Variations in Regulated Cannabis Sales Conditions and Their Effect on Purchase Behaviour: Findings from an Observational Study

Abstract

Introduction

Many countries are exploring regulations for non-medical cannabis markets to mitigate cannabis-related health risks. Regulations such as restricting opening hours or product variability may be useful tools, yet their effects on cannabis purchases remain understudied due to a lack of comprehensive sales data.

Methods

The Züri Can study implements a framework for strictly regulated cannabis sales, with a fixed product portfolio and a maximum purchase volume. The dataset contains all tracked purchases of 2,507 participants from August 22, 2023, to January 17, 2025, across ten Cannabis Social Clubs (CSCs), ten pharmacies, and the municipal Drug Information Centre Zurich (DIZ). It captures detailed information on sales conditions (opening hours, product variety, customers per hour) and purchase behaviour (purchase frequency, number of packages per purchase, THC volume, and expenditure per package). Generalised estimating equations provided means and confidence intervals by type of point of sale (POS). Multilevel survival analysis and random coefficient models examined how sales conditions influenced purchase behaviour, and how purchase behaviour evolved over time.

Results

Pharmacies had notably more weekly operating days (5.66), longer daily opening hours (6.16), and fewer customers per hour (1.08) than CSCs (3.81 days; 3.76 h; 4.62 customers per hour) and the DIZ (2.85 days; 3.28 h, 3.37 customers per hour). However, compared to the POS average, longer opening hours and more customers per hour only slightly increased purchase frequency; their impact on purchase amount was negligible. Variations in the number of available products only marginally affected purchasing behaviour. Purchase frequency was higher in the 30 days following the first purchase and declined over 12 months (CSCs, pharmacies) or remained stable (DIZ) while amount obtained per purchase remained stable.

Conclusion

These findings suggest that in a strictly regulated non-medical cannabis market similar to the framework of the Züri Can study, variations in sales conditions only have a small impact on purchase behaviour. Legal access to cannabis does not inherently lead to increased purchasing and may even contribute to a decline in purchase frequency.

Plain Language Summary

Many countries are considering new rules for legal cannabis sales. Rules about when and where people can buy cannabis may help reduce how much they buy, which could lower health risks. But we know little about how different rules affect buying behaviour over time. This article uses data from the Züri Can study in Zurich, Switzerland. There are three types of cannabis dispensaries: Cannabis Social Clubs, pharmacies, and a city-run Drug Information Centre. About 2,500 adults can choose one dispensary and buy cannabis for non-medical use. From August 2023 to January 2025, we recorded how often people bought cannabis, and how much they bought each time. We also looked at dispensary opening hours, how many customers they had per hour, and how many different products were available. Pharmacies had the longest opening hours and fewest customers per hour. Within a dispensary, changes in opening hours and number of customers led to slightly more frequent purchases, but it did not change the amount that people bought each time. The number of available products increased over time and was always similar between the dispensary types. It did not influence buying behaviour. People bought cannabis more often in the first month, but in Cannabis Social Clubs and pharmacies, they bought less over time. This suggests that changes in sales conditions have little effect on how much cannabis people buy. Over time, legal access with clear rules, such as letting people buy only from one dispensary, may lead people to buy cannabis less often.

Introduction

Despite its illegal status in many countries, an estimated 228 million people worldwide used cannabis in 2022 [1]. In Switzerland, where non-medical cannabis is illegal, approximately one-third of the population over the age of 15 report having used cannabis before [2]. Although many people who use cannabis (PWUC) do not experience negative effects, heavy cannabis use, characterised by high frequency and quantity of use, and use of high-potency (THC level) products have been linked to increased risks of adverse health outcomes, including psychosis, dependency, anxiety, and depressive symptoms [3, 4].

Given these risks, there is an interest in minimising the public and individual health consequences associated with cannabis use [5]. As a result, several countries are moving away from prohibition towards regulated markets for non-medical cannabis. Legal markets, as opposed to illegal ones, allow for transparency and control regarding some key factors such as products, pricing, availability, and points of sale (POS). Evidence suggests that such factors can significantly influence use behaviour [5–7]. Therefore, implementing regulatory policies could be effective in reducing cannabis use and related risks. On the other hand, previous research found that strict regulations – such as too limited opening hours or severe restrictions on product potency – can pose barriers to switching from the illicit to the legal market [8, 9].

Trying to establish a competitive legal market, several US states have introduced a commercial, profit-driven cannabis market [10]. Although several states implemented regulatory measures, such as limits on POS opening hours [11], many decisions were left to profit-oriented producers and retailers, with minimal restrictions on products, pricing, and advertising. This focus on commercialisation has come at the expense of public health and prevention goals [10] while also leading to increased cannabis availability, higher potency products, greater product variety, competitive pricing, and widespread advertising [12–14]. Such developments are likely to contribute to increased cannabis use and related harms [12, 15–18]. For instance, international evidence indicates that longer opening hours of alcohol dispensaries can lead to increased consumption and higher rates of alcohol-related harms [19]. Moreover, cannabis product variety in terms of form and potency, as well as advertising, appears to significantly influence use behaviour [19]. Indeed, adult cannabis use has increased in US states that have legalised it [20]. However, the extent to which individual sales conditions impact purchasing and use behaviour remains underexplored [21].

One reason for the limited scientific evidence even from jurisdictions with a legal cannabis market is the difficulty to accurately measure psychoactive substance use. PWUC often struggle to recall or accurately estimate their use or purchase behaviour [22] which makes objective sales data preferable. Sales data, however, are often difficult or expensive to obtain. Local authorities are reluctant to disclose sales figures, leaving researchers to rely on indirect methods and estimations to approximate key variables [23]. Data on individual customer purchases, such as those analysed in this article, were previously unavailable.

Evidence-based insights into PWUC purchase behaviour are especially valuable for countries considering cannabis regulation. In Europe, support for legal regulation is growing, often driven by grassroots initiatives such as Spain’s Cannabis Social Clubs (CSCs) [24]. These models promote small POS, run by individuals committed to the health and well-being of PWUC, rather than profit-driven enterprises [25]. A key challenge for policymakers is to create regulations that grant these POS enough autonomy to develop viable business models while also implementing necessary restrictions to prevent an increase in cannabis use.

To address this challenge, the Züri Can study evaluates non-medical cannabis sales across 21 small POS in Zurich, Switzerland (study protocol by Buschner et al. [26]). The POS are granted significant autonomy and responsibility, but operate within a health-focused framework with strict regulations, mandated by legislation and the study design. This controlled setting offers a unique opportunity to analyse longitudinal sales data from 2,507 participants, linking individual purchases to sales conditions. This article examines whether longer opening hours, higher customer turnover, and greater product variety lead to increased purchasing.

The study also tracks how behaviour changes over time, controlling for these sales conditions. Stable or slightly declining purchases suggest adequate access to cannabis, while increases may imply too easy availability or too cheap prices. A sharp decline could indicate a shift back to the illicit market, for instance, due to limited access or poor product quality.

Methods

Züri Can: Study Framework

In 2021, Switzerland introduced a legislative amendment permitting pilot studies on the sale of non-medical cannabis, an otherwise illegal substance [27]. The Züri Can study was approved as one of these pilot studies by the Federal Office of Public Health. Starting in August 2023, over the course of 3 years, each participant can buy cannabis from one of 10 pharmacies, 10 CSCs, or the municipal Drug Information Centre Zurich (Drogeninformationszentrum Zurich [DIZ]). The DIZ has expertise in tailored promotion of lower risk cannabis use, while pharmacies provide professional knowledge in selling quality-controlled products and offering credible health advice. In addition, CSCs offer social proximity and peer support for lower risk use [26]. All POS are located within the city of Zurich which has a highly accessible public transportation system. Therefore, each participant should have easy access to their chosen POS. As recommended by previous research, locations are at a considerable distance from schools, kindergartens, and playgrounds, and POS clustering was prevented [19]. In June 2024, one CSC closed and their participants could either leave the study or transfer to a different CSC.

Participants

Participant recruitment began in March 2023 and was conducted by the POS which could include 50–150 participants each. This limitation was defined by the study design to allow for personalised harm reduction measures while also making the project economically feasible for POS. Participants were required to be at least 18 years old, residents of the City of Zurich, and had to be proficient in German. Following legislation, people without previous use of cannabis were not eligible to participate in the study to avoid incentivising use. Therefore, participants were required to report cannabis use for at least 1 year. A THC urine test confirmed recent cannabis use. Professional drivers were excluded due to the increased risk of traffic accidents associated with regular cannabis use [28]. Medical exclusion criteria were current psychiatric hospitalisation, suicidality and history of suicide attempts, severe depression, history of psychosis or bipolar disorder, cognitive impairment, pregnancy (pregnancy test), or breastfeeding [26].

Product Variability and Pricing

Restriction of product variability was guided by both legislation and study design. In accordance with the legislation, the THC content was capped at 20%, and prices were tied to potency, with higher prices for more potent products. Participants can purchase up to 10 g of cannabis per day (total across flowers and hash) and 10 g of THC per month (50–200 g of cannabis, depending on potency).

Product pricing and selection were defined by the study design to mirror the illicit market. In Switzerland, dried cannabis flowers and resin (hash) are the most commonly used forms of cannabis, while other products like edibles are less common [29]. Two Swiss producers supplied flower and hash products with varying THC and CBD ratios (online suppl. material A, Table A.1; for all online suppl. material, see https://doi.org/10.1159/000551031). Products were sold in 5 g packages. To prevent POS from favouring sales of the pricier high-potency products, profit margins were equal across all products. Moreover, prices were standardised across all POS to avoid competition.

Data Collection

On behalf of the Federal Office of Public Health, all sales were recorded in a software. The dataset contains anonymised sales data for all participants until January 17, 2025, capturing the date and time of each purchase, product type, quantity, THC content, expenditure in Swiss Francs (CHF), and the associated POS. Sales began at 16 POS in the first week of the study (starting August 22, 2023), with three more launching later in 2023. The remaining two opened in early 2024 due to challenges in finding suitable locations.

Measures

This article uses three measures for sales conditions at the POS level and purchasing behaviour (frequency and amount) of participants (Fig. 1).

Fig. 1.

Model overview: measures for sales conditions and purchasing behaviour. CHF = Swiss Francs. This articles assesses three measures for sales conditions at the POS level. Purchase frequency and the purchase amount were assessed on the participant level. The effect of sales conditions on participants’ purchase behaviour was examined.

Sales Conditions at Points of Sale

Product Variety. Due to production difficulties, sales began with three flower and two hash products. Five flower products and two hash products became available later on (online suppl. material A, Table A.1). Although POS were required to keep all products in stock, supply shortages and delayed reordering occasionally led to temporary gaps in availability.

• Number of available products: number of distinct products that were purchased from the POS per study week.

Opening Hours. POS opening hours were determined by the individual POS without restrictions by legislation or study design. Two variables were calculated regarding opening hours:

• Weekly operating days: the number of days in a study week when participants bought cannabis from a POS.

• Daily opening hours: the average daily hours a POS was open during a study week. This was calculated by calculating the time between the first and last purchase each day (plus 5 min), averaged over the POS’s operating days that week (online suppl. Equations 1, 2, online suppl. material C).

This approach may underestimate actual opening hours but was preferred over relying on official schedules, as it better reflects the diverse operating conditions across POS. Swiss pharmacies typically open Monday to Saturday for at least 8 h. However, only staff trained for the study could sell cannabis, and they were not always available. Some CSCs operated in rented spaces with limited and sometimes changing opening hours. At the DIZ, cannabis sales were restricted to designated evening time slots booked in advance by participants.

Number of Customers per Hour. The number of customers per hour at a POS provides insight into how much time staff may have for personal interactions, such as offering safer-use advice. POS can influence customer turnover by choosing their number of participants (from 50 to 150) and their opening hours.

• Customers per hour: the average number of participants per hour purchasing cannabis at the POS during a given study week. This is calculated by dividing the total number of purchases in a study week by the total number of operating hours that week (online suppl. Equation 3, online suppl. material C).

Outcome Variables: Purchase Behaviour

Purchase behaviour was assessed using four measures:

• Duration between purchases: all cannabis purchased on the same day was counted as a single purchase. The number of days from the day after a purchase until the next purchase indicates how long the purchased cannabis lasted a person (online suppl. material D). For instance, if a person purchased cannabis on January 5, January 25, and March 16, 2025, the average duration between purchases is 35 days ([20 days + 50 days]/2) with a standard deviation of 21.21. A longer duration reflects a lower purchase frequency.

• Purchase amount:

  • –Number of packages per purchase: the average number of 5-g packages a participant obtains per purchase (maximum 2). For example, if a participant makes four purchases and buys two packages each time, their average is 2.
  • –THC per package (grams): the average amount of THC per package purchased. The THC content differs by product, ranging from 0.25 to 1.00 g per package.
  • –Expenditures per package (CHF): the average amount spent per package per purchase. Prices per package ranged from 36.00 CHF to 48.00 CHF initially, and from 37.00 CHF to 49.00 CHF after a price increase in August 2024. Higher potency products were more expensive.

Data Analysis

Drop-Outs

Züri Can participants can leave the study any time or be excluded, for example, if they no longer meet the inclusion criteria. Therefore, after a participant’s most recent recorded purchase, it is uncertain whether they will return or have dropped out. How this uncertainty is handled can significantly bias statistical outcomes. To identify likely drop-outs, the duration between purchases was analysed (online suppl. material D, online suppl. Tables D.1–D.9; online suppl. Fig. D.1). Importantly, since the duration between purchases approximately follows a log-normal distribution, the natural logarithm was used.

If the time between a participant’s last purchase and the data export date (January 17, 2025) exceeded the mean duration between their previous purchases plus twice the standard deviation, it was assumed that they would not make another purchase. In these cases, the mean duration between prior purchases replaced the actual time since their last purchase. To account for absences due to holidays or illness up to 30 days, if this mean duration was less than 3.4 (the natural logarithm of 30), the value 3.4 was used as time since their last purchase. A parametric survival analysis with exponential distribution estimated the annual drop-out rate.

Analysis of Sales Conditions and Purchase Behaviour

The data were unbalanced because participants joined the study on different days, participated for varying time periods, and made purchases at different time points. This was considered in the statistical analysis to avoid biased results.

Firstly, generalised estimating equations with exchangeable working correlations were used to calculate the mean and confidence interval (CI) of sales conditions and purchasing behaviour variables separately for CSCs, pharmacies, and the DIZ. Monthly purchase amounts per participant were estimated for each POS type using random coefficient models with a log-normal distribution based on the outcome variables.

A multilevel survival analysis was conducted separately for each POS type to examine how the duration between purchases changed with varying sales conditions relative to the respective mean of each POS. The following predictors were added to the model: one additional available product, one additional weekly operating day, one additional daily opening hour, and one additional customer per hour.

To assess changes over time, a time variable, scaled to represent 12-month periods (days since first purchase divided by 365), was added as a predictor. Moreover, a binary variable indicating whether a purchase occurred within the first 30 days after a participant’s initial purchase was added to capture whether the duration between purchases is shorter or longer during this early phase than in subsequent months. The model included a random intercept and random slopes for time (in 12-month periods) and the indicator for early purchases (initial 30 days). Since one CSC withdrew prematurely, 32 participants switched to a different CSC and had an artificial gap between their last purchase in the original CSC and the first in the new CSC (40 to 250 days, mean = 91.6 days). These data points were excluded from analysis. For each of the three purchase amount variables, random coefficient models with random intercept, random slope for time (in 12-month periods), and log-normal distribution were conducted separately for CSCs, pharmacies, and the DIZ.

Results

For the present analysis, 2,507 participants who purchased study cannabis at least once until January 17, 2025, were included (Table 1). In total, 53,806 purchases were recorded across all POS (per participant between 1 and 209). An annual drop-out rate of 20.8% (95% CI = 19.1–22.8%) was estimated.

Table 1.

Mean values and CIs of main variables per POS type

​	Pharmacies: 24,901 purchases, 1,127 participants	CSCs: 26,828 purchases, 1,278 participants	DIZ: 2,077 purchases, 102 participants
Predictor variables1
Weekly operating days	5.66 [5.64; 5.68]	3.81 [3.71; 3.90]	2.85 [2.84; 2.87]
Daily opening hours	6.16 [6.13; 6.19]	3.83 [3.76; 3.91]	3.28 [3.26; 3.29]
Customers per hour, n	1.08 [1.07; 1.08]	4.62 [4.48; 4.76]	3.37 [3.34; 3.40]
Available products, n	8.57 [8.51; 8.62]	8.81 [8.74; 8.88]	8.42 [8.24; 8.60]
Outcome variables1
Days between purchases	17.28 [16.13; 18.44]	14.06 [13.27; 14.85]	20.88 [16.17; 25.59]
Packages per purchase, n	1.64 [1.62; 1.66]	1.73 [1.71; 1.74]	1.84 [1.79; 1.88]
Volume THC per package, g	0.78 [0.77; 0.78]	0.78 [0.77; 0.79]	0.75 [0.72; 0.78]
Expenditures per package, CHF	45.37 [45.24; 45.51]	45.67 [45.56; 45.77]	44.84 [44.41; 45.28]
Estimated monthly purchase amount2
Cannabis volume over 30 days, g	14.17 [13.20; 15.14]	18.45 [17.41; 19.49]	13.18 [10.18; 16.19]
THC volume over 30 days, g	2.19 [2.03; 2.35]	2.87 [2.70; 3.04]	1.97 [1.49; 2.44]
Expenditures over 30 days, CHF	128.47 [119.58; 137.36]	168.45 [158.86; 178.03]	117.88 [90.64; 145.12]

Data are presented as mean [95% CI]. GEE, generalised estimating equation; 95% CI, 95% confidence interval.

¹Marginal means (GEE with exchangeable working correlation).

²Estimated monthly purchase amount is derived from outcome variables using random coefficient models with log-normal distribution. Data from August 22, 2023, until January 17, 2025; average days between purchases were estimated with a repeated survival model with log-normal distribution; 32 purchases were excluded because participants changed CSC afterwards.

Pharmacies had more weekly operating days (5.66) and longer daily opening hours (6.16) than CSCs (3.81 days; 3.83 h) and the DIZ (2.85 days; 3.28 h) (Table 1). Pharmacies had fewer customers per hour (1.08) than CSCs (4.62) and the DIZ (3.37). The number of available products was similar across POS types, with CSCs having a slightly higher average (8.81) than pharmacies (8.57) and the DIZ (8.42), though these differences were negligible. Participants at CSCs, on average, purchased cannabis every 14.06 days (95% CI = 13.27–14.85), which was more often than participants at pharmacies (every 17.28 days; 95% CI = 16.13–18.44) and the DIZ (every 20.88 days; 95% CI = 16.17–25.59). No significant difference was found between the DIZ and pharmacies. Participants at the DIZ on average obtained a higher number of packages per purchase (1.84) compared to participants at CSCs (1.73) and pharmacies (1.64). Volume of THC per package (0.75–0.78 g) and expenditures per package (44.84–45.67 CHF) were very similar between the three types of POS.

The estimated monthly purchase amount per person, measured in volume of cannabis, volume of THC, and monthly expenditures, was higher for participants at CSCs (cannabis: 18.45 g; THC: 2.87 g; expenditures: 168.45 CHF) than for participants at pharmacies (cannabis: 14.17 g; THC: 2.19 g; expenditures: 128.47 CHF) and at the DIZ (cannabis: 13.18 g; THC: 1.97 g; expenditures: 117.88 CHF). Between the DIZ and pharmacies, there were no notable differences in monthly purchase amount.

Changes in Duration between Purchases

Deviations from the usual average sales conditions at each POS had little effect on the duration between purchases. An additional weekly operating day reduced the duration by 2.4% at pharmacies, 5.0% at CSCs, and 11.0% at the DIZ (Table 2). An additional daily opening hour led to a decrease in duration of 2.0% at pharmacies, 2.4% at CSCs, and 2.5% at the DIZ. Similarly, an additional customer per hour decreased the duration between purchases by 8.1% at pharmacies, 2.5% at CSCs, and 4.5% at the DIZ. Having an additional available product had little or no significant effect.

Table 2.

Percentage changes in duration between purchases by type of POS

	Pharmacies, %	CSCs, %	DIZ, %
Due to an additional
Product	−1.1 [−2.0; −0.2]	−0.5 [−1.5; 0.4]	0.9 [−1.9; 3.6]
Weekly operating day	−2.4 [−4.1; −0.8]	−5.0 [−6.7; −3.4]	−11.0 [−18.5; −3.4]
Daily opening hour	−2.0 [−3.0; −0.9]	−2.4 [−3.9; −1.0]	−2.5 [−10.7; 5.8]
Customer per hour	−8.1 [−13.2; −3.1]	−2.5 [−3.6; −1.4]	−4.5 [−8.7; −0.3]
Time frame
Initial 30 days	−18.3 [−22.5; −14.1]	−12.7 [−17.1; −8.2]	−24.5 [−40.4; −8.6]
After 12 months	18.3 [9.3; 27.3]	23.9 [13.1; 34.7]	−6.2 [−32.2; 19.9]

Values are presented as the mean [95% CI]. Data from August 22, 2023, until January 17, 2025. Multilevel survival analysis with random intercept, random slopes for time, and for the first 30 days after the first purchase. Number of participants and purchases per type of POS can be found in Table 1.

95% CI, 95% confidence interval.

When controlling for sales condition variables, the duration between purchases was significantly shorter within the initial 30 days following the first purchase. This effect was strongest at the DIZ, where it was 24.5% shorter during the initial 30 days. The effect was smaller, yet noticeable, at CSCs (−12.7%) and pharmacies (−18.3%).

Twelve months after their first purchase, the duration between purchases had increased by 18.3% for participants at pharmacies and by 23.9% for those at CSCs. In contrast, there was no significant change for participants at the DIZ. However, due to the small sample size and large CI, it remains unclear whether there is a significant difference between the DIZ and the two other POS types.

Changes in Amount per Purchase

Detailed results of the mixed models analysis can be found in online supplementary material B. The effects of having an additional product, weekly operating day, opening hour, or customer per hour, on purchase amount were mostly not significant. While a few effects reached statistical significance, they were all negligible (≤2.9%). Within 12 months after their first purchase, participants had increased their expenditures per package by 2.3% to 2.6% across all types of POS, a significant but minimal effect (online suppl. Table B.3). At the DIZ, the number of packages per purchase rose slightly by 5.8% (online suppl. Table B.1). No significant change in THC per package was observed for any POS type over the first 12 months (online suppl. Table B.2).

Discussion

The Züri Can study offers a unique opportunity to examine legal cannabis sales within a regulatory framework, including a fixed product portfolio, a maximum purchase volume, and the requirement that participants purchase from a single POS. While previous studies on purchasing behaviour have relied on aggregated sales data, limiting their analyses to regional trends [17, 30], the present article used individual purchase data. Every purchase was recorded, generating the first comprehensive dataset for objectively analysing participants’ purchasing behaviour and how it is influenced by variations in sales conditions at their POS.

In Züri Can, the three types of POS, despite operating under the same regulatory framework, differed significantly in their business models and sales conditions. For instance, at pharmacies, cannabis was available to participants for more hours per week (34.9 h = weekly operating days multiplied by daily opening hours), than at CSCs (14.33 h) and the DIZ (9.3 h).

Despite this greater availability, participants at pharmacies purchased less cannabis per month on average (14.17 g) than participants at CSCs (18.45 g). A potential reason for the higher average monthly purchase volume at CSCs might be that people who buy more cannabis prefer a CSC as their POS. Notably, most previous research has used the number, density, or proximity of cannabis stores as indicators of availability, whereas our study offers a different perspective by focusing on opening times.

The longer opening times at pharmacies likely contributed to the notably lower average number of customers per hour (1.08) compared with CSCs (4.62) and the DIZ (3.37). Within each POS, adding an extra operating day, extending daily opening hours by 1 h, or having one additional customer per hour was associated with a statistically significant but small increase in how often participants made purchases, and little to no effect on how much cannabis participants obtained per purchase. Having an additional product in stock did not significantly affect purchase frequency or the amount obtained per purchase. These findings suggest that small adjustments in sales conditions are unlikely to strongly influence purchasing behaviour. Consequently, a limited degree of flexibility for POS in setting opening hours, managing customer numbers, and selecting the product portfolio may be feasible.

These results contrast with earlier studies reporting that greater availability of cannabis is linked to higher sales, use, and related harms and that product variety appears to influence use behaviour [16, 17, 19, 30, 31]. This contrast is likely due to differing focuses and policy contexts. Almost all prior studies were conducted in Canada and the USA, where few restrictions are in place and the focus lies on commercialisation. In several US states, POS offer a wide range of products, including vape oils and edibles, some exceeding 90% THC, and competitive pricing and advertising are permitted [14].

By contrast, the Züri Can study prioritises health protection and maintains this focus through several restrictions. These include limiting products to flower and hash with a maximum THC content of 20%, prohibiting advertising, and capping each POS at 150 participants. While POS may be granted some autonomy, such measures are likely necessary to prevent incentives for PWUC to try new products and increase their purchases [32, 33]. These restrictions also support harm reduction efforts by ensuring staff have sufficient time for personal interactions during each sale, allowing them to offer guidance and lower risk use suggestions when needed and appropriate. Harm reduction measures may include asking participants about their well-being and offering them access to professional medical advice, if necessary. Staff could also suggest effective filtering methods or encourage participants to vaporise cannabis instead of smoking it [28]. Furthermore, using cannabis without tobacco could be encouraged by offering low-risk alternatives with similar properties [34]. 

These contrasts underscore that our findings partly reflect the specific regulatory framework under which the Züri Can study was conducted. This should be considered when interpreting our results. While other countries may be unable to implement identical regulations due to regional differences, the results offer valuable guidance on how sales conditions can influence purchasing behaviour, particularly in contexts that prioritise public health.

As Myran and colleagues [30] demonstrated in their scoping review, outcomes are oftentimes strongly influenced by the policy context. The authors conclude that increasing market maturation and access may be associated with increases in some types of cannabis-related harms. However, in jurisdictions with no commercialisation and restricted access to retail markets, legalisation does not appear to lead to substantial short-term increases in cannabis-related harms. This aligns with our findings regarding time trends in purchasing behaviour independent of variations in sales conditions.

After gaining legal access, some PWUC initially bought less cannabis, potentially out of caution while determining whether they like the products. Nevertheless, our results show that, overall, purchases occurred more frequently during the 30 days following participants’ first purchase. This suggests that participants may, more commonly, buy more cannabis within the first month. This probably indicates that they are exploring products with different THC and CBD ratios or building a personal stock. However, 12 months after the first purchase, participants at pharmacies and CSCs had significantly reduced their purchase frequency, while frequency at the DIZ remained stable. The amount obtained per purchase remained unchanged or increased only slightly over time. This pattern suggests that access to legal cannabis does not inherently lead to greater overall purchase volumes over time. On the contrary, it may contribute to a slight decrease. Harm reduction measures implemented in the study – such as personal interactions with sales staff – may have played a role in the decline in purchase frequency.

Strengths

This article presents results from the first available dataset containing data on individual legal cannabis purchases within a controlled setting. It covers all purchases made by 2,507 participants across 21 POS over more than 16 months. While previous research often struggled to gather data on purchase quantity and had to rely on aggregated data or self-reports [22, 35], the present study analysed several metrics of purchase behaviour, providing objective and nuanced insights into purchase behaviour.

Moreover, this study is the first to directly compare three different types of POS operating under the same regulatory framework. These POS types differed notably in their business concepts and sales conditions, making the comparison particularly insightful. Despite these differences, the detailed sales data allowed for consistent estimation of the specific sales conditions across all three types.

Limitations and Future Research

Despite the extensive and reliable dataset, the specific setting of Züri Can introduces limitations. Previous research suggests that factors like pricing structure or POS location can influence cannabis purchase behaviour [6, 7, 11, 23, 36, 37]. However, these factors could not be explored in this article, for instance, because prices were fixed across POS. Future studies should prioritise investigating these factors and their influence on purchase behaviour.

Participants were informed at registration that reselling, gifting, or sharing study cannabis was illegal. Therefore, their purchase behaviour likely reflects their personal use. However, it remains unclear whether participants also purchased cannabis from the illicit market or shared study cannabis with others. Therefore, within the scope of this article, no direct conclusions about personal use can be drawn. Moreover, as tracking of purchases was anonymised, no demographic data were included in the analysed dataset. The Züri Can study collects participant information via online surveys. This will allow us to link sales data with self-reported use and give an estimation of how much of participants’ use is covered by study cannabis. Moreover, we will be able to examine differences in purchase behaviour between different demographic groups.

Conclusion

The Züri Can study evaluates a health-focused regulatory framework for legal non-medical cannabis sales through pharmacies, CSCs, and the municipal DIZ. Opening hours and customer turnover varied notably across POS types. Within each POS, an additional operating day, daily opening hour, or customer per hour had little to no significant effect on the amount obtained per purchase but were associated with a small increase in purchase frequency. Variations in product selection only marginally affected purchasing behaviour.

These findings suggest that basic operational decisions could be delegated to POS, as adjustments are unlikely to strongly influence purchasing. However, substantial deviations from the study’s regulations – such as allowing advertising, competitive pricing, new product types, or an unlimited number of customers per POS – may lead to different outcomes. In more commercialised markets, such as in the USA, these factors have been linked to increased sales, use, and related harms. Therefore, certain restrictions are likely still necessary.

Purchase frequency was highest during the 30 days following participants’ initial purchase and then declined over 12 months at CSCs and pharmacies, while remaining stable at the DIZ. In contrast, the number of packages per purchase, as well as THC content and expenditure per package, remained stable over time. This suggests that a regulated cannabis market does not inherently lead to increased purchasing and, especially when combined with harm reduction measures, may even contribute to a decrease in purchase frequency.

Statement of Ethics

The Züri Can study was reviewed and approved by the Cantonal Ethics Committee Zurich (Kantonale Ethikkommission; BASEC research project number: 2022-01066). Written informed consent was obtained from all participants to participate in the study.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

The study is mainly funded by the City of Zurich. The University of Zurich provides additional resources for scientific personnel. The funder had no role in the design, data collection, data analysis, and reporting of this study.

Author Contributions

Nadine Heckel: writing – original draft, project administration, methodology, visualisation, conceptualisation, formal analysis, investigation, and data curation. Maximilian Buschner: writing – review and editing, methodology, and conceptualisation. Etna J.E. Engeli: methodology and conceptualisation. Patricia Dürler: project administration, methodology, and conceptualisation. Eva Maria Havelka and Maximilian Haas: investigation. Marcus Herdener: writing – review and editing, supervision, resources, project administration, methodology, and conceptualisation. Carlos Nordt: writing – review and editing, project administration, formal analysis, supervision, methodology, conceptualisation, visualisation, investigation, and data curation.

Funding Statement

The study is mainly funded by the City of Zurich. The University of Zurich provides additional resources for scientific personnel. The funder had no role in the design, data collection, data analysis, and reporting of this study.

Data Availability Statement

The research data are not publicly available due to data privacy. The syntax for running a simulation of the analysis in this article can be requested from the authors.