Historically teams I've worked with have taken a few varying approaches when designing tests against acceptance criteria. One is to have the business define the feature, while the team help define the acceptance criteria. Ultimately the business gets the final say if they agree, and further acceptance criteria is either added or removed. The strength of this approach is everyone is involved with the process so nothing is missed or misunderstood. The biggest flaw with this style is that the documentation produced is often verbose, using wordy Given-When-Then scenarios. Using this plan a test plan is then created, mapping tests to acceptance criteria.
An alternative approach is have the business define both the feature and acceptance criteria while the team come up with a corresponding test strategy. This more technical approach allows for a separation of testing activities and test categories. Finally the test plan is replayed back to the business and correlated against acceptance criteria. A negative of this approach is not everyone is involved with the task at the same time. This means there can be some disconnect with what the business is actually asking for. Both approaches work though they can yield mixed results on a case by case basis.
The QA Matrix
I've recently been introduced to the concept of a testing/QA matrix, which is a far more condensed and simplified solution. It has the benefit of the whole team being engaged, while producing nothing more than a simple table that can fit comfortably on a A4 page. The left hand column includes each condition of acceptance, while the other columns should have a mark to indicate the type of test that will cover this functionality. An example is below.
Unit Integration Acceptance Contract Manual COA X COA X X COA X ...
The beauty of this matrix is that at a glance you can see where you testing efforts lie. If too much occurs on the right of the matrix you may need to re-consider and question your approach. Is there a way to limit the more expensive style of tests and still gain confidence? Other questions can arise around test coverage and whether higher level tests are needed.
When producing this matrix the whole team including the business should be involved. By having everyone together, decisions can be made quickly with everyone in agreement. Additionally it allows debate and discussion around how each feature should be tested.
For higher level tests these can be directly translated into automated tests. While the lower level tests need to confirmed at a later date once the code is complete.
Along side the QA matrix it may be worth while adding a simple diagram of the components that will be involved such as web servers, databases and so on. This can aid discussion and highlight hot spots for changes or tests.
Finally for demonstration to the business the matrix can be used as a form contract for signing off functionality. Once the feature is complete it is simply a case of finding the corresponding tests, confirming their existence and making a note of the commit that included them.