The early voice-of-customer research steps described in parts 10 through 12 will yield a tremendous rich set of data that will provide insights as to what true jobs customers are trying to get done, their desired outcomes and their definition of how satisfied they are in achieving their desired outcomes. The challenge of course is mining and transforming the raw data into market insights we can use to develop solutions for important jobs to be done.
The reason we conduct “formal” voice of customer research versus random conversations with potential market participants is to allow us to structure the data so we can compare apples to apples and develop basic trend lines to focus our innovation efforts on products that appeal to a community of customers versus one-off highly customized products.
Recall from part 10 (Testing the Key Assumptions with VoC) we created a quantitative component of the research based on rating scales associated with key questions. The rating scale is a modified Likert scale with a numerical range of 1-10 and descriptive anchors at each end.
The key to this kind of research is that the real value of the numerical rating lies not in the number itself but in the explanation of why the number was chosen. When the rating scale is combined with a follow-up question asking “What would it take to make it a 10?” the discussion yields in-depth information about the specific components of the score, what elements the respondent used to judge the item in the question, and the missing “wow’s.”
Additional analysis of the combined scores of all respondents helps us understand the range of answers, high and low scores, and trends. Fore example scores of 8 – 10 might be considered “high” ratings. Scores of 1 -4 might be considered “low” ratings. Again, the real value of the numbers is to act as a focal point for analyzing the verbatim transcripts of the interviews. Opinions (and quotes) associated with high scores are compared to those associated with low scores. Definite themes can be identified and used to validate and/or refine a core customer value proposition.
If you were able to record the interviews, definitely transcribe the audio files into word files. There are many good resources that provide transcription services and at a great value. Most if not all of the data mining will come from the transcripts – though you might want to listen to selected parts of the audio files to uncover nuances that might provide additional insights. If you didn’t capture the audio then you will have to rely on your copious notes.
Use Excel to organize your data. Each anchor question (quantitative questions, demographic questions, and the scripted qualitative questions) is place on an Excel sheet. Each sheet will include the all respondents inputs including their Likert scores and follow on qualitative responses to the drill down questions.
You will have to decide how many quotes (and their lengths) you want to include for each question. There will be a lot of information to sort and you might feel a bit overwhelmed as you get into the analysis. My recommendation capture all the quotes you can that reflect insights and paste them into the appropriate Excel sheet. You will do several recursions of data mining and report formatting to identify unique and insightful information that contains specific desired outcomes. We will talk more about extracting information for the qualitative research in a future blog.
To mine the data you will use both the scripted questions to search the word files as well as notes you take during the interviews. On your notes try to capture key phrases during the interview so you can do a simple text search to locate the quote in the text (if you mark the time you can also search the audio file). There will be amazing nuggets for you to discover so do a little upfront planning to help guide your mining efforts into the right spots.
Creating the quantitative charts and stats are relatively straight forward in Excel. The challenges you will have is in deciding how to “quantize” the data (i.e. deciding what range of numbers represent “high” ratings and what range of numbers represent “low” ratings). This can be especially tricky if the data set is relatively small (i.e. less than 25) – don’t sweat it though. We are just looking for trends to develop a general idea of what potential customers want and value.
In this early stage research you will get some level of validation that the “jobs-to-be-done” are important or not important as well as discovering new insights and opportunities you may not have otherwise been aware of. This is the power of the discovery driven design process – and it’s a process. You still have more work to do in mining the data and uncovering in all likelihood more questions to test.
In our final blog of the series I will talk about team Teknovantage (remember them?) next steps in the discovery driven design process and how they can leverage the information they gathered in the initial VoC research.
Until then – design well!
Kevin