Discussion

Our results show that prompts in the form of outright asking customers if they want a reusable cup was the most influential intervention to increase the use of reusable cups. This is not surprising, as it makes the option very clear for the customer, and among the nudges tested, this is the one closest to a default nudge, which is often regarded as the most impactful form of nudges (Mertens et al., 2022). However, many representatives of coffee shops have also expressed a reluctance towards disturbing customers with questions. Staff is already instructed to ask customers whether they want the beverage to go or to sit in, the size of the beverage and possibly additional questions. Too many questions are perceived to be a burden for customers. This claim is also substantiated in research, where studies have found that questions or reminders may backfire. For example, Damgaard & Gravert (2018) found that reminders for donations increases donations, but also the amount of unsubscriptions to newsletters and communication, as they are perceived as a burden.

Posters increasing the saliency of the reusable cups, along with various types of messages, were not by themselves sufficient in substantially affecting behaviour. This result is also highlighted in the tests at Original Coffee, where there was a substantial drop in the use of reusable cups when prompts were no longer provided, but posters remained. Visitors in coffee shops, restaurants, gas stations and similar stores are exposed to many different messages about offers to buy, or actions to take. Therefore signs, no matter how good they are, run the risk of disappearing in the overwhelming communication, or written communication is simply not sufficient to change behaviour, when other obstacles are present. We believe increasing saliency is still a good step towards enhancing the desired behaviour, but it should not be expected to influence behaviour in isolation. It may, however, facilitate the prompts, as it is easier to ask the customer about reusable cups if it is also communicated on posters visible at the right moment. It should also be noted that our interventions increased the saliency of reusable cups, but the standard option of single-use cups are still even more salient, as many single-use cups are often located in plain sight where customers can choose their beverage, while reusable cups are typically more difficult to spot for the customer. In general, there are clear opportunities for further saliency improvements in most coffee shops.

Some coffee shops offer a discount for beverages when customers choose or bring their own reusable cups. Financial incentives are usually believed to be a strong motivator for choices, but we find no significant impact of visually increasing the saliency of this discount on posters in Sweden. The lottery mechanism tested at some of the Original Coffee locations also provided a form of monetary incentive. Our results showed a strong positive effect of this lottery mechanism, but only being significant in the buying moment opposed to the return moment. It should be noted that the share of the reusable cups increased significantly in intervention period with the return moment, but that the effect is most likely to be driven by the prompt. The incentive of a chance for a reduced price when returning the cup might have been too distant from the moment customers decide between a reusable and single-use cup.

Our findings on the impact of social norms differ depending on how the social norms were communicated. The Green Nudges Playbook advocates communicating social norms to leverage behaviour change, but it is important to highlight the different possibilities with social norms. Previous research has found that simply stating how much others are engaged in a desired behaviour, can have a substantial effect on the increase of this behaviour. In one of the most influential studies on the effect of communicating social norms, Allcott (2011) showed that household energy consumption could be lowered by communicating differences in consumption with similar households. However, subsequent research has had varying success in replicating this finding. For example, Gravert & Olsson Collentine (2021) showed that communicating social norms had minimal effect on the uptake of public transport. Our results also indicate that this approach may not be enough. However, when communicating social norms more actively with a continuously updated graph and numbers on the use of reusable cups at the location where customers buy their beverages, our results show stronger effect. Whether social norms impact our behaviour also depends on which group of people we receive information about, and whether we identify ourselves within that group. Because of this, social norm nudging is also more likely to succeed in a closed setting rather than an open setting.

Coffee is everywhere, bought in all kinds of settings, from offices and gas stations to universities and sports stadiums. Therefore, it’s important to make a few notes on these different settings, because they have relevance for the applicability of nudges. As the results illustrate, using the same nudge in different settings can result in different impact. A truck driver might not react to a nudge implemented at a gas station, that works on the professor who just needs a coffee while performing a lecture. We advocate for the need to investigate one’s context to understand what might work and what might not, and not to think that an effective nudge would be effective everywhere.

Two overall settings can be identified to have an influence on the applicability of the nudges. We call them open- and closed settings. Closed settings refer to contexts where people are consuming their beverages within the surrounding area of the purchase. The customers are usually regulars and examples could be workplaces like Nordea, Universities like Nordrest or events and festivals. Open settings are the opposite. Here there are fewer regulars and people do not necessarily consume their hot beverages in the surrounding area. These two settings have a substantial impact on the applicability of nudges. As shown in the experiment, the prompt is more likely to successfully be implemented by organisations that operates in open settings. That is because employees can feel uneased or unprofessional if they are performing the prompt continuously to the same regulars, which are more frequent to happen in closes settings opposed to open settings. The social norm nudges also differ in results depending on the setting. In the closed setting of Nordea, the result was significantly higher than in the open system of Original Coffee’s at Store Kongensgade. This is logical, since social norm nudges leverages on already existing communities, and in closed settings customers are more likely to identify stronger with each other and thereby influence each other more.

The aim for the experiments is to evaluate and test if different nudges can be used to limit the use of single-use cups by getting people to choose a reusable alternative. These alternative options can vary in both material and how they are systemized. In the Swedish tests, Panter was used as the reusable alternative which is like the Kleen Hub cup tested in Nordea, Denmark. In these scenarios, customers can for free choose the more sustainable option instead of the single-use cup. It is though only free if they return it, as they otherwise they will be fined. This system is based on the idea that a reusable alternative will be chosen at a higher frequency if the initial payment isn’t different from ordinary single-use cups. The obstacle is here another behavioural barrier, where people need to use relatively long time to sign up to the system. We tried testing sign-up time with Kleen Hub and it took little over 3 minutes, which in theory is a small investment to make, but in practise this can be too much of a barrier for people to opt-in.

The system used at Original Coffee operates in a different way, where customers pay a deposit fee when buying the hot beverage (5DKK), which they will get back when they return the New Loop cup. Here the initial barrier isn’t time to sign up, but the additional 5DKK that one need to pay as a deposit.

What is interesting with these two systems is that they try to overcome two different behavioural barriers for people choosing the reusable option. The results from the test show that it in practise is easier to nudge people to opt-in to a system like New Loop, since the initial time barrier simply is too inconvenient compared to the extra money. At Nordea the main driver for the effect wasn’t the Kleen Hub cup, but the option of bringing one’s own cups emphasising the sign-up barrier. Kleen Hub have post-experiment chanced their registration process to be easier for people to opt in to.

As stated, Nordea had a third option of a reusable cup, being that customers could bring their own cup and get 2DKK discount. In theory, this is a great alternative, but in practice it’s very important to notice the previous distinction between open- vs closed settings. In a closed setting, this option can work since people are regulars and maybe even have a personal space or desk, like in Nordea. This setting provides the possibility for people to store their cups close to where they buy their coffee and not necessarily have to remember to bring their cups from home every day. Here people have both the possibility of allocating their used cups after they are done drinking their coffee and have it at easy hand when needing it the next time. In the open setting, people would not have the ability to allocate their used cup nor have it at easy access in the coffee-buying moment. They would at all times need to remember their cup every time they were about to buy coffee.

The above highlights the difference between the two initial barriers for the systems, but a combined barrier also exist. It is the inconvenience of carrying a reusable cup around until you can return it, which naturally is more present in open settings opposed to closed settings. This barrier is a major implication for shifting consumer behaviour towards sustainability in open settings, which quickly becomes a broad problem to solve and therefore relevant for future research.

When conducting field experiments, one serves the practical interest by using the appropriate experience, theory, and scientific methodology. To serve the practical interest is to develop theories and insights that is applicable to the real-world, which is why we distinguish between scientific praxis and applied scientific praxis, leading in this case to the discussion between optimal and normal implementation (Hansen, in press). When implementing nudges in the real-world that incorporate a human element like instruction to a prompt, nudging can rarely be without noise.

Noise occurs for example when employees forget to prompt, if the coffee machine shuts down or if employees get sick – it’s all that cannot be controlled for when you’re testing in the real world. Optimal implementation is what happens in laboratory studies, where the curiosity of science is of focus. Conducting field experiment often results in normal implementations where noise cannot be avoided – only limited.

To minimise noise, we did daily monitory visits and two undercover visits per week at Original Coffee. Every potential noise was written down, where employees were asked if, for instance, they had performed the prompt correctly, how they felt about doing it, any reaction from customers, or if anything unusual had happened. The undercover visits were performed to double-check their answers in performing the prompt. When reading these notes, it’s clear that we are talking about a normal implementation, meaning that some of our undercover visits show that not every employee did the prompt in every transaction. It was also openly reported in the monitory visits, where employees expressed a struggle in always remembering the prompt, and that they were not comfortable in continuing to do this to regulars or tourists. During the experiment, we therefore tried to tackle these barriers with a dynamic approach by correcting the prompt a bit (not a yes or no question, like some accidentally did), stressing the importance of always asking from an experimental point of view, and making a highlighted poster at counter for the employees to be reminded to prompt.

In the tests with a more static approach, monitoring and controlling that instructions were followed in detail was kept at a minimum. Instead, we had follow-up discussions with the collaborators to gauge how instructions were followed. Indeed, noise was present during the intervention and likely to a significant amount. Instructed prompts were not provided zealously, to avoid inconveniencing the customer or the staff themselves. The reason for the static approach was to test the results of an approach that is likely to be followed when scaling up chain- or market wide. When scaling, a dynamic approach with monitoring becomes time-consuming and expensive. Managers (or consultants) may instruct staff to follow a certain practice, but ensuring full compliance is far from guaranteed. Using a more static approach for some of these tests also allows comparisons with the results from the more dynamic approach. These results are important for conclusions regarding possibilities for scaling. One should expect results similar to test results only if the methodological approach is similar at scale. If the approach is dynamic (monitoring and controlling that instructions are followed), one should expect results similar to the tests with a dynamic approach. If the approach is static (instructing prompts, but limited monitoring and control), one should expect weaker results, similar to tests with a static approach.

To sum up, the nudges can be said to have a normal implementation, both with a dynamic and static approach. However, the implementation is closer to an optimal implementation using a more dynamic approach. The fact that implementation was not optimal makes the results even better – just imagine what could have happened if we could make sure that the nudges was 100% correctly implemented each time someone needed a to-go cup.

The experiments conducted within this project give the opportunity to test in to-go settings, where a reusable alternative for the single-used cup already had been implemented but with no great success. As the Playbook tries to assist, the goal is to get more people to freely choose a reusable cup with the overall intention to reduce plastic waste and environmental pollution. It’s hard to argue against this objective. However, it's crucial for regulators and decision-makers to consider potential unintended consequences that can arise during such changes. In the current case of this project, the objective is to reduce plastic waste and environmental pollution by encouraging more customers to choose reusable cups and using nudging as a tool for this. This is also the main objective of the playbook with no success criteria of the cup being returned, leaving the return of cups more as a sidenote.

To solve a complex problem effectively, we must consider all critical elements of the system in which the problem operates. The potential of using nudging as a policy tool becomes much greater if we take all critical elements into account when we use it. The challenge then becomes how to incentivize people to reuse these products consistently and to actually reuse the reusable products.

Before the experiment at Original Coffee, we wanted to tackle this issue and decided to track return data as well. We also placed donation boxes for reusable cups in each shop, providing customers with an easy way to return cups for those who didn't care about the discount. Additionally, with the 2B nudge (Bredgade) we wanted to see if offering rewards upon cup return would increase return rates. The add-on nudges of 2B and 2C were designed to see if a nudge effectively could include making customers both buy and return the reusable cups.

The data showed no immediate change in return rates during the intervention, but a higher return rate in the subsequent weeks at Bredgade compared to other locations. This suggests that customers may have been motivated to opt in initially but either forgot about or were unaware of the expiration of the return discount. The data also revealed that only 51% of reusable cups were returned between all shops during the experiment. With only about half of reusable cups returned over three months, it's clear that successfully changing consumers preferred choice in a to-go-coffee setting – getting them to reuse the reusable products – requires addressing the ‘entire’ desired behaviour, including considerations of return as well.

A final note to mention to the return cup discussion, is that the share of reusable cups sold is also affected by how convenient it is to return the cups. Turning to the Fogg-behavioural-model (2009), a behaviour is determined by a cue (like a prompt), the ability to perform it and one´s motivation behind it. As stated, the designs of reusable systems differ, and may require varying levels of motivations for customers to select the reusable cups. The Kleen Hub and Panter option require time for sign-up, and you need pay 5 DKK extra for the New Loop cup. The results show that by increasing the motivation, e.g. with the of rolling a die (Istedgade) or contributing to the community (Nordea), there is an increase in the share of reusable cups. However, none of the nudges are created to influence the ability of the behaviour, meaning in this case the return of the reusable cups. When deciding between a reusable- and single-use cup it’s natural to think of how convenient it will be to get rid of. The ability factor is here much lower for reusable cups compared to single-use cups, and thereby a higher motivation is needed to choose the reusable cup. This argues for the need to test interventions aiming at increasing the ability factor for customers and making it easier for customers to return the cups.

When discussing normal vs optimal implementation, dynamic vs static approaches, and the issue of cup return, the importance of context and setting becomes evident. We therefore wish to emphasise the importance of this to all policymakers and private agents who want to apply nudging to help citizens and customers make more sustainable choices. When using nudges and behavioural insights in practice, one needs to consider the following perspectives before applying strategies blindly.

Strategies such as prompt or social norms are overall broad categories which in practice can be conceptualised in many different ways with equally different effects. The nudges tested in this report present substantial differences in their effect even though some of them are similar and fall under the same strategy, such as prompt or social norms. Even within the same test setting using the same strategy of reward, we see different effects based only on a differentiation of how the strategy is conceptualised – the lottery mechanisms. Therefore, strategies cannot be implemented as plug-and-play solutions for those seeking to apply behavioural insights to encourage the use of sustainable coffee cups.

To identify what strategy to use and how to conceptualise them, one needs an examination of the issue relative to the specific context. Using a behavioural model and undergoing an extensive behaviour project might not be possible nor feasible for the layman. However, an examination of the behavioural context and setting should at minimum precede any nudge development. The goal is here to objectively identify different behavioural barriers and to prioritize their importance.

Both in theory and in practice the diagnostic approach is of most importance when working within the behavioural science paradigm (Hansen, 2019). One needs to develop a nudge-solution based on overcoming the barriers that exists in the specific context, which is why the identification of the barriers is important. For instance, you cannot nudge a consumer to choose a sustainable cup by using social norms, if the customer does not identify with the norm-group. Nor is the nudge effective to change behaviour if the issue is a matter costumer belief about sustainable cups.

When applying the nudge-approach to create any kind of behaviour change, practitioners need also to be aware that unforeseen side-effects might happen. This furthermore stresses the importance of sufficient examination of the context, as some side effects create an outcome worse than the starting point. As so, when applying the nudge approach to influence cup choice practitioners needs to consider what side effect might follow the implementation of the nudge – such as how to get customers to reuse the reusable cups.

The results of controlled tests at a limited number of establishments are interesting in themselves, but the real social impact is possible only once solutions with positive effect are successfully scaled. Our recommendations are to focus on scaling interventions first in the same domain (reusable cups) and in the same regions (Nordic cities), before possibly expanding to other domains (reusable food containers) and other regions. These are also important forms of scaling, but careful considerations when scaling in the same domain and region is a necessary first step. Our recommendations for scaling are heavily influenced by research by John List and other researchers, and readers specifically interested in recommendations for scaling behavioural interventions (and ideas in general) are advised to read The Voltage Effect (List, 2022).

When scaling, it is important to base implementation strategies on evidence of which implementations that work. When doing so, you are creating evidence-based policies. However, it is equally, or even more important to consider policy-based evidence. This is evidence that is generated specifically to be relevant for when policies are implemented at scale. When scaling, there’s a risk that interventions are not controlled as carefully as they were during pilot testing. If the implementation design and control is not the same as during pilot testing, there’s also a substantial risk that results will differ. As our results indicate that a dynamic approach leads to greater impact than a static approach, any scaling attempt should also be conducted with a more dynamic approach, to make sure that nudges are implemented as intended.

When scaling, it is also crucial to consider what can go wrong as small-scale tests are implemented much wider. First of all, it is important to have reasonable expectations of possible results. In the nudging literature, and especially outside of academia, you are often confronted with a widespread optimism of the impacts of nudges. This is partly an effect of successful implementations being dis­proportio­nate­ly showcased and used as promising examples. These successful implemen­ta­tions are not always as successful at scale, or when implemented in other settings. Similarly, scaling up in closed settings and in open settings are likely to yield different results. Circular systems for reusable cups are likely more attractive for customers in closed setting, and social norm nudges are likely more effective in these settings, as customers may know or identify more with other customers. At the same time, prompts may be perceived more as an inconvenience or burden at closed settings, since they are more likely to be repeated for each customer.

With a successful intervention, one has to understand why the intervention is successful. If any important ingredient does not scale well, attempts of scaling is likely to fail. In our tests, we have noted that interventions were most successful when managers and staff were enthusiastic about interventions. Ensuring staff enthusiasm will then be a key component when scaling interventions widely. Enthusiasm has likely influenced our tests also in the form of a form of selection bias, since certain coffee shops or chains were more enthusiastic toward this project. Coffee shops choosing to collaborate were of course more enthusiastic than those not choosing to participate, and when attempting to scale this should be kept in mind.

When scaling an intervention, it is also important to consider possible unintended consequences. For example, regardless of whether businesses instruct their staff to provide prompts to customers, or if this were to be required or somehow promoted by official agencies, possible inconveniences for customers or staff needs to be considered. As research has shown, too frequent prompts can be a burden for people, and may result in actions taken to avoid the prompts, as shown by Damsgaard & Gravert (2018) in the context of prompts for charitable giving.

Additional impacts of scaling may include various spill-over effects that may arise at scale, but not for smaller implementations. In our case, the most likely spill-over effect of upscaling may be a positive, based on additional people observing the use of reusable cups. Reusable cups are currently uncommon, while single-use cups are ubiquitous. If more and more people use reusable cups, they would naturally become more salient, and additional people making this choice would become more likely. It is also important to note that with increased use of reusable cups, the follow-up action of returning the cups also becomes crucial. Successfully nudging customers to buy coffee in reusable cups in circular systems should not be considered a success unless customers also return the cups.

For businesses that want to increase the use of reusable cups, we recommend using prompts asking customers if they want their beverage in a reusable cup. Among our interventions, this is proven to be the most impactful. As these questions are also perceived as an inconvenience, strategies where they are used for a certain initial time period may also be considered. However, one should also be aware that the positive effect is likely to disappear once the prompts are no longer used. An alternative would be that staff ask questions, but at their own impression of when this would not be burdensome. For example, it could be avoided for repeat customers who always choose single-use cups, when queues are long, or when the customer is in a hurry. This approach would not yield the same effect as when always asking but would also have less unintended consequences of being an inconvenience for customers and staff.

We also recommend increasing the salience of reusable cups in comparison to single-use cups. Information material such as posters are one step. This has been proven insufficient for behaviour change, but we still believe increasing saliency is necessary to facilitate behaviour change and also increase the chances of successfully implemented prompts. Additional steps to increase saliency can also be taken, as the single-use alternative is still the most salient, even with posters highlighting the reusable alternative. If reusable cups were made clearly visible, similar to how single-use cups are visible, the alternative would become more on top of the mind of the customers.

We also want to emphasize the need to focus not only on choosing a reusable cup, but also on returning the cups after use. Similar to the choice of using them, this has to be made as easy as possible. At university cafeterias, office coffee shops and similar locations, we also recommend installing return stations not only at the coffee shop, but also at the exit of the building. This would further facilitate the return of the cups and increase the likelihood of habit formation that increases the success of circular systems for reusable cups.

For policy development, we believe additional steps are necessary. The ordinance recently implemented in Sweden is one step towards more reusable cups, but in isolation will likely lead to no impact on behaviour change. Forcing coffee shops to use verbal prompts would likely increase the uptake, but this would be ill-advised, as there is an aversion towards more questions for customers, which may very well be justified. However, we recommend promoting the use of prompts by showcasing their potential and the importance of behaviour change.

Similarly, we also recommend promoting increased saliency of the reusable alternative compared to the single-use cups. This creates awareness that can facilitate behaviour change, even though more would be needed than only installing posters.

Another possible venue would be to focus on incentives, both for customers and coffee shops. While some coffee shops already provide monetary incentives for customers choosing a reusable cup, our studies are not enough to determine their potential impacts, since these tests focused on green nudging. Our tests revealed positive impacts from lottery mechanisms that leverage incentives, but simply increasing the saliency of price discounts was not enough to meaningfully increase the uptake of reusable cups. While price discounts may not have led to an important uptake, it may be warranted to investigate possible behavioural effects of some form of tax on the unsustainable options, or subsidy for the sustainable options. More testing would however be recommended before establishing any recommendations, as increasing saliency of discounts was only tested at one of the coffee shops in this study. Incentives are also important for coffee shop managers and staff. Currently, there are no financial incentives for promoting reusable cups, which is typically more expensive for the coffee shop.

To address multiple obstacles facing increased uptake of reusable cups, one could consider an awareness campaign led by national environmental protection agencies in collaboration with coffee shops. The focus should be on increasing awareness, enhancing the saliency of reusable cups, and providing motivation for coffee shops to increase their efforts. Campaign administrators could work with collaborating coffee shops by providing guidance and communication materials. Tracking the uptake of reusable cups should be central, and a friendly competition could be organized, providing motivation for coffee shops to become the most sustainable in terms of their share of to-go beverages sold in reusable cups. Various funding models could be considered.

It is important to consider both making the desired behaviour easy and making customers motivated to adopt the desired behaviour. Results from the Green nudges playbook show that there is already a substantial level of motivation among customers. However, making the behaviour easy should not only concern the choice of the reusable cup, but also making it easy to return the cup after use. Individual businesses can make this easier, but only to some degree, for example by allowing to return at their respective establishments and increasing the saliency of this option. For customers, it would facilitate a lot with a common system where various brands (ideally all) work together, providing the same option to customers and establishing multiple locations where returns can be made easily. What policymakers can do would be to facilitate this networking between actors, and possibly offer some incentives for the establishment or operation of such system.

Finally, we recommend additional testing on the effects of behavioural interventions, with a specific focus on the effects at scale. We recommend scaling specific nudge types (ie verbal prompts, with and without increased saliency) at many locations and over a longer time-period. We also recommend scaling different nudges at similar locations (within the same coffee shop chains), to carefully evaluate the effect of different nudges while keeping the surroundings as even as possible between test locations.

Recommended approach for scaling:

Recommendations for businesses that want to increase the use of reusable cups:

Recommendations for policy development:

Recommendations for additional tests: