Now, let’s start creating the survey. First, we would like to focus on general music websites or app usage behaviour. This serves as a hook for our survey, and more importantly, this would later on help us segment our users into cohorts and define behaviour based on these cohorts.
Naturally, the first question we would ask to assess music app usage behaviour could be: Do you use any of the following music streaming services? However, this would not be a good choice of question. The reason being lack of specificity of the time period. That is, it may simply be that you have used a music service once in your lifetime. The majority of users would report yes to such generic questions and it doesn’t yield any valuable information, and the second, it is a yes, no question. While it is okay to ask users for a yes, no question in a survey, a better way to frame the same question would be.
In the past 30 days, which of the following online music services have you used? Please select all that apply, and here we display a bunch of options. This question is more specific and it has a time period associated with it as well. So, we get a more comprehensive response from the users. Next, we could potentially focus on the frequency of usage.
Hence a question like do you use any of the streaming services daily would seem like a valid question. However, this is a leading question, and hence people might end up answering yes sure, since the survey is all about music. Another interesting thing to note is regarding user psychology. Users would often over report positive behaviours. For example, for the survey question: Do you work out or how many times a week do you typically work out? Majority of the people would report a frequency higher than their actual workout frequency, because it is a positive behaviour.
So, instead we can ask in a typical day on a scale of 1 to 5, how likely are you to use music streaming services? Moving on, we would like to even go deeper and understand the R spent on music streaming apps. We can ask in a typical day how many hours do you spend listening to music, on music streaming services, and this can be a range-bound question. So, we can have options like 0-1, 2-3, 4-5, 6-7 and more than 7.
If you were to create questions for a survey on analysing people’s taste in music, you could focus on the following traits:
- App-usage behaviour
- Frequency of usage
- Duration of usage
Since we are interested in mood-based music recommendation software, we should understand if users are familiar with recommendations at all and where they find the best recommendations. This can later serve as a benchmark to measure the strength of recommendations on our platform. We could ask questions like which of the following music streaming websites is best at recommending music you like, and we can list a bunch of options?
While we would want to understand whether users listen to songs based on their mood or how they want to feel a question like do you subscribe to music therapy would not be a good question to ask. It serves the purpose; however, it is a loaded question. It is unclear. People may not understand what music therapy is and you may not get valid results. Instead, we can go for a question like sometimes people use music to help them feel a certain way like happy, sad, calm, etc. On a scale of 1 to 5, how does music affect your mood, where 1 is not at all and 5 is extremely.
We might also want to ask: Does your music preference change according to your mood? But again, this is a leading question and most users would say yes. So, we could rather ask this same question in a better way. Sometimes, people listen to different music or have different playlists based on how they feel, how often does your music preference change based on your mood, where 1 can mean not at all and 5 can mean always.
Next, we would like to focus on getting insights into the usage of some of the potential premium features, and if users might be willing to pay a premium for these features. Let’s start off with a simple feature that majority of us use, downloading songs. We might want to ask: How often do you download music from file sharing websites like ZipShare, DoReGaMa, 123 music, etc.? It can also be rephrased, as in the last 30 days, how many times have you downloaded music from a file sharing website?
You learnt how to ask questions that can help analyse people’s taste in music based on the following traits.
- Familiarity with recommendations
- Preference of streaming service (with ads or paying a monthly subscription fee)
- Usage of some potential premium features
You also learnt how to frame survey questions and how certain traits characterise different questions. You learnt about yes/no questions, leading questions and loaded questions. These three types of questions are not desirable.
Now, let us go ahead and see what kind of questions we can ask for our second key hypothesis, that is: Indian users prefer streaming music with ads rather than paying a subscription fee.
Our next hypothesis was majority of Indians with a smartphone would prefer music streaming services with ads compared to paying a monthly subscription fee.
So, probably we would want to start with do you actually own a smartphone? This question would also help us avoid coverage error. We could then ask what kind of apps do you find the most appealing? The options here could include free or low-cost apps that are supported by in-app ads, second option free or low-cost apps that are supported by in-app purchases, you see these kind of things in games or music services that charge for download. Third option here is more expensive apps that have no ads or no in-app purchases, and then the default options of no opinion.
A bad question here would be: Do you prefer ad-free music streaming apps, because the usual answer would be yes. We would also like to know if people are open to using multiple music streaming services, giving us an indication that they are open to trying new apps, like the one we are planning to develop.
One of the best practices is to end the survey with personal questions and not to begin with them. The primary reason is that they may be perceived to be sensitive, and since we have already established a connection with the users by questions that we asked earlier, it is more likely that you will get an answer and also a more precise answer. Also, it is always better to ask for ranges in these kind of questions instead of the exact numbers. Instead of providing a text box to enter age or a text box to enter the family income, post them as range questions. For example, how old are you? 17 or younger, and then divide the ranges as 18 to 27, 28 to 37, 38 to 47, 47 or older. Similarly with family income?
A few important things to note here; one, these are not exhaustive set of questions for this particular survey. They are just indicative in nature, and two, you need not frame all of these questions by yourself. Majority of questions are tried and tested and are available online. Based on the principles we have learnt, we can choose the one question that suits you the best in the given context. Majority of surveys are designed in tools that provide a ton of templates and visually appealing surveys like SurveyMonkey and SurveyGizmo. So, better to leverage them than to reinvent the wheel.
It is important to ask questions that enable users to clearly explain their position and report their situation as is. Here are some points that you can keep in mind while framing questions:
- Ask personal questions, but it is better to ask them at the end of the survey rather than at the start
- Ask for answers in terms of ranges for these questions instead of exact numbers.
- Ask questions that will help you segment users into cohorts and define behaviour based on these cohorts.
To create a survey form, you can use tools such as SurveyMonkey and Alchemer (previously SurveyGizmo).
For the music streaming app survey, you can use this survey questionnaire. This survey questionnaire contains the ‘good questions’ that Saketh was referring to in the video. You can also use this template to create other surveys as per your requirements.