, as your application undergoes changes; automation solution needs to be maintained. Plus there are recurring tool costs, training costs and test case/project management over heads. Is your vendor telling about this sufficiently? Is your vendor downplaying this aspect? Watch out …

Anything that is quickly creatable – is quickly perishable Do not believe claims of “script creation in minutes”, “Automated script generation” etc. Any enterprise level serious automation is software development and requires clear attention and methodical approach to design/create and maintain.

Your business users can not create and own Automation solutions Let your business users/SME’s and domain experts do what they are best at – business support. Do not believe claims of business users creating and maintaining automation solutions. It requires good amount of testing and automation knowledge.

Judge your vendor by the questions they ask about automation Suggesting an automation solution requires thorough study of context of project, release cycles, current state of project, future expected life of application, business expectations, state of automation practice etc. Vendors who propose a “standard”, “gold plated” automation solution without asking questions to probe the context are “simply” selling you something – be skeptical about such vendors.
Higher % of offshoring in automation higher investments to make While moving work to offshore will give cost benefits, when it comes to automation there are number of factors to be considered while deciding how much of automation can happen at offshore. Key thing is about creation of application environment at offshore. If a local application environment can be created at offshore with tool licenses – higher degree of automation work can happen at offshore. It is important to note that for GUI centric/COTS automation tool based automation, both automation tool and application should reside on same machine. Hence pay attention to your application environment and feasibility of creating local environment at vendors offshore location before deciding the amount of work that can move to offshore. Even when such local environment is available at offshore, certain activities like acceptance testing, demo to users, and certain types of test cases will need to be done only at onsite. Be sensitive to these factors. Make sure your vendor asks about this and suggests suitable alternatives.

Pay attention to dependencies and Quality of current test Artifacts A successful outsourced automation requires that project dependencies are well understood by all stakeholders. Access of application from offshore, access to testers/developers/SMEs on test cases, Access to reviewers and code acceptance people are few dependencies that one needs to track as part of project. If automation is on the basis of existing manual test cases- make sure that these are detailed enough and available in a form that can be sent across to vendor offshore team.

Decide Acceptance criteria Formulation of Acceptance criteria is not item that is often given least importance while planning outsourced automation projects. Identify an internal owner in your organization who will accept the code/solution delivered by the vendor. Make sure that this person gets engaged in the project from the beginning and formulates the acceptance criteria along with vendor technical team. Failure to identify automation acceptance person from your end and getting a formal agreement on acceptance criteria can leave you “High and Dry” and leaves an open space for the vendor to deliver any “working” automation code but not necessarily the one that stays for long time.

Avoid linking an automation project (and its deliverables) to application release dates one common mistake committed by IT groups outsourcing automation is to link the automation deliverables to immediate ensuing product/application release and cut down time/effort for manual testing. One IT manager said to his project team “We are having a major release for application XXX in January and we have automation solution coming from a vendor by December. Since Vendor has promised that 90% of manual testing will be automated – let us plan to allocate 2 weeks of testing instead of 10 weeks planned earlier. As promised by vendor we can deploy automation and cut down testing by more than 50%”. What is the problem here? What if Automation delivery slips for the reasons beyond control of everyone? What if development delays? What if due to poor test case quality and fast changing application, automation scripts give inconsistent results? This will result in conflict - “believe automation” or “get some good manual testing done”? Such conflicts can severely impact your releases and hence business plans.

Record and Playback (RP) Automation is for Kids This one point can never be overemphasized. Many IT managers still feel that record playback features of industry standard automation tools can help them to create automation quickly and thereafter their own resources should be able to record and create/maintain automation scripts. However, the experience has shown time and again that RP approach is not beyond learning about how automation tool works with application and can not be used for real time, sustainable automation. If vendor proposes this as a part of the solution – you should be alert and suspect the abilities of this vendor to deliver automation solution

Do you agree with these commandments? Any different experience?

Some additional tips :
What a Vendor should ask you
- questions about your manual testing practice
- Your objectives of Automation and expectations
- Readiness to take up automation in terms of test cases, application state and environment
- Ask about your expectations on ROI

What vendor should suggest you
- Plan for automation maintenance when supplier is gone.
- Automation may or may not reduce cycle time – that depends upon nature of tests, application technology stack, nature of tool etc
- Automation may or may not reduce the cost of testing.

What to look for in an automation proposal
- Acceptance criteria
- Automation design details – how tests will be structured
- Environment related assumptions
- Pre-requisites about tool licenses, test cases, test data, access to developers, testers and business users (for clarifications about test cases)

You can not automate testing – all of testing as testing is intrinsically a human thinking and investigation activity. What you in reality claim to automate is some portions of testing – namely “test execution” of some select test cases. Note that activities like test design, bug investigation and logging, Test results verification etc are still to be done by a human tester. Automation will take you some places but not all. Unfortunately, the places where it does not take you – are the ones where real problems lie – only a skilled human tester can help.