6.1 Key considerations
One good way to go about testing a complex form is to test it at each stage of the form building process. This can avoid having a mountain of technical errors when finalising the questionnaire (for example when you pass the form through a technical verification tool). How often you want to test it will depend on the length and settings you have planned, but we suggest that you do so at the end of each section or after a particularly tricky set of questions. Use the checklist in the following section to determine what you might need to test in your own form.
You can try and follow a pattern when you check all the settings you have built into your form:
- Once deployed, try to enter deliberately impossible combinations or wrong values of answers into the form. For example numbers which are excessively large or small, dates in the future, or answers to different questions which contradict each other. This will highlight the need for more constraints.
- On your second submission, select as many possible combinations of each answer as possible and check which of the remaining questions appear or are hidden. This will help test the relevancies.
- To test skip patterns, check all the possible “paths”. To design the form, you will have set up skip patterns where some questions appear (or not) depending on prior answers. Ensure you have mapped these potential combinations (“paths”) and ensure each one is tested.
- In order to test if your calculations are working as expected, the best approach is to enter random responses and then after completing the form a number of times, export the data to excel. In Excel you can then add a column and re-create the calculation with an excel formula to see if the results are the same in all cases. If you prefer, you can also add a working note presenting the result to make your tests easier.
- Verify that the mandatory single-choice/multiple option questions all have the necessary safeguards (“Don’t know”, “Unknown”, “Other”, “N/A”, “None of the above”, “Refuse to answer”, etc.) to avoid forcing the enumerator to enter incorrect information to be able to submit his data.
- Make sure that for cascading lists the “other” options are included at every stage (Check the settings that permits to repeat the same answer at different level of your cascading list).
- Check appearance settings on the different interfaces you might be using, such as a mobile, a tablet or a web browser… Make sure theses settings are compatible with your device size (grouping questions on the same screen or using big matrix-type questions when you have a small screen is not recommended) and your question meaning (using Likert scales for questions where it does not make sense).
- Verify that all is translated which means: questions, hints, customized constraint messages and the question choices.