The Chosen Construct
The construct feedback was decided as the most viable choice for the intervention. The intention was to increase low carbon behavioural performance by ‘nudging’ individuals to act on a series of existing prompts in order to increase the speed of the stop/start procedure. It was decided that the central hub was the ideal location for the pilot study.
Owing to budgetary limitations, this intervention proved to be the only viable option and was achievable as it would build on an existing system and simplify an extremely complex task. However it was made mandatory that every shift be given the opportunity to take part. Whilst it would have been preferred to establish a control group, the team leaders stated it could bring unrest between shifts that need to be in complete symbiosis with one another. Any detrimental effect in one shift can affect all shifts and due to the extreme economies of scale with regards to energy usage at the plant, any slight change could have a costly negative effect.
An energy saving intervention derived and created from identified behavioural constructs will result in lower utility costs. Based on the evidence within the literature review and the interpretation of the results from the thematic analysis and semantic questionnaires, the introduction of a feedback dashboard will significantly lower energy use at the plant.
The 55 participants all worked within or were connected to the open plan area of the central hub. They consisted of shift managers, group leaders, team leaders, logistics, schedulers and other operatives. Three shifts operated from this section over a 24hour period.
The dashboard was provided and designed by the target company. Prior to the intervention the dashboard was available by logging into a PC however this system was seldom, if ever used by the employees. In addition the researcher discovered a switched off 40’ LCD TV placed in an obscure location at a material delivery point, when the researcher switched this on, it displayed a version of the dashboard below. The intervention materials required to display this dashboard live within the central hub involved the use of a 50” plasma TV screen, a computer (Revo PC) and Ethernet connection. The dashboard was mounted in a central location which provided easy visual access to every operative who worked within the open plan offices of the central hub.
Figure 14 – The Feedback Dashboard
Design & Implementation
The above dashboard works in the following manner. The 10 process lines are all monitored via numerous energy meters placed at key locations along the process line. That data is fed into a central processing unit and then displayed as a dashboard in visual form. The green running lady indicates a line is running and everything is good. If a line stops for whatever reason, the person turns and holds up a hand indicating a stop. If the person is indicating a stop and is red then the line has stopped and components are drawing power. The viewer then picks up the schedule of works – stop start system of prompts and follows the prompts to power down components. If the person is indicating a stop on the dashboard and is green, then this means enough items have been turned off and the line is at an acceptable level of usage.
The feedback dashboard was installed and operated live non-stop for 8 weeks. The effects of the monitor were analysed not by retesting the constructs and disrupting the workforce, but through examining the weekly utility consumption and comparing that to past utility data. If the performance dashboard has an effect then savings in energy should be made. In addition for this period no new technology was added that could affect the energy use of the plant.
Post Intervention Results
The regression graph titled ‘Energy Conversion Costs (?/Tpiw) over last 4 years’ was generated by the targeted workplace energy team in order to assess if the intervention has been successful. The graph demonstrates that savings occurred. The energy management team reported that two weekly meter reads were the lowest ever achieved for throughput at the plant. In addition in was the first time that savings were achieved consecutively i.e. both months back to back.
This graph contains the energy data from 2010 to 2014 used to create the product at this plant. The FC 10/11 dots indicate how much energy was used in the year 2010 to create the product. The X axis shows how much product was made in tonnes per 1000. The Y axis shows how much energy, expressed in British Pounds, was needed per tonne to produce the product. The FC10/11 line, is a fit line. The 11/12 & 12/13 fit lines were left off to simplify the interpretation of the graph. However their location is slightly under the FC10/11 for the 11/12 fit line and slightly above the FC13/14 for the 12/13 fit line.
The two fit lines of interest are the FC13/14 and the FC14/15. The FC13/14 is the energy used in the year 2013 – 2014 up until the point of the implementation of the intervention. This fit line is referred to by the company as the tracking utility standard. The FC14/15 fit line is generated from the 8 weekly meter reads during the intervention. The large black squares are the weekly meter reads that create this fit line.
The FC14/15 intervention fit line indicates that energy usage for this period is significantly under the FC13/14 target utility standard and that savings have been achieved.
In addition the shape of the line is different. As throughput increases there is more opportunity to activate the stop/start procedure and therefore more opportunity to engage in energy saving behaviour. If the outlier at 6.2KT was closer to the FC13/14 line then the new FC14/15 fit line would initially sit on top or close to the FC13/14 standard. This is an important factor, because the normal fit line pattern over the years due to technology improvements has produced the same curve fit, the change in curve shape indicates that behaviour as opposed to technology is probably driving the utility savings.
In order to provide some clarity of the effect of the intervention, a CUSUM chart was created. A CUSUM is a sequential analysis technique developed by E. S. Page with the purpose of monitoring change by calculation of a cumulative sum. The 8 weekly meter readings were used to create the CUSUM by adding how much in British Pounds was actually saved or lost over the intervention period. The X axis is the meter reads. The Y axis is the win or loss expressed in British pounds in comparison to the FC13/14 fit line.
Figure 14 CUSUM Graph
The CUSUM graph above indicates that approximately for the 8 week period that a saving of ?84,000 was achieved when compared to the FC13/14 energy target tracker. If these savings are sustained for a 12 month period then a saving of ?546,000 will occur (6.5 X ?84,000). The utility data and the CUSUM indicate that the intervention has been successful. This was also confirmed by the energy team who stated there had been no changes in product or significant technology upgrades. One energy manager confirmed that within this intervention period that two of the lowest meter readings for production had been achieved and for the first time in its history the plant recorded two consecutively monthly savings. However one possible confounding variable is outside temperature i.e. degree days. The energy team explained this will have an effect but the affect is not large enough to negate the findings. Three months after the intervention, the energy team are still reporting improved savings indicating degree days are not having a negative impact.
Validity refers to the level of knowing that what a researcher believes is being measured is actually being measured. The different types of validity fall into two categories internal and external.
The measuring tools for this study did not rely on one method to identified constructs and ascertain how active the constructs were within a targeted environment. Regarding internal validity, the method used a range of interviews, questionnaires and open-ended questions to obtain this knowledge. A similar pattern or analysis was derived from the thematic analysis, percentile data and multiple regressions indicating a level of face validity. The utility data provided some external validity as the change in PBC via an intervention was measured in real time in a real working plant. As this was expected, then predictive validity increases, as operatives need the tools to achieve the set goal. Overall this allows for some generalisation to other populations involved in the same manufacturing procedure but it does not necessarily translate into generalisations to other industries. However the TPB has been shown to be effective in many working environments.
There were no technology changes during the intervention that could affect the energy usage of the plant. Likewise there were no product changes as the company makes the same thing repeatedly. Therefore it is probable that the cause precedes the effect in this situation. The introduction of the smart dashboard reduced the time operatives took to activate the stop start system of shut down prompts resulting in saving energy. However without retesting the model after the intervention or having data to show the increase reaction time it is only possible to claim a covariation effect as oppose to temporal precedence.
Reliability refers to how consistent is the observed measure. In order to thoroughly test reliability then the study needs to be replicated at another plant. The results of these findings were presented to the company’s board, energy team, EU energy team and various plants in person and live via WebEx (Web & Video Conference). The outcome of this presentation was the offer by the board to repeat the study at other plants. In essence, this can be interpreted as a measure of external reliability because the most knowledgeable individuals within this company made comparisons with this intervention and other interventions. By proxy this created a level of inter-rater reliability (Appendix E for presentation and notes).
Communications with the plant have been maintained and savings are still being reported at the plant and have been sustained and improved on for the 12 weeks after the intervention research deadline. There is no indication of the utility usage returning to baseline at this present time (08/01/2015).
Field work issues
Whilst all effort has been used to increase validity and reliability to acceptable levels, the fact of the matter is that this is not a laboratory based experiment. As a result there are trade-offs to be made requiring a degree of psychological bricolage to achieve the desired outcome. There are a mass of problems to overcome. For example the N, the number of participants was low. However despite the company having numerous employees, not all employees have influence over the energy usage. In this particular environment 55 operatives control the ?23,000,000 energy bill. Therefore they are the prime candidates for saving energy and including others would be of little value. This in turn presents challenges when using questionnaires as a Cronbach Alpha / Factor analysis would not produce the desired results due to the low N. This was catered for with the use of judges as reported earlier. Similarly the low N is not ideal for conducting multiple regression, this is why the thematic analysis was also created so comparisons could be made between the quantitative and qualitative data. If they produced similar results then this could cater for the low N. For example if the regression showed social norms to be low, then this should also be present in the thematic analysis, which it was.
There are positives to these trade-offs as high internal validity i.e. random selection, random assignment, control group etc. can limit the generalisability / external validity of the findings. These validity factors will not exist when the study is used in the real world. This is of critical importance as research on saving carbon / energy in the workplace needs to have some form of scalability, practical value and achieve real-life results for the benefit of all.
There are a range of issues and theoretical questions that need to be raised and considered between psychology conducted in strict academic settings and psychology conducted in working environments, which simply cannot be covered within this thesis.
The study examined if an intervention derived from Theory of Planned Behaviour combined with psychology knowledge could reduce energy consumption in a metallurgy plant. The results indicate that TPB plus added constructs can be an effective system for developing energy saving interventions. The results clearly showed that employees saved more energy during the intervention period than at any time in the 4yrs prior to the intervention. If sustained, the intervention resulted in energy savings circa ?500,000 pa. The intervention worked by increasing the group’s perceived behavioural control via feedback. This enabled group members to use the monitor to inform them when ‘action’ was needed. The schedule of works / system of prompts enabled them to act out the required behaviours. In essence, any group member could view the live dashboard and observe a line is down and drawing power due to the red lady indicating a stop. Then they have the choice to inform other members about the stop or take action themselves. In order to take action they pick up the schedule of works mounted on the office wall and follow the instructions to contact engineers to shut certain items off. The schedule is an already established and familiar system to the group members, so no new learning was required to activate the behaviour. The dashboard provided the missing trigger which in turn increased PBC and influenced behaviour.
In this particular study the increase in PBC can only be inferred as it was measured via savings in energy as opposed to measuring the constructs after the intervention. However Siero (1996) argues that within metallurgy plants employees who control energy work in predominantly small groups, and therefore talk more to each other with regards to energy. Similarly comparisons and competition may have occurred between employees who were responsible for certain productions lines. This could result in begin peer pressure i.e. an operative who is not responsible for a line that is down could notify another operative that his line is down. This could lead to a conclusion that perhaps subjective norms increased, resulting in savings. However this would be a mistake, if they did exist they would be a contributory factor as opposed to a confounding variable but in order to act on them an operative would need to have the belief and means to do so i.e. PBC & actual control. The research also indicates that behaviour can change without changes in attitude. However the operative’s attitude towards energy were measured as part of the model and shown to be exceptionally positive. Perhaps attitude did play a role by increasing operatives ‘buy-in’ for the intervention?
Whilst the TPB was used as a developmental tool to create the intervention, the same process could be developed by simply understanding the taxonomy of constructs identified by existing academic work in psychology. Behaviour itself can be broken down in this fashion as shown by many of the available meta-analysis of which constructs affect energy saving behaviour in given environments. A research could take measurements of these constructs and make a decisions on what construct to positively or negatively influence for the targeted environment. This notion brings into question the idea of a model. Perhaps models have more use for those who are not familiar with the taxonomy of constructs, and can be used to simplify the behaviour change process and achieve results over a short timeframe due to limited time to study behaviour at this level of reductionism.
TPB with the use of added constructs was used as a framework to ‘develop’ an intervention, as opposed to using TPB as ‘predictive’ tool to reduce energy consumption. This methodology appeared logical for field work and was well received by the target company. However this methodology required balancing act between the scientific method and practical application. Thus creating a form of psychological bricolage to achieve a working model to produce field based results. It can be concluded that this methodology based on TPB plus added constructs identified in the literature review significantly reduced energy consumption through behavioural means at this workplace. It is critical to measure utility data prior to and after behaviour change programs as the results are then truly judged in real life settings. The goal is to save energy and save carbon not theoretically but actually, by specifying the elements that make up behaviour in quantifiable terms will one should be able to effectively change behaviour via intervention.