I'm sure you've all heard of the automated testing pyramid, I'll describe it briefly here, but you can read all about it here.
.jpg)
It's essentially a strategy that shows good practice ratio of Acceptance Tests (generally UI) to Integration Tests to Unit Tests, and here it is here in a simple form.
It states that it is a good ratio to have your testing covered with 10% of acceptance tests, 20 % integration tests and 70% unit tests. Why is that you may ask? The primary focus of this is on Return on Investment, by finding bugs/breakages at Unit test level you are finding cheap bugs, as Unit Tests are quick and easy to maintain, whereas acceptance tests, whilst having value, are harder to maintain and take longer to run.
Obviously, it's not a strict ratio, but I think it's a good practice to try and live by.
However, I digress, the main point of this post is to put another spin on the automation triangle, and is possibly more QA centric than the automation pyramid, as I don't often see (rightly or wrongly) QA getting involved in creation of Unit Tests/Integration tests, I feel you can visualise the ratio in an easier way, and that is in the form of an Iceberg.
Interesting Fact: only around 11% of an iceberg is actually visible...
Now if we apply that interesting fact to automated testing, (I guess you can see where I am going with this) we can say that the 11% that we can see as QA is the acceptance tests, these are tests that in general have been driven by the QA and are what will often fall into our domain to maintain and create.
The remaining 89%, are more dev focused, in that unit tests and integration tests are generally maintained by the developers, and not the QA department (at least in my experience). So this allows the QA team to work on acceptance tests and get them working effectively, and leaving the Unit and Integration tests to the developers (although it is definitely wise to get involved as much as you can).
Please don't think I'm saying that unit tests aren't important (quite the opposite as my previous blog posts will attest to), as QA need to be aware of what Unit Tests and Integration tests are to be run, as this will influence what tests are run as part of the acceptance tests, you only have to look at the Titanic to see what happened there when the rest of the iceberg was ignored.
.jpg)
It's essentially a strategy that shows good practice ratio of Acceptance Tests (generally UI) to Integration Tests to Unit Tests, and here it is here in a simple form.
It states that it is a good ratio to have your testing covered with 10% of acceptance tests, 20 % integration tests and 70% unit tests. Why is that you may ask? The primary focus of this is on Return on Investment, by finding bugs/breakages at Unit test level you are finding cheap bugs, as Unit Tests are quick and easy to maintain, whereas acceptance tests, whilst having value, are harder to maintain and take longer to run.
Obviously, it's not a strict ratio, but I think it's a good practice to try and live by.
However, I digress, the main point of this post is to put another spin on the automation triangle, and is possibly more QA centric than the automation pyramid, as I don't often see (rightly or wrongly) QA getting involved in creation of Unit Tests/Integration tests, I feel you can visualise the ratio in an easier way, and that is in the form of an Iceberg.
Interesting Fact: only around 11% of an iceberg is actually visible...
Now if we apply that interesting fact to automated testing, (I guess you can see where I am going with this) we can say that the 11% that we can see as QA is the acceptance tests, these are tests that in general have been driven by the QA and are what will often fall into our domain to maintain and create.
The remaining 89%, are more dev focused, in that unit tests and integration tests are generally maintained by the developers, and not the QA department (at least in my experience). So this allows the QA team to work on acceptance tests and get them working effectively, and leaving the Unit and Integration tests to the developers (although it is definitely wise to get involved as much as you can).
Please don't think I'm saying that unit tests aren't important (quite the opposite as my previous blog posts will attest to), as QA need to be aware of what Unit Tests and Integration tests are to be run, as this will influence what tests are run as part of the acceptance tests, you only have to look at the Titanic to see what happened there when the rest of the iceberg was ignored.
Nice post Gareth. I've been working the same iceberg metaphor around in my mind for a while now.
ReplyDeleteI hadn't limited the metaphor to just automation though - in my opinion, the iceberg is a metaphor for the test strategy.
Every member on the development team should be aware of just how much iceberg there is, both above & below the water - i.e. what is the entire test strategy of the development team.
I really should get my thoughts onto paper for some open criticism.
Thanks for the post Gareth,
Duncs
One thing to note is that Mike Cohn has stated that because of the types of tests and where they are at in the pyramid (which causes it to be a pyramid shape) structure you will get this distribution.
ReplyDeleteHe doesn't necessarily say this is the way it should be as common practice, he states that it comes out this way because at the lower levels you have more 'atomic' tests (Unit/Code level tests/asserts) in order to get coverage of the code.
The benefit is finding issues/defects sooner and reducing the amount of churn/rework (lower cost ratio in comparison to later on down the road to release).
The whole purpose IMO of the pyramid is to get testing going earlier and to leverage the tools at hand (xUnit Harnesses and automation). Acceptance Test level tests are valuable, don't disregard them because of the smaller percentage. These tests via proper framework architecture and planning will be easier to build and maintain. They have value as they represent the final integration of the system with the user.
By following the suggestions of the pyramid you start at the micro level and work towards the macro. This provides greater coverage and exercising of the software under test. By using automation you leverage a tool/machine to aid in the execution of those tests and get efficiency gains as a benefit. That is what Mike is really pushing.
Jim Hazen