BUILDING SURVEYS | BEST PRACTICE Scoring will be driven by the questionnaires themes and your choice of questions. Text boxes, however, give respondents free rein. Carrying out a survey Further reading: Analyzing Social Science Data, Sage Publications De Vaus D (2002) Surveys in Social Research, 5th edition, Routledge, De Vaus D (2002). Qualitative Research Practice, NatCen, SAGE (Second edition), Ritchie, J., et al. (2014) D R RODERIC BUNN is a consultant in Soft landings, energy analysis and occupant surveys. Success depends largely on how a survey is carried out. Raise awareness among building occupants so they welcome you with open arms rather than indifference or resistance. High survey response rates require honed communication skills. Be polite but insistent when asking people to take part. Changes to data protection legislation require respondents to opt in formally. Surveys should be carried out over a single day, in case conditions change. You need enough resources to hand out and gather up paper questionnaires. One person can feasibly survey 500 people in a day. For populations above 600 people or multiple tenancies above 300 in total budget for two surveyors. You should always aim for a 100% response rate of the people in the building on the day. Be pleased with anything above 80% and content with 60% and above. A 40% response rate will generate shaky statistics. Web-based surveys may be needed for large buildings, or those where access or security is an issue. Elsewhere a paper-based Figure 3: An example of a semantic differential bi-polar rating scale used in many occupant surveys. So-called, as the respondent is required to translate their subjective perceptions of a range of conditions into a numerical score. This enables direct translation of perceptions into statistics. Used notably in the Building Use Studies (BUS) occupant survey Still 1 2 3 4 5 6 7 Draughty Figure 4: Examples of abstract response scales using terms or icons. This is a form of attitudinal or Likert scale, with or without numerical references. Many built environment surveys favour these scales, as scaling intervals based on emotions are believed to help respondents rank their subjective feelings towards conditions Very unsatisfied Unsatisfied Neutral Satisfied Very satisfied Amenities Figure 5: A version of a semantic differential scale where a respondent is required to select their opinion on a seven-point scale between two polar adjectives (not shown). The green and red tinting implies a sliding scale between good and bad. This scale is used in the CBE survey available from Berkeley University California (certified by Well and Leed, so has the same status as BUS) survey should be the default. It may require more work, but you get to speak to the occupants directly, and you experience the building yourself. Web-based surveys tend to generate lower response rates than paper-based surveys, which is why some people devise incentives. However, any form of incentive, such as prize draws, runs the risk of introducing hidden bias in responses. It might also motivate people to cheat for example, some people might use multiple email accounts to increase their chance of winning a prize. Keep data analysis simple. There are many analytical and statistical tests suitable for survey data. However, as respondents are delivering whole integer scores on a small categorical scale (for example 1-7), attempts to transform it in any way might fatally distort it. Some surveyors indulge in complex transformations to turn non-normal distributions into normal distributions (bellshaped curves) so the data is parametric. However, in building studies, non-normal distributions describe how occupants perceive actual conditions. They are insightful distributions. Normalising them is unhelpful. The mean value of a distribution and its variance (a measure of data dispersion that may be insightful) should be calculated, and always reported in any graphs along with the sample size (Figure 1). Free-text analysis Comments are best kept in their original form rather than subjected to some form of thematic analysis, as comments can easily be re-categorised into themes chosen to suit the ambitions of a researcher. Again, this is a form of bias. In any case, thematic analysis is best for longer tracts of feedback. Word clouds are often used to display comments. They are compelling to look at, but these can be prone to error if no attempt has been made to resolve subtle differences, such as too hot and a little hot. It may be insightful to categorise freetext responses as positive or negative, with a balanced category for any comments that are neutral or only mildly critical. People are generally more motivated to complain rather than to praise. People who are happy generally dont make comment. Which brings us back to the beginning: ask people a sensible question and they tend to give you an honest answer. It may not be the one you want or the one you expected, but thats why perceptions gained from occupant surveys are so important: they tell us what we need to know, not serve to confirm the surveyors expectations. CJ 38 February 2019 www.cibsejournal.com CIBSE Feb19 pp36-38 Occupant survey JR.indd 38 25/01/2019 16:28