In this tutorial, we will learn what is Sanity Testing and Smoke Testing in Software Testing. We will also learn the key difference between Sanity and Smoke testing with examples.
Most of the times we get confused between the meaning of Sanity testing and Smoke testing. First of all, these two testing are way “different” and performed during different stages of a testing cycle.
What You Will Learn:
- Sanity Testing
- Smoke Testing
- Difference Between Smoke and Sanity Testing
Sanity testing is done when as a QA we do not have sufficient time to run all the test cases, be it functional, UI, OS or browser testing.
Hence, I would define sanity testing as a test execution which is done to touch each implementation its impact but not thoroughly or in-depth, it may include functional, UI, version etc. testing depending on the implementation and its impact.
Don’t we all fall in a situation when we have to sign off in a day or two but the build for testing is still not released?
Oh yes, I bet that you also must have faced this situation at least once. Well, I faced it a lot because my project(s) were mostly agile and at times we were asked to deliver the same day. Oops, how could I test and release the build in a stretch of hours?
I used to go nuts at times because even if it was a small functionality, the implication could be tremendous. And as an icing on the cake, clients sometimes simply refused to give extra time. How could I complete the whole testing in few hours, verify every functionality, bugs and release it?
The answer to such problem was very simple, i.e nothing but using Sanity testing strategy.
When we do this testing for a module or functionality or a complete system, the test cases for execution are selected such that they will touch all the important bits and pieces of the same i.e. wide but shallow testing.
At times the testing is even done randomly with no test cases. But remember, Sanity test should be done only when you are running short of time, never use this for your regular releases. Theoretically, this testing is a subset of the regression testing.
Out of my 8+ years career in Software Testing, for 3 years I was working in Agile methodology and that was the time when I mostly used sanity test.
All the big releases were planned and executed in a systematic manner but at times, small releases were asked to be delivered as soon as possible. We didn’t get much time to document the test cases, execute, do the bug documentation, do the regression and follow the whole process.
Hence following are some of the key pointers that I used to follow under such situations:
=> Sit with the manager and the dev team when they are discussing the implementation because they have to work fast and hence we can’t expect them to explain us separately.
This would also help you to get an idea about what they are going to implement, which area will it be affecting etc., this is a very important thing to do because at times we simply don’t realize the implications and if any existing functionality is going to be hampered (at worst).
=> As you are short of time, by the time the development team is working on the implementation, you can note down the test cases roughly in tools like Evernote etc. But make sure that you write them somewhere so that you can add them later to the test case tool.
=> Keep your testbed ready as per the implementation and if you feel that there are any red flags like some specific data creation if a test bed will take time (and it’s an important test for the release), raise those flags immediately and inform your manager or PO about the roadblock. Just because the client wants asap, it doesn’t mean that a QA will release even if it is half tested.
=> Make an agreement with your team and manager that due to time crunch you will only communicate the bugs to the development team and the formal process of adding, marking the bugs for different stages in the bug tracking tool will be done later in order to save time.
=> When the development team is testing at their end, try to pair with them (called dev-QA pairing) and do a basic round on their setup itself, this will help to avoid the to and fro of the build if the basic implementation is failing.
=> Now that you have the build, test the business rules and all the use cases first. You can keep tests like validation of a field, navigation etc for later.
=> Whatever bugs you find, make a note of all of them and try to report all of them together to the developers rather than reporting individually because it will be easy for them to work on a bunch.
=> If you have a requirement for the overall performance or stress or load testing, then make sure that you have a proper automation framework for the same. Because it is near to impossible to manually test these with sanity test.
=> This is the most important part, the last step of your sanity test strategy – “When you draft the release email or the document, mention all the test cases that you executed, the bugs found with a status marker and if anything was left untested mention it with the reasons” Try to write a crisp story about your testing which will convey everyone about what has been tested, verified and what has not been.
I followed this religiously when I was using this testing.
Let me share my own experience:
#1) We were working on a website and it used to popup ads based on the keywords. The advertisers used to place the bid for particular keywords which had a screen designed for the same. The default bid value used to be shown as 0.25$, which the bidder could change.
There was one more place where this default bid used to show up and it could be changed to other value as well. The client came with a request to change the default value from 0.25$ to 0.5$ but he mentioned only the obvious screen.
During our brainstorming discussion, we forgot (?) about this other screen because it wasn’t used much for that purpose. But while testing when I ran the basic case of the bid being 0.5$ and checked end to end, I found that the cronjob for the same was failing because at one place it was finding 0.25$.
I reported this to my team and we made the change and successfully delivered it the same day itself.
#2) Under the same project (above mentioned), we were asked to add a small text field for notes/comments for bidding. It was a very simple implementation and we committed to delivering it the same day.
Hence like I mentioned above, I tested all the business rules and use cases around it and when I did some validation testing, I found that when I entered a combination of special characters like </>, the page crashed.
We thought over it and figured out that the actual bidders won’t in any case use such combinations. Hence, we released with a well-drafted note about the issue. The client accepted that as a bug but agreed with us to implement it later because it was a severe bug but not a prior one.
#3) Recently, I was working on a mobile app project, we had a requirement to update the time of delivery shown in the app as per the time zone. It was not only to be tested in the app but also for the web service.
While the development team was working on the implementation, I created the automation scripts for the web service testing and the DB scripts for changing the time zone of the delivery item. This saved my efforts and we could achieve better results in a short duration.
Sanity Testing Vs Regression Testing
Given below are few differences between these:
|Regression Testing|| |
|1||Regression testing is done to verify that the complete system and bug fixes are working fine.||Sanity testing is done at random to verify that each functionality is working as expected.|
|2||Every tiniest part is regressed in this testing.||This is not a planned testing and is done only when there’s a time crunch.|
It is a well elaborate and planned testing.
|This is not a planned testing and is done only when there’s a time crunch.|
|4||An appropriately designed suite of test cases is created for this testing.||It may not every time be possible to create the test cases; a rough set of test cases is created usually.|
|5||This includes in-depth verification of functionality, UI, performance, browser/OS testing etc. i.e. every aspect of the system is regressed.||This mainly includes verification of business rules, functionality.|
|6||This is a wide and deep testing.||This is a wide and shallow testing.|
|7||This testing is at times scheduled for weeks or even month(s).||This mostly spans over 2-3 days max.|
Strategy for Mobile Apps Testing
You must be wondering why I am mentioning specifically about the mobile apps here?
The reason is that OS and browser version for web or desktop apps do not vary much and especially the screen sizes are standard. But with mobile apps, the screen size, the mobile network, the OS versions etc affect the stability, look and in short, the success of your mobile app.
Hence a strategy formulation becomes critical when you are performing this testing on a mobile app because one failure can land you in big trouble. The testing must be done smartly and with caution.
Following are some pointers to help you perform the this testing successfully on a ‘mobile app’:
#1) First of all, analyze the impact of OS version on the implementation with your team.
Try to find answers to questions like, will the behavior be different across versions? Will the implementation work on the lowest supported version or not? Will there be performance issues for the implementation of versions? Is there any specific feature of the OS that may impact the behavior of the implementation? etc.
#2) On the above note, analyze for the phone models also i.e. are there any features of the phone that will impact the implementation? Is the implementation behavior changing with GPS? Is the implementation behavior changing with the phone’s camera? etc. If you find that there’s no impact, avoid testing on different phone models.
#3) Unless there are any UI changes for the implementation I would recommend keeping UI testing on the least priority, you can inform the team (if you want) that UI will not be tested.
#4) To save your time, avoid testing on good networks because it is obvious that the implementation is going to work as expected on a strong network. I would recommend to start with testing on a 4G or 3G network.
#5) This testing is to be done in less time but make sure that you do at least one field test unless it’s a mere UI change.
#6) If you must test for a matrix of different OS and their version, I would suggest that you do it in a smart way. For instance, choose the lowest, medium and the latest OS-version pairs for testing. You can mention in the release document that not every combination is tested.
#7) On a similar line, for UI implementation sanity test, use small, medium and large screen sizes to save time. You can also use simulator and emulator etc.
Sanity testing is performed when you are running short of time and hence it is not possible for you to run each and every test case and most importantly you are not given enough time to plan out your testing. To avoid the blame games, it is better to take precautions.
In such cases lack of written communication, test documentation and miss outs are quite common.
To ensure that you don’t fall prey to this, make sure that:
- Never accept a build for testing until you are not given a written requirement shared by the client. It happens that clients communicate changes or new implementations verbally or in chat or a simple 1 liner in an email and expect us to treat that as the requirement. Compel your client to provide some basic functionality points and acceptance criteria.
- Always make rough notes of your test cases and bugs if you are not having sufficient time to write them neatly. Never leave these undocumented. If there is some time, share it with your lead or team so that if anything is missing they can point it out.
- If you and your team are short of time, make sure that the bugs are marked in the appropriate state in an email? You can email the complete list of bugs to the team and make the devs mark them appropriately. Always keep the ball in the other’s court.?
- If you have automation framework ready, use that and avoid doing manual testing, that way in less time you can cover more.
- Avoid scenario of “release in 1 hour” unless you are 100% sure that you will be able to deliver.
- Last but not the least, as mentioned above, draft a detailed release email communicating what is tested, what is left out, reasons, risks, which bugs are resolved, what are ‘latered’ etc.
As a QA, you should judge what is the most important part of that implementation which needs to be tested and what are the parts that can be left out or basic-tested.
Even in a short time, plan a strategy about how you want to do and you will be able to achieve the best in the given time frame.
Smoke testing is not an exhaustive testing but it is a group of tests that are executed to verify if the basic functionalities of that particular build are working fine as expected or not. This is and should always be the first test to be done on any ‘new’ build.
When the development team releases a build to the QA for testing, it is obviously not possible to test the entire build and verify immediately if any of the implementations is having bugs or if any of the working functionality is broken.
In the light of this, how will a QA make sure that the basic functionalities are working fine?
The answer to this will be to perform a Smoke Testing.
Once the tests are marked as Smoke tests (in the test suite) pass, only then the build is accepted by the QA for in-depth testing and/or regression. If any of the smoke tests fail, the build is rejected and the development team needs to fix the issue and release a new build for testing.
Theoretically, Smoke test is defined as a surface-level testing to certify that the build provided by the development team to the QA team is ready for further testing. This testing is also performed by the development team before releasing the build to the QA team.
This testing is normally used in Integration testing, System testing, and Acceptance level testing. Never treat this as a substitute for the actual end to end complete testing. It comprises of both positive and negative tests depending on the build implementation.
Smoke Testing Examples
This testing is normally used for Integration, System and Acceptance testing.
In my career as a QA, I always accepted a build only after I had performed a smoke test. So, let’s understand what is smoke test from the perspective of all these three testing, with some examples.
#1) Acceptance Testing:
Whenever a build is released to the QA, smoke test in the form of an acceptance testing should be done.
In this test, the first and most important smoke test is to verify the basic expected functionality of the implementation. Like this, you should verify all the implementations for that particular build.
Let us take the following examples as in implementations done in a build to understand the smoke tests for those:
- Implemented the login functionality to allow the registered drivers to log in successfully.
- Implemented the dashboard functionality to show the routes that a driver is to execute today.
- Implemented the functionality to show an appropriate message if no routes exist for a given day.
In the above build, at acceptance level, the smoke test will mean to verify that the basic three implementations are working fine. If any of these three is broken, then the QA should reject the build.
#2) Integration Testing:
This testing is usually done when the individual modules are implemented and tested. In the Integration level, this testing is performed to make sure that all the basic integration and end to end functionalities are working fine as expected.
It may be an integration of two modules or all modules together, hence the complexity of the smoke test will vary depending on the level of integration.
Let us take the following example of integration implementation for this testing:
- Implemented the integration of route and stops modules.
- Implemented the integration of arrival status update and reflecting the same on the stops screen.
- Implemented the integration of complete pick up till the delivery functionality modules.
In this build, the smoke test will not only verify these three basic implementations but for the third implementation, a few cases for the complete integration too. It helps a lot to find out the issues that get introduced in integration and the ones that went unnoticed by the development team.
#3) System Testing:
As the name itself suggests, for system level, the smoke testing includes tests for the most important and commonly used workflows of the system. This is done only after the complete system is ready & tested, and this testing for system level can be referred as smoke testing before regression testing also.
Before starting the regression of the complete system, the basic end to end features is tested as a part of smoke test. The smoke test suite for the complete system comprises of the end to end test cases that the end users are going to use very frequently.
This is usually done with the help of automation tools.
Importance in SCRUM Methodology
Nowadays, the projects hardly follow the Waterfall methodology in project implementation, mostly all the projects follow Agile and SCRUM only. Compared to the traditional waterfall method, smoke testing holds high regards in SCRUM and Agile.
I worked for 4 years in SCRUM. And as we know that in SCRUM, the sprints are of shorter duration and hence it is of extreme importance to do this testing so that the failed builds can immediately be reported to the development team and fixed.
Following are some takeaway on the importance of this testing in SCRUM:
- Out of the fortnight sprint, halftime is allocated to the QA but at times builds to the QA are delayed.
- In sprints, it is best for the team that the issues are reported at the early stage.
- Each story has a set of acceptance criteria, hence testing the first 2-3 acceptance criteria is equal to smoke testing of that functionality. Customers reject the delivery if a single criterion is failing.
- Just imagine what will happen if its 2 days that the development team delivered you the build and only 3 days are remaining for the demo and you come across a basic functionality failure.
- On an average, a sprint has stories ranging from 5-10, hence when the build is given it is important to make sure that each story is implemented as expected before accepting the build into testing.
- When the complete system is to be tested and regressed, a sprint is dedicated to the activity. A fortnight maybe little less to test the whole system, hence it is very important to verify the most basic functionalities before starting the regression.
Smoke Test Vs Build Acceptance Testing
Smoke testing is directly related to Build Acceptance Testing (BAT).
In BAT, we do the same testing – to verify if the build has not failed and that the system is working fine or not. Sometimes, it happens that when the build is created, some issues get introduced and when it is delivered, the build doesn’t work for the QA.
I would say that BAT is a part of smoke check because if the system is failing how can you as a QA accept the build for testing? Not just the functionalities, the system itself has to work before the QA’s proceed with the in-depth testing.
Smoke Test Cycle
The following flowchart explains the smoke testing cycle. Once a build is deployed to a QA, the basic cycle followed is that if the smoke test passes, the build is accepted by the QA team for further testing but if it fails, the build is rejected until the reported issues are fixed.
Who Should Perform Smoke Test?
Not the whole team is involved in this type of testing to avoid the wastage of time of all the QA’s.
Smoke testing is ideally performed by the QA lead who decides based on the the result as for whether to pass the build to the team for further testing or reject it. Or in the absence of the lead, the QA’s themselves can also perform this testing.
At times, when the project is a large scale one, a group of QA can also perform this testing to check for any showstoppers. But this is not so in the case of SCRUM because SCRUM is a flat structure with no Leads or Managers and each tester has their own responsibilities towards their stories.
Hence individual QA’s perform this testing for the stories that they own.
Why should we Automate Smoke Tests?
This testing is the first test to be done on a build released by the development team(s). Based on the results of this testing, the further testing is done (or the build is rejected).
The best way to do this testing is to use a automation tool and schedule the smoke suite to run when a new build is created. You may be thinking why should I “automate the smoke testing suite”?
Let us look at the following case:
Let’s say that you are a week away from your release and out of the total 500 test cases, your smoke test suite comprises of 80-90. If you start executing all these 80-90 test cases manually, imagine how much time will you take? I think 4-5 days (minimum).
But if you use automation and create scripts to run all these 80-90 test cases then ideally, these will be run in 2-3 hours and you will have the results with you instantly. Didn’t it save your precious time and give you the results about the build in much less time?
5 years back, I was testing a financial projection app, which took inputs about your salary, savings etc., and projected your taxes, savings, profits depending on the financial rules. Along with this, we had customization for countries that depend on the country and its tax rules used to change (in the code).
For this project, I had 800 test cases and 250 were smoke test cases. With the use of Selenium, we could easily automate and get the results of those 250 test cases in 3-4 hours. It not only saved our time but showed us ASAP about the showstoppers.
Hence, unless it is impossible to automate, do take the help of automation for this testing.
Advantages and Disadvantages
Let us first take a look at the advantages as it has a lot to offer when compared to the few disadvantages.
Following are the Advantages:
- Easy to perform.
- Reduces the risk.
- Defects are identified at a very early stage.
- Saves efforts, time and money.
- Runs quickly if automated.
- Least integration risks and issues.
- Improves the overall quality of the system.
- This testing is not equal to or a substitute for complete functional testing.
- Even after smoke test passes, you may find showstopper bugs.
- This type of testing is best suitable if you can automate else a lot of time is spent in manually executing the test cases especially in large-scale projects having around 700-800 test cases.
Smoke testing should definitely be done on every build as it points out the major failures and showstoppers at a very early stage. This applies not only to new functionalities but also to the integration of modules, fixing of issues and improvisation as well. It is a very simple process to perform and get the correct result.
This testing can be treated as the entry point for complete functional testing of a functionality or system (as a whole). But before that, the QA team should be very clear about what tests are to be done as smoke tests. This testing can minimize the efforts, save time and improve the quality of the system. It holds a very important place in sprints as the time in sprints is less.
This testing can be done both manually and also with the help of automation tools. But the best and preferred way is to use automation tools to save time.
Difference Between Smoke and Sanity Testing
Most of the times we get confused between the meaning of Sanity testing and Smoke testing. First of all, these two testing are way “different” and performed during different stages of a testing cycle.
|S. No.||Smoke Testing||Sanity Testing|
|1||Smoke testing means to verify (basic) that the implementations done in a build are working fine.||Sanity testing means to verify the newly added functionalities, bugs etc. are working fine.|
|2||This is the first testing on the initial build.||Done when the build is relatively stable.|
|3||Done on every build.||Done on stable builds post regression.|
Following is a diagrammatic representation of the difference:
- This testing originated in the hardware testing practice of turning on a new piece of hardware for the first time and considering it a success if it does not catch fire and smoke. In the software industry, this testing is a shallow and wide approach whereby all areas of the application without getting into too deep, is tested.
- A smoke test is scripted, either using a written set of tests or an automated test
- A Smoke test is designed to touch every part of the application in a cursory way. It’s shallow and wide.
- This testing is conducted to ensure whether the most crucial functions of a program are working, but not bothering with finer details. (Such as build verification).
- This testing is normal health check up to a build of an application before taking it to testing in depth.
- A sanity test is a narrow regression test that focuses on one or a few areas of functionality. Sanity testing is usually narrow and deep.
- This test is usually unscripted.
- This test is used to determine a small section of the application is still working after a minor change.
- This testing is a cursory testing, it is performed whenever a cursory testing is sufficient to prove the application is functioning according to specifications. This level of testing is a subset of regression testing.
- This is to verify whether requirements are met or not, checking all features breadth-first.
Hope we have clearly explained these two different and important software testing types. Let us know if you have any queries.