Chapter 2: Research design
In this section I go through the method used during this research project to answer the research question “How might we enable startups to build their culture?”. I touch on subjects like my scientific approach, implementation, analysis, diversity, and ethics, along with a reflection on the chosen research design.
Scientific approach
Easterby-Smith, Thorpe and Jackson (2015) mentions four distinct, but closely related, main headlines for research design. The first two are ontology, how I as a researcher view the world and the assumptions I make, and epistemology, the assumptions I make about the best way of investigating the world and reality. In this project I’ve used an interpretivistic approach, which leans toward theory-building and facilitates the understanding of how and why as well as allowing for contextual factors and complexity (Raddon, n.d.). This aligns with the research questions that aims to explore and come up with a theory of how startups can build their culture. From an epistemology standpoint I needed to understand another person’s worldview (Anderson, 2013) in order to be able to interpret their answers and analyze it in relation to data I gained from other people in the research process, leading to me choosing to take a qualitative approach to my research.
During the project I’ve mainly used a relativistic school of ontology, meaning I believed there to be many truths (Easterby-Smith, Thorpe and Jackson, 2015). This is why I interviewed people from various viewpoints and fields to gain a better understanding of the different contextual factors that were at play in the realm of the research question. This viewpoint is also evident in the designed intervention, which is built on the insight that every intervention has to be tailor-made to the organization that’s applying it to some extent. Depending on what the startup need or the values the people in it carries with them different exercises or even starting points could be used, essentially reconfirming the relativistic construct of reality.
However, this research project also has a pragmatic blanket wrapped around it due to the feature of an actual intervention, where the paradigm focuses on design-based research and that reality is a practical adoption of an idea or ideas (Anderson, 2013). Drawing from my learnings from Hyper Island and the design thinking approach as a way to solve problems, the pragmatic paradigm has been evident in the design of this research project.
Double diamond
One of the frameworks used for design-based research in the industry is the Double Diamond, which is developed by the British Design Council (Design Council, n.d.). This project has been heavily influenced by the same framework. This is a quick overview of how the different phases relate to my research.
Phase 1 — Discover
I started the research project by interviewing people first, meaning there was no theory my questions were based on. Instead I aimed to discover the truths behind each individual I was meeting and staying true to the discover-part of this phase.
Phase 2 — Define
I transcribed all the interviews and went through them while writing keywords on post-its that were put on a wall for clustering themes.
Phase 3 — Developing
Using the data clusters as a starting point I defined my own theory by connecting them to academic research and industry sources and developed insights. From these I designed a possible process framework that startups could use to build their culture.
Phase 4 — Deliver
The process framework was sent out to people in the industry, both people I’ve interviewed and people I believe would be interested in this type of work, for feedback which was used to iterate the process and provide a guide moving forward.
Data collection
Due to the interpretivistic, relativistic and qualitative research approach underpinning this study in-depth interviews have been used as a method of the primary data collection. As Kvale (1996) mentions, the main task in interviewing is to understand the meaning behind the interviewees words as well as covering facts. By interviewing people from the industry I could therefore get access to their best practices and their own values, as well as trying to get behind the deeper roots of those practices and values. Because of this I wanted to adopt a very conversational approach to the interviews.
The interviews were unstructured in nature but leaned towards semi-structured, meaning I was focused on building rapport with the respondents and having a topic that guided the conversation instead of specific questions I wanted answered. This allowed me to, as Cohen and Crabtree (2006) mentions, follow the respondents’ particular interests and to test out my own preliminary understanding while still being open for new perspectives. All the respondents knew about my research question beforehand and had that as a hook to hang their expectations of the conversation. I did however develop interview guides for each respondent to prepare myself for the interviews and to open myself up to each of their specific interests. As Cohen and Crabtree writes, this semi-structured approach is best used when you only get one chance to interview a respondent and it also provides some sort of reliability to a study. Because of my more unstructured approach and the fact that I was exploring my research question from a wide range of different professional angles reliability is something that this research project struggles to uphold, along with many other qualitative studies.
Implementation of interviews
A majority of the interviews were done face to face which allows for the reading of social clues like body language, but as Opdenakker (2006) mentions because of the synchronized nature of this type of conversation the interviewer can tend to focus on the next question to ask instead of practicing active listening. With this in mind I often used an interview trick from my days as a journalist and remained quiet when the person I interviewed had answered the question. This allowed for me to pause and think about a next question in response to what they’ve just said, but it also opened up for the respondent to add more if they feel the need for it. This technique was also used, but to a lesser extent, when I had a Skype- or telephone-interview. But due to the nature of lag and not being able to see the person I was talking to these interviews were more straight to the point.
Out of the 15 interviews only 1 was conducted via Skype and 1 via telephone due to geographical distance. It’s great that technical tools like this exist that allows people to meet and share their viewpoints from across the world, but there was less rapport-building in these interviews especially the telephone interview which was in nature much more structured then the other one’s conducted. This could be because I’m not that comfortable speaking on the phone, even with people I know, and felt that I needed a “script” to stay on track. It might have influenced the way I approached the interview which in turn might have influenced the answers I received. However, since this was more structured it was also more on point regarding the specific area of this research project which might have made it more focused and therefore receiving more relevant answers.
The face to face interviews were conducted either at a co-working space or at a coffee shop, meaning there were other people around that could serve as distraction (as well as music and traffic outside). I chose this because of the laidback environment and the fact that a lot of meetings in New York are done in coffee shops. I also felt that this context provided a more leveled playing field where the respondent and I could meet as equals
The interviews were recorded and after I transcribed them I did writeups of the interviews that were sent as drafts to all respondents except for the first two. Consent for both recording and the writeups were done orally, with recording before the interview officially started and the writeup after the interview was finished in order not to put the respondent in the mindset of having to answer “perfect” or “smart” in the beginning of the interview. The respondents had the option to add comments/notes to the drafts of the writeups if they felt that they wanted to clarify or delete something. The majority of the time it was either a clarification or a typo that was edited. The data that is described on this website comes straight from the raw data. At the same time the writeups provide a very close correlation of what was said during the actual interview and how the conversation was structured. After the edits were done the writeups were published on Medium to spread the knowledge in the industry.
Sampling
In total 15 interviews were conducted using two different sampling methods. In the beginning I used purposive sampling, focusing on particular characteristics and expertise that were of interest and would enable me to answer my research question (Lærd Dissertation, n.d.a). Since I wanted to explore how we might enable startups to build their culture I focused on three groups/characteristics that would provide different perspectives and expertise; Startup founders, Accelerators and Organization Consultants. I did research on who were working in the realm of culture building and the future of work, along with accelerators and startups in the New York startup scene, and contacted people via e-mail or Twitter. After interviewing some people I asked them to refer me to people who they thought would be of interest for me to talk to, meaning I started to use a snowball sampling. Mainly this was used for getting in touch with startup founders, since I had problems getting contact with them. Lærd Dissertation (n.d.b) mentions that this sampling method is good for getting in touch with people who are hard to reach and by getting an intro to other founders from a person I already interviewed I was able to get in touch with more people and provide more depth and data to my research.
Because of the nature of both sampling methods being non-probability ones and therefore up to the judgment of me, the researcher, it could be argued that there is difficulties in reaching generalizability with this study. As mentioned at Lærd Dissertation (n.d.a), this method isn’t designed to reduce bias as much as a probability sampling technique. However, because of my relativistic ontology this study doesn’t claim to be able to generalize. Instead, it provides another truth to the many other truths that are out there. Because of my bias and unique experiences as a human being I’m also interpreting the data through my filters, which means another person might interpret the data differently. By bringing in the pragmatic paradigm, which focuses on providing practical tools or techniques, and focusing on iterating the process I’ve developed from my interpretation of the data during this project one could argue that what matters is that the process (intervention) works for the people who are using it. This study is therefore more focused on the transferability, that invites the readers of this project to develop their own connection between the study and their own reality (Writing@CSU, n.d.).
Diversity
For me it was important that this project was diverse, due to my personal values and beliefs in equality. The startup world is also predominantly male. Because of this I made a conscious effort to make it gender balanced and reached out to more women. In the end the respondents were 47% female and 53% male, meaning I reached my goal of gender balance. In future research it would be of interest to the study to include more perspectives from minorities, especially since the startups scene is very white male-dominated.
Another factor I kept track of was connections to Hyper Island. Since Hyper Island works in the realm of the study and since they tie a lot of experts to them. A connection to Hyper Island means they’ve worked for them or have enrolled in any of their programs. 60% of the respondents have a connection to the institution, which means it could have effected the outcome of the study. This is because Hyper Island is very process-oriented and focused on using reflection as a tool, something that has come up a lot during my interviews. This could put the internal validity of this study at risk. However, as stated by Trochim (2006) internal validity is only relevant in studies that try to establish a cause and effect-relationship which this study is not trying to prove. The theme of reflection etc. was also evident for the people who did not have a connection to Hyper Island. Reflecting back on the project I would’ve focused a bit more on getting respondents that didn’t have a connection to Hyper Island to increase the external validity and possibly reduce the bias. However, as mentioned earlier, the generalizability for this study is limited due to the ontological approach.
Analysis
After transcribing the interviews I went through them in batches, doing 5 interviews at a time, and used a method from IDEO called Download Your Learnings (IDEO, n.d.a), a method designed to be used in a team which meant I had to adapt it to fit my individual work. As mentioned I went through each transcription and captured information on post-its in the form of keywords. I organized the post-its by respondent at first until I’d captured all of the data. After that I used the method Find Themes, also from IDEO (n.d.b), and clustered keywords together as I saw relations between them. This partly mirrors the inductive approach of a qualitative data analysis, which is looking for emergent frameworks through relationships (Nigatu, 2012) and using content analysis to categorize and highlight important findings (University of Surrey, n.d.). The four clusters that emerged are chapters 5–8.
Ethics
Dara Blumenthal says that ethics is fundamentally about how people interact with each other and that there are small ethical building blocks, like being present, practicing active listening and treating your respondent as a human being. These are the blocks that have been used during the data collection to build rapport with the respondents and to ensure that I’ve met and interviewed them in an ethical way.
Informed consent was collected in writing, via email, to participate in the study and orally for recording the conversation and to gain consent to publish writeups on Medium. The recordings have been stored on an iPhone, a computer, and an account on Dropbox, all of which are password-protected. As mentioned by Houghton et al. (2010) informed process consent allows for negotiation and revision of the arrangement throughout the course of study. By first checking via email, then again during the interview and then through sending the respondents a draft of the writeup of the specific conversation that they could comment on consent has been collected throughout the process.
In the case of Dante Federici’s interview — it was ad hoc and his boss Sanjay Venkat, who I was originally there to interview, suggested I’d interview someone more from the team to gain more perspectives. Dante came out to the lounge area at their co-working space and agreed to do an interview then and there. He might’ve felt coerced to say yes since it was his boss asking, and to decide on the spot. Before I began interviewing Dante Sanjay went back to work in a different room, and was therefore not present. Because of my decision to publish the interviews I made here on Medium I had the opportunity to write a draft and send to Dante, making sure he was okay with what he said and that he could stand for it, which he was. However, I’ve decided to incorporate very few quotes from him (all in all there are three quotes from him in the following text) because of the circumstances surrounding the interview.
This study has been approved by the the ethics committee at Teesside University and follow Teesside University’s principles of ethical conduct.
This study has been conducted independently and there are no conflicts of interest or partiality to be disclosed.
Reflection on method
Reflecting back on my method choices I believe I’ve chosen the right method to answer the research question. What I would do differently is to make the respondents more diverse, for example bringing in more people of color, in order to gain more perspectives and therefore build a more robust theory that the intervention is designed on. I would’ve also been more iterative during the process, writing as I went along instead of taking two weeks off to write my whole theory. This might’ve opened up new perspectives and ideas that might’ve added more rigor to this project. I would’ve also liked the method to be more participatory, adding workshops that would’ve informed the development of the framework more.
Due to issues with time management I did not have the opportunity to properly test out the process with a startup, which is something that would’ve added more rigor to the intervention and the research design, which is built on the Double Diamond. If I‘d have had a workshop, as mentioned above, I would’ve been able to collect feedback through them as well. However, I believe the prototype of the framework is sufficient to move forward and test it further with actual startups after this project has been turned in and there are no time constraints to consider.
Next: Chapter 3: The interviews
Previous: Chapter 1: Introduction
If you’d like to get in touch, you can find me on Twitter.
Hyper Island — MA Digital Media Management
Industry Research Project
