DIYing along with DIYers: Juggling with Scales During a Home-Improvement Research

Research Strategy

Given the diversity and the complexity of building materials, tools, methods and home-improvement practices, rushing headlong into the field would have resulted in leaving the team stuck, confused by the flood of technical terms and gestures. Even more, trying to make a Prévert-ish list of all home-improvement activities could have led the team into a trap. Indeed, such a list is impossible to define a priori and impossible to make in practice: a new activity could always be added to the list. More than that, setting such a list would have led the team to consider DIY in a strict and rigid way, potentially far from the DIYers’ own definition of DIY. The team tilted towards an extensive and phenomenological approach: investigate what is DIY from DIYers’ point of view and what makes them engage in doing something by themselves, in their home.

Building the Research Framework

To understand the logic of DYIng practices, the team had to take a step back to grasp DIYers and home-improvement in a broader perspective. The study was framed around several main aspects:

• The social anchors of DIY. The team’s goal was to study DIY in its environment and social contexts. Who are the DIYers in terms of social origins or positions, family structures, political and moral beliefs and inclinations? The hypothesis was that home-improvement are socially situated practices (Bourdieu 1979, Bonnette-Lucat 1991).
• The embedding of DIY. The team needed to study the social structures in which DIY takes places: the family organization with its different roles, division of labor, rules, issues, way of life and how it affects DIY itself. As most activities are embedded in wider social structures (Granovetter 1985), the hypothesis was that DIY practices are shaped by the social structure where it takes place.
• The learning logics of DIY. An emphasis was placed on observing how the skills were learned in practice. From an hand-objects systems perspective (Sigaud 2012), a special focus was put on how learnings are incorporated through practice (Wacquant 2015) considering the body both as a tool for learning and action.

The main research techniques were in-depth interviews with DIYers, combined with in-home observations of improvements and alterations already realized. The team stayed for about a day at our informants home, the interviews were carried out with the DIYers and their family. A formal interview was carried out then followed by informal discussions during the visit of their home. Given the subject and the mission’s constraint, the team estimated that meeting with 15 DIYers would be sufficient enough. The guidelines for conducting the interviews were to make an archeology of DIYing: starting the conversation with DIYers as a biography through the prism of DIY and let it slip to memorable past achievements, ongoing works and future projects. For each case, the researchers focused on action-in-the-making: how the DIYers did it by themselves. Some even went so far as to play the game of “recreating” some of their achievements. It helped to grasp the practice with details and to re-stimulate the memory of informants.

Secondly, a quantitative study was carried out using an online questionnaire administered to a representative sample (N = 1200) of the French population. The original intention was to measure the behaviours and properties of DIYers and then to extrapolate the data to the general population. The initial analysis was planned as part of descriptive and inferential statistics. The sampling and constitution of the panel was carried out by a specialised market research company. The questionnaire design and the data analysis were carried out by the _unknowns team. On paper, the quantitative part seemed relatively simple, as its articulation with the ethnographic study sounded logical at least initially. It was naive, it turned out to be a bit more complicated.

Having built their research protocol, the researchers were set to go. And no sooner had they set foot in field, they understood that DIY couldn’t be reduced to a predefined and limited set of activities.

DIYing, Defining what DIY Is

Starting from the DIYers point of view, the researchers quickly realized that DIYing couldn’t be defined so easily. Indeed, far from being boiled down to a predetermined set of activities, defining what DIY is a part of the activity of DIYing. Each DIYer builds its own definition of DIYing both in terms of activities (such as electricity, building work, carpentry or woodworking, etc.) and the intensity with which they are practised. This tension between versatility and specialization (Bonnette-Lucat 1991) is coupled with a tension between what can be done by yourself and what can (or should) be delegated to a professional.

As a result, there is no such thing as degree 0 of DIYing. Everybody is compelled to DIY to a certain extent. It would be shameful not to change a lightbulb or build a furniture kit by one’s self. This social constraint implies taking care of a certain number of maintenance tasks by yourself and is rooted in family and intimate contexts and issues. Indeed, home-improvement practices are directly intertwined to a household’s lifestyle and ambitions. The stakes are high: the possibility for a child to play again with an accidentally broken toy, the possibility of regaining the use of the only shower in the dwelling before a week of work. All home-improvements or alterations are made in the name of a desired way of life, couple and family agreements, and constraints of (working) life. For the researchers, the word “ethnography” took on its full meaning. From the Greek “ethnos” for family, tribe, culture, ethnographying DIYers meant studying their family culture as a whole and not DIY as a disembodied practice, floating in the air like an ectoplasm.

However, DIYing practices are fragile and constantly re-assessed through action-in-the-making. At any time, the possibility of failure can arise, which calls into question the decision to DIY.

DIYing, Coping with Doubts and Fear

Uncertainty is a major component of DIYing: at any time, anything can go wrong. The DIYers’ daily challenge is to manage these doubts and find the boldness and courage of taking action. To get a feel for the DIYers challenge, one of them, Michael a 35 y.o. engineer living in Paris recalled the story of an epic fail, best known as the shower tray incident: “The shower tray is made of solid stone, 115 kg […] we laid it, we glued it, I started to lay the tiles at the same time. And during the night, I said to myself: “Uh uh, I forgot to do the levels”. The next morning, I arrive, I pour water: it was stagnant…” In the end, Mike had it changed by a professional.

The shower tray incident underlines the importance of trials and errors. Making mistakes and learning from them is an important part of DIYing. Advanced DIYers are used to make mistakes and to find ways to overcome them. Indeed, most of the knowledge and skills are built and incorporated while practicing. Having experience, using their senses and relying on them to “see”, to “know”, and of course to “feel” while in action, is at the heart of DIY. There is no substitute for experience, not even a good handbook. And getting experience requires trying, i.e. self-confidence, a sense of authority beyond intellectual and manual skills. Thus, the ability to deal with doubt, fear and incidents is what most differentiates DIYers as well as their proclivity to engage in diverse activities and to learn by themselves.

4 DIYers Profiles Uncovered

The research team observed four different DIY profiles, based on their attitudes dealing with doubts and fear, playing around with tools, and building materials.

1. Compelled DIYers. They only tend to do small maintenance work by themselves such as minor repairs, building furniture kit, etc. Their home-improvement practices are driven by social constraints. Some even have chosen their house because no work needed to be done. Michel, a bachelor in Clermont-Ferrand said to the team: “the apartment was in this state, I didn’t redo the wallpaper, it was generally quite clean, so it was a stroke of luck! […] I do the minimum, I know how to drive a nail to put a frame, but I have no interest”. What prevent them to engage in DIYing was they don’t have enough resources (skills, sense of authority, incorporated knowledge) to take action and overcome their fear and feeling of incompetence.
2. Hobbyist DIYers. They carry out maintenance and repair tasks to comply with their obligations, but they also invest in some specialized activities as a hobby. Not all DIY activities are eligible to become a hobby, this mainly concerns the creation or repair of furniture, gardening, car repairs, and all sorts of activities that don’t impact on the family lifestyle nor the usage of an important home feature (such as the shower). In some of these activities, DIYers can even compete with professional craftspeople.
3. Self-Sufficient DIYers. They tend to great versatility and expertise. They engage in DIY activities with an ideological dimension: DIY is a valorization of autonomy, resourcefulness and individual responsibility. They value the fact that they are not dependent on anyone, sometimes to the point of challenging professionals. Mark, a 38 y.o. living with his wife and a 2 y.o. baby told the team: “I’ve laid steel-pan roofing, I’ve done roofing before. I could even do whole roofs. Honestly, […] if I had a job that would allow me to take a year off, you know, to be out of work… if I had a job that would allow me to do that, I could build my house by myself”. In addition, they do not hesitate to tackle home-improvement issues that could impact the family lifestyle.
4. Semi-professional DIYers. This profile is a continuation of the self-sufficient DIYers to the point where it may be considered turning DIY into a profession: putting themselves at the service of something other than a housing ambition, as Franck, a 43 y.o. engineer turned entrepreneur: “The apartment I buy it partly to rent it… that is the difference, it is important. Typically, I don’t have nice furniture, I bought furniture from Emmaüs [local charity] […] this summer, we removed some cables to be able to insulate the oven […] and put a specific circuit breaker for the plates, in 32A, because they were in 20A […] I brought the electricity up to standard, because that was dangerous, and in terms of… When you’re renting, you’d better… make sure everything is up to standard as much as possible”.

The purchased-patterns home-improvement retailer customer segmentation was surely challenged. But was it enough to make it shift? Seeing these results, the home-improvement retailers executives empathize with different types of DIYers and their everyday challenges, some even recognize themselves among the profiles presented. Ethnography surely demonstrated its interest to reveal the logics behind the practices, but a concern quickly arose. Could these findings be quantified? Indeed, even if these insights could fuel the design process to imagine new services, it was not enough to identify where the business opportunities were located – if it was in their interest in specifically targeting one or more of the DIYers profiles. The strategy couldn’t be defined yet, we were still in the middle of the fold. This is where the numbers come in.

Two Scales of a Different Nature

When the team started to design the questionnaire at the beginning of the quantitative phase, several problems arose. Indeed, ethnographic materials and quantitative-questionnaire data are not easy to connect. Each one has its own scale, properties and captures different aspects of reality.

As ethnography captures various kinds of traces – audio, transcription, pictures, diagrams, drawings, gestures but also field notes that reflect the lay of the land – its research unit is multifaceted and diffuse. Indeed, the ethnography grasped the complex DIYers’ experience intertwined within the symbolic and material reality of their home and family. DIYers were not alone, they had spouses or wife, children and all the family participated in home-improvement to a certain point. The ethnography captured not only the DIYers experience of DIY but a part of their tribe’s experience in situations.

On its side, quantitative-questionnaire captures standardized answers to a predefined set of questions but provides little information on context – especially when administered online. Thus, its research unit is unitary and discontinuous: a series of context-blind individuals reactions to the questionnaire. The research team was aware that an online questionnaire would only provide artifacts generated by the questionnaire itself.

Do ethnography and quantitative questionnaires belong to two different and irreducible knowledge scales? Realizing these differences, it seemed to the team that no direct and continuous link could be established between these two scales. The researchers wonder if they could really quantify their ethnographic finding with an online questionnaire. They felt they were trying to fit a square peg in a round hole. If it was the case, it would have meant that it was impossible to quantify ethnographic findings. And the team would have been in a good mess! But there is only one real world, and multiple ways of describing and understanding it. Ethnography and quantitative questionnaires can only extract traces from this real world – different kind of traces. And the diffuse nature of ethnographic traces should be approximated through a standardized questionnaire. The researchers were facing the same doubt the DIYers had to go through. They decided to give it a try and started thinking by putting themselves in their informants shoes: what would they answer if they were confronted with this questionnaire?

Designing the Questionnaire: Translating Ethnography

The behaviors observed by the ethnography could not be directly measured by a questionnaire. But a questionnaire could capture specific and revealing information about behaviors or attitudes observed in the field. In a word, the questionnaire could provide approximations and hints that would need to be interpreted and combined to make sense of its data. The researchers thus designed each question – or group of questions – as a test to provide specific information on DIY practices. And in order not to be completely off the mark, it was necessary to start from the concrete reality of DIYers, i.e. ethnography.

Thus, ethnography was a valuable resource for questionnaire design. The team focused on translating the ethnographic findings into testable hypotheses within the questionnaire. The idea was to identify items or groups of items whose responses would most differentiate DIYers – well, according to the ethnographic study. To verify the existence and relevance of the profiles identified, the team gradually designed specific tests to assess DIYers profiles, their attitudes when facing doubt, their practices or their arbitrations. To achieve phrasing relevant questions required to look for details in the ethnographic material that would have remained left out otherwise. The team needed to plunge back into the interviews to analyse in detail all the activities (electricity, plumbery, building work, carpentry and woodworking, etc.), their underlying culture, vocabulary, tasks, tools and gestures.

Getting into the DIYers’ shoes: that was the team’s approach. 1) Use the lessons of the ethnographic study to make hypotheses (how to identify this profile knowing this?), 2) translate them into sufficiently sensitive and specific tests, and 3) use the richness and depth of the ethnographic material collected to find the most precise and relevant items and formulations. The researchers felt there was no such thing as a one-best-way to design the questionnaire. They tested it themselves and had it tested by other members of the unknowns team in order to assess whether the questions were unequivocally understood and to get an idea of what the results could be. It was more a case for trials and errors than a deterministic science: they DIYed.

The questionnaire was administered to a representative sample of the French population. 1,200 responses were collected in the end. At this point, the researchers didn’t really know what to expect, they were not at the end of their surprises.

Looking for the DIYers

While filling the questionnaire, respondents were asked to report for about 40 DIY activities – of different types, issues, and levels of difficulty – whether or not they had been carried out in their dwelling and by whom (by themselves or delegated to someone else: partner, relative or professional). The underlying hypothesis came from what the ethnographic study had revealed: that Compelled DIYers only did maintenance tasks, that Hobbyists DIYers invested in a few activities without stakes, that Self-sufficient DIYers were not afraid to engage in many types of work by themselves even with high stakes. Finally, the Semi-professional DIYers tackled all types of tasks, no matter how difficult they were. Thus, each type of DIYers should have had a particular profile of response.

To cross validate their results, the researchers ran two types of clustering techniques – Kmeans and Hierarchical Classification on Principal Components (HCPC) – after a dimensionality reduction – Multiple Component Analysis (MCA). The algorithms were running, drum roll… and three groups came out on the first draft. The team felt disappointed, if not frightened. Things seemed to go wrong.

Before plunging into the intricacies of data analysis, side note: for those who aren’t familiar with Euclidean distances, minimizing inertia methods and other mathematical matters HCPC iteratively builds a hierarchical tree (the so called dendrogram) by coupling two by two the closest measures in pairs and then repeating the operation, this time coupling the closest pairs. At the end, it results in a tree whose different branches partition the data into several groups. HCPC algorithms usually suggest the smallest number of ramifications to take into account by isolating the branches with the smallest variance.

The researchers analyzed the hierarchical tree provided by the HCPC and there was no reason whatsoever to think there were four clear distinct groups. In fact it was even questionable to be able to clearly distinguish groups.

Figure 1. Hierarchical tree built by the HCPC algorithm ©Guillaume Montagu

So the algorithm suggested three groups. Even if the researchers found that disappointing, they decided to give a chance to the machine. They looked into the data in detail and the results appeared to be even worse than expected: the groupings made by the algorithm were unclear and remained difficult to interpret. The first group seemed consistent and could match the Compelled DIYers profile previously identified, but the two others seemed to be muddled and heterogenous. That’s how the researchers got a feel for the DIYers’ experience with uncertainty.

Figure 2. First hierarchical tree and cut section suggested by the algorithm that separate the tree in three groups ©Guillaume Montagu

Experiencing this dead-end, the team had to find a way to get out of this impasse. In an act of despair they decided to consider lower branches of the three. They looked at a finer level of details and considered four groups. But still, the possible interpretations of the groupings remained unclear. Without really believing in it, the team continued to descend the tree. It was only by considering nine groups that the results became clearer. The groups composed by the algorithm made more sense and, fortunately, the team found the characteristics identified in the ethnographic part.

Figure 3. The same hierarchical tree with the cut section selected by the research team, it divides the data in nine groups ©Guillaume Montagu

• Groups 1, 2, 3 & 4 looked like the profile of Compelled DIYers: they stick to maintenance tasks (with some little variations). They seemed to easily give up on DIYing or getting work done. What differentiated the groups are the kind of activities delegated to professionals – usually “big” works (group 2 & 4).
• Groups 5, 6 & 9 looked like the profile of Hobbyist DIYers: beyond maintenance tasks some activities were particularly invested in, especially gardening, auto repairing, decorating, furniture repairing/building – i.e. activities that don’t impact the family everyday lifestyle. Other “big” works were usually delegated to professionals. One interesting thing: group 5 looked more like they were spouses of Hobbyist DIYers, the activities beyond maintenance tasks were invested by their partner.
• Groups 7 & 8 looked like the profile of Self-sufficient DIYers: most activities were invested in whatever the complexity or importance, including technical activities such as electricity, plumbery, roofing or levelling. Another interesting thing: group 7 looked like they were spouses of Self-Sufficient DIYers. Most of the activities and especially the “big” and technical ones were invested by their partner.
• But no clear traces of the Semi-professional DIYers at this point… Some converging hints make the team think that they were hiding in groups 7 & 8 but without enough evidence to isolate them as a distinct group from the Self-sufficient DIYers.

Why did the algorithm fail to suggest meaningful groups? The problem relied on a tacit assumption during the questionnaire design. Designing the questionnaire led the team to normalize behaviours into comparable variables. Thus, all activities were considered equivalent and comparable on the same basis. And that’s exactly what the algorithm did: it attributed the same “weight” to all the variables – which is disputable. How can carpentry activities be compared on the same basis with plumbery or leveling ones? The variable couldn’t be “weighted” properly to be read without bias by the algorithm. Then all the suggestions and predictions based on the inertia-minimizing criterion were not completely reliable. The algorithm did well in gathering the answers that looked alike, but did not properly identify the boundaries between groups. That was a task that could only belong to a human eye.

Retracing the Tribes

For a moment the researchers forgot that they were interrogating individuals and not tribes anymore – as they were doing in the ethnographic part. Indeed, a “tribe” answering a questionnaire does not make sense but a member of a tribe does. And it was obvious that different kinds of “tribe” members would answer the questionnaire, especially when the sample had been drawn from the general population. Here again, the team was not operating on the same scale that what was identified first in the ethnographic material: DIYers profiles were DIYer-tribes profiles instead.

Seeing things from this perspective allowed the team to “see” the tribes in the data, beyond individual answers. It brought out a mosaic aspect of the reality of DIYers and their families. The DIYers tribes couldn’t be quantified “as tribes”, it was only possible to collect data from their members. Thus, the team had to reconcile this data in order to interpret it. Thus, the DIYers-tribe clusters were made on this basis and deviated from what the HCPC algorithm originally suggested.

However, the classification algorithm made it possible to highlight aspects that remained unnoticed during the ethnographic study. In particular, it was able to highlight a clear dividing line between the Compelled DIYers and the other profiles. Putting aside maintenance tasks, it became obvious and objectified that Hobbyist DIYers and Self-sufficient DIYers do more work in general in their dwelling than compelled DIYers – by themselves or by delegating it to professionals. Seeing that Hobbyist DIYers were getting significantly more work done in their homes, a new hypothesis emerged: practice begets practice, the more you DIY, the more you transform your home, by yourself or not.

But confirming and consolidating the findings of the ethnographic study was not the only goal of the quantitative part. Since the analysis was performed on a representative sample of the population, it was possible to infer the likely shares in the general population and thus the market size represented by each profile.