Understanding the water consumption associated with the use of hair care products: The impact of six hair characteristics on rinsing shampoos and conditioners

Abstract

Objective

Environmental life cycle assessment of hair care products shows that the highest environmental impact is associated with the use phase, rather than conception, production, packaging, distribution or disposal of the products themselves. To measure the water consumed in the use phase, an innovative and cost‐effective methodology was developed and tested to measure the water consumed to rinse off hair care products (rinsability).

Methods

Over 4 months, we tested the rinsability of 10 shampoos and 10 hair conditioners applied to 148 females, split between six hair characteristics: length, volume, dryness, thickness, curliness and damage. The volunteers were received in a hair salon on 20 different occasions for about 30 min each time. A team of hairdressers was specifically trained to detect two indicators of when a product is rinsed: a visual disappearance of the product and a clean touch. The volunteers were asked to have their hair washed at home 48 h before their arrival, using a standardized shampoo to control for sebum apparition.

Results

According to this test, on average, 7.1 L of water are needed to rinse a shampoo and 6.3 L to rinse a hair conditioner. However, there are significant differences depending on hair types: long and abundant hair requires more water to rinse shampoos and conditioners, whereas hair thickness, curliness, dryness and damage do not significantly affect the water required.

Conclusion

We suggest that data on product rinsability are essential when considering the water footprint for shampoos and hair conditioners. This method could be adopted for industry‐wide experimentation to assess the water footprint of products and set reduction targets.

Most water during a shower is used to rinse shampoo (average 7.1 L) and conditioner (average 6.3 L). We developed a methodology to test how rinsability varies according to hair characteristics, finding that hair length and volume significantly affect water use, while thickness, curliness, dryness and damage have little impact.

INTRODUCTION

Most development research on shampoos and conditioners focuses on their physical properties and their performance in relation to cleanliness and user experience, and more recently, the environmental impacts of the production of the product. Life cycle analysis studies on shampoos are common, and the impact of the use phase has been clearly demonstrated. However, the quantity of water consumed during the rinsing phase has not been rigorously quantified, and values generally used range between 7 and 15 L [1]. Water footprint assessment of the product life cycle of cosmetic products is necessary to promote consumer transparency and to guide manufacturers' efforts [2, 3]. We follow the recommendation from Golsteijn et al. [4] to provide quantifiable data on water consumed to inform life cycle assessment of environmental footprint methodologies for shampoos, and by extension other shower products (see [3] on shower gels). The aim of this study was to develop and test a methodology to measure the volume of water required to rinse the products off. To do so, we test a range of shampoos and conditioners in a sample of women to measure the effect of hair characteristics (i.e. length, volume, dryness, thickness, curliness and damage) on product rinsability, as part of efforts to incorporate equality, diversity and inclusion within product design.

An industry agreed protocol for the measurement of rinsing shampoos and conditioners would be a substantial contribution towards the definition of a shampoo life cycle assessment which can inform water footprint assessment and Product Environmental Footprint Category Rules for hair care products. At present, we know little about the methodologies followed to test rinsability of hair products. For example, two articles often quoted about this topic [5, 6] say very little about the methodology followed, and do not specify the type of hair that was tested. Because these methodologies are lacking, the development of water footprint category rules for shower products is based on assumptions regarding rinsability properties [4]. The first contribution of this study is the development of a reliable yet simple and replicable methodology to measure rinsability of hair care products.

Those studies that test for rinsability in relation to the chemical composition of the products rarely report the hair characteristics of the test population [7]. We argue that a universal methodology needs to be tested with a diverse sample to ensure that the results are as widely applicable as possible. To the best of our knowledge, no study has shown how hair characteristics will determine product rinsability. Women typically have four to eight square meters of hair surface area [8]; however, there is no such thing as average hair. When shampoos are tested for suitability for different hair types for different ethnic groups, it is from a cleanliness point of view and not for product rinsability [9]. The second contribution of this study is providing the first test of how hair care product rinsability differs against six hair characteristics (i.e. length, volume, dryness, thickness, curliness and damage).

LITERATURE REVIEW

The scarcity of water makes it crucial to reduce the amount of water collected, purified, transported and subsequently the treatment and the disposal of wastewater. This will also yield positive effects on associated chemicals and energy. Enhancing the availability of data for measuring and modelling household water usage is essential [10]. In Australia, it has been estimated that showers are responsible for the highest proportion of household water usage, accounting for 33% [11]. The differences in the environmental footprint of showers primarily depend on the type of water heater and the duration of the shower [12]. Females, children and especially teenagers tend to consume more water during showers [13]. Morizet et al. [3] confirmed that the water used in a shower can be divided into two categories: ‘useful water’, which represents the water objectively required for personal hygiene and remains relatively consistent among users, and ‘used water’, which encompasses the total volume of water used, significantly larger but subject to greater variability due to individual behaviour [14]. Manufacturers bear responsibility for providing products that inherently promote reduced water usage among consumers, rather than relying solely on social marketing nudges, which may initially seem effective, such as in the case of shower duration [15], only to see consumers revert to their customary behaviours [16].

Rinsability based on the hair product composition

One of the key determinants of a shampoo's rinsability is how it can be influenced by its chemical composition. Understanding how shampoo ingredients affect rinsability is not only important for the cosmetic industry but also for environmental sustainability. Research on shampoo design has made significant strides, particularly concerning its detergent, foaming, thickening, opacifying, sequestering and conditioning properties [17, 18]. However, the assessment of shampoo surfactant technologies' rinsability has not received substantial attention in the review and screening processes [19]. Surfactants, the primary cleansing agents in shampoo, play a role in determining rinsability. Common surfactants like lauryl sulfates and laureth sulfates are known for their ability to produce a rich lather and facilitate rinsing. However, their harshness on hair may necessitate the inclusion of conditioning agents, potentially impacting rinsability [17]. Non‐sulfate surfactants offer an alternative but may interact with lather formation and silicone deposition, making rinsing more challenging [19]. Shampoo formulations often include conditioning agents to counteract the drying effects of surfactants. While they can enhance hair texture, they may also affect rinsability. The choice of conditioning ingredients and their concentration is crucial in determining whether the shampoo can be easily rinsed away [17]. Thickeners and opacifiers contribute to the shampoo's texture and appearance but must be carefully balanced to avoid making the product too viscous, which can impede rinsability [17]. Other ingredients in the formulation, such as thickeners, emollients and preservatives, can impact rinsability. Thickeners, such as xanthan gum, can increase the viscosity of the product, making it harder to rinse away. Emollients, such as glycerin and silicone derivatives, may also leave a residue difficult to rinse away. Research has shown that it is possible to enhance shampoo rinsability through the use of amino acids [20], and tea‐tree extracts [18], among other methods.

In conclusion, the link between shampoo formulation and rinsability is a complex one that is impacted by a range of factors. Surfactant type, surfactant concentration, thickeners, pH and emollients, can all impact the rinsability of shampoo. Further research is needed to understand precisely the link between hair care chemical composition and rinsability.

While global brands have made significant strides in the initial stages of this life cycle, particularly in production and distribution [2], they have largely neglected the responsibility of assisting households in reducing their water consumption and the associated energy required for water heating. This oversight is evident despite the development of Product Environmental Footprint Category Rules for shampoos, which have identified product usage as the most critical stage for several environmentally impactful categories [4].

Rinsability based on hair characteristics

It is important that the design of new products is informed by equality, diversity and inclusion principles. In the past, few studies reported on hair characteristics when testing rinsability of products, and the summary below mostly demonstrates how much there still is to learn in this subject. Hair care products have often been designed according to hair conditions by race, such as hair dryness, thickness, curliness and damage (Kahre [21]). Although there are parallelisms between these characteristics and being of African, Asian and Caucasian descent [22], it is more appropriate to categorize hair differences beyond geo‐racial classifications [7]. Describing hair characteristics by race does not consider variability and cannot apply to mixed‐origin populations. For example, curliness can be described by morphological parameters on an 8‐point scale [23].

We know little about the impact of hair dryness, thickness, curliness and damage in relation to the rinsability of hair products, however. There is some evidence that African hair has a lower water content than Asian and Caucasian hair and does not become greasy like naturally straight hair (due to not becoming coated in sebum/oil [24]). As African hair is naturally dry, hair products will need to contain less harsh surfactants to avoid stripping away moisture and natural oils, and these products may therefore rinse differently.

There are also differences in the rules of cationic surfactants in hair conditioners depending on the hair length and volume of hair. Saturated single chain cationic surfactants with shorter chain lengths (C₁₄‐C₁₆) have faster desorption rates (will leave hair faster) and are used on short, thin or normal hair, whereas surfactants with longer chain lengths (C₁₈‐C₂₂) have higher substantivity (will stick more strongly to hair) and are used on long, coarse or damaged hair [25]. Shampoo formulation for fine hair should have a neutral/slightly acidic pH and a mild surfactant (Kahre, 2001). Products that are more easily rinsable will have ingredients with a lower affinity to hair, leaving less residue and therefore eliminating electrostatic charges; however, they will leave the hair a little more damaged and more difficult to style [25].

MATERIALS AND METHODS

Sensory evaluation methodology

The methodology designed for this study was a modification of L'Oréal's standard methodology, referred to as the Sensory Expert Evaluation and used to objectively assess the performances of shampoos and hair conditioners' formulae during their development. The Sensory Expert Evaluation is inspired by Sensory profile (ISO 13299) and performed by Sensory Experts who are trained hairdressers that assess sensory attributes from visual appearance, application and use for product like ease of detangling and hair smoothness. The standard assessment consists in comparing two products on a volunteer's head by applying each product on half a head (i.e. a within‐subject test). More recently, a new protocol was proposed to assess product rinsability, the volume of water required to rinse a product—shampoo and conditioner. To design the protocol, three Sensory Experts assessed the rinsability of two products, compared by half‐head, on 30 volunteers. Tests were developed using a constant flow of 9 L of water per minute and with an average hardness of 26°F (i.e. 260 mg/L of CaCO₃).

The protocol consisted of the following steps: (1) wet the volunteer's hair thoroughly; (2) divide the head longitudinally into two sides; (3) randomly apply each product (i.e. the control and the product to be tested) at the same time on each side of the head. Each test used the same quantity of product, 5.23 g for shampoos and 7 g for conditioners, as per AFNOR norm BPX30‐323‐5; (4) for shampoos only, massage the scalp on both sides to develop foam; (5) start the manual timer and rinse off the product on the left side; (6) stop the timer and collect the data; (7) repeat the operation on the right side. The dependent variable, the volume of water, was the result of multiplying the flow rate by the rinsing time.

This study followed that protocol, with some adjustments, as follows: At the end of Step 1, the volume of water, known as the wetting step, was collected. In steps 2 and 3, each product was tested in the whole head (i.e. between‐subjects). No control product was used. For the last steps, a commercially available smart handheld shower called Amphiro Digital Hand Shower V2. It accurately and consistently measured the volume of water used, improving the reliability of the measurements. The hand shower was developed by Amphiro AG, founded in 2009 as a spin‐off of the ETH Zurich (Swiss Federal Institute of Technology), following the successful development and testing of the Amphiro a1 and Amphiro b1, smart shower monitors that provide real‐time feedback during showers (e.g. energy and water consumed and other parameters) to encourage users to save energy (see Tiefenbeck et al. [26]). The handheld shower measured, in real time, the water temperature and the volume of water. The volume of water appeared on the handheld showerhead at the end of the rinsing as total rinsing volume. That data were written down by the Sensory Expert after each test. The rinsing volume of water, our dependent variable, was calculated as total rinsing volume minus wetting volume.

The first stage of this study was to develop the training in preparation for data collection. Five hairdressers were recruited and engaged in a training by L'Oréal Sensory Expert team. The 2‐day training followed the above rinsing protocol developed by L'Oréal, consisting of standardizing the rinsing process, specifically the quantity of product to be applied, the rinsing motion and the sensorial cues to end the rinsing. The quantity was established at 5.23 g for shampoos and 7 g for conditioners, independent of the hair characteristics, to be consistent with the AFNOR norm BPX30‐323‐5. All parameters linked to water quality and quantity (flow rate, temperature and water hardness) were standardized and fixed for the study, depending on the building facility. The first part of the training was theoretical, explaining the rinsing process for shampoos and conditioners, as follows:

• Step 1. Adjust the flow rate and the water temperature. The tap had to be opened to the maximum to get a flow rate of 10,4 L (±0.8 L) with a water temperature of 37°C (±2°C) and a water hardness level of 14°F (±4°F). Stop the water and record the water consumed.

• Step 2. Divide the participant's head hair longitudinally.

• Step 3. Open the tap and prepare to start on the left side of the head.

• Step 4. Start rinsing from the top of the head and pass the fingers between the hair and make lateral and vertical movements from the head to the ends to eliminate the product.

• Step 5. [Only for shampoos]. Rinse until the foam disappears. [For shampoo and hair conditioner]. Rinse until the residual coating becomes homogeneous and no longer varies.

• Step 6. Repeat the same process on the right side of the head.

• Step 7. Once the rinsing is finished, stop the water and record the volume of water.

• Step 8. [Only for shampoos]. Lift the hair slightly and make three very light circular movements with the fingertips to check that the foam does not restart.

During the practical part, each hairdresser used three products (of both shampoos and conditioners) defined by the previous methodology as having low, medium and high rinsability levels. The hairdressers rinsed each of these products twice, following the above steps, being unaware of the brand and level of rinsability, and with the products being randomly allocated. A total of 30 samples were taken for each product [shampoo/conditioner] (i.e. 5 hairdressers × 3 rinsability levels × two repetitions) on volunteers with the same hair characteristics. The volume of water was measured during steps 1 (i.e. wetting) and 7 (i.e. rinsing).

Volunteers' recruitment and hair characterization

Participants for this study were recruited through mailings, advertisements and social media. To participate in the study, volunteers subscribed to RedJade, a platform to manage recruitment for sensory research. Volunteers had to complete their personal data and sign up to participate in studies recruiting participants, so the RedJade team could select participants based on their match with the research conducted. All participants were informed that their personal data would be kept for 2 years from their registration or their last request to participate in a study, unless they requested the deletion during that time. In the event of participation in a study, they were also informed that their personal data would be kept for 5 years from the end of the study. However, volunteers could request the deletion of their data at any time by sending either an email to the Call Centre or to the Data Protection Officer. Each participant signed an agreement outlining the study's terms of reference and acknowledgment of GDPR compliance. Each participant was incentivized with a €20 gift voucher per visit, or a total of €400 if they completed the whole study. Participants could withdraw from the study at any time.

On signing the agreement, participants agreed to visit the evaluation centre (in Saint‐Herblain, Nantes, France) on 20 occasions for about 30 min each time. They also agreed not to wash their hair 48 h before their arrival and to use a standardized shampoo at home to control for sebum apparition. Data collection lasted 4 months. One hundred forty‐eight females were recruited after screening them to be over 18 years old, not working in advertisement/market research/cosmetics/chemistry/radio/TV, not currently involved in a hair product study, not pregnant, not intolerant to hair products, with no scalp problems, with no significant or very important presence of residual coating by other hair treatments, with no dreadlocks, not hair topped asymmetrically between right and left side, with no extensions and with no alopecia. Participants were classified according to six hair characteristics (i.e. curliness, damage, dryness, length, thickness and volume). Despite some parallelisms between two types of hair characteristics (e.g. curly hair is mostly dry rather than oily), we chose to treat these as separate independent variables. For each volunteer, a Sensory Expert was in charge of characterizing the hair according to the six hair characteristics. Each attribute was defined by visual and tactile characterization:

1. Length is the distance between the root and the ends,

2. Volume is measured by the quantity of hair in a ponytail,

3. Dryness describes the quantity of sebum generated (or not) after a 48 h wash,

4. Thickness is defined by the diameter of the fibre,

5. Curliness is defined following the visual atlas according to morphologic parameters, without referring to ethnic origin [23] and

6. Damage depends on whether the hair is natural or if a chemical treatment has been applied (and frequency of chemical treatment).

Figure 1 shows the distribution of each of the hair characteristics among the participants and their related categories. Due to the uneven distribution of participants among the categories in each hair type and the very low number of participants for some of the categories, we classified participants into two categories (i.e. low and high) for each hair characteristic, indicated in Table 1 and reflected in Figure 2.

FIGURE 1.

Distribution of participants across categories for each hair characteristic.

TABLE 1.

Re‐distribution of participants across two categories for each hair characteristic.

	Low	High
Curliness	Straight	Curly 1
	Wavy	Curly 2
		Kinky
		Very kinky
Damage	1	3
	2	4
Dryness	Mixed	Very dry
	Greasy	Dry
	Very greasy	Normal
Length	Very, very short	Half‐long
	Very short	Long
	Short	Very long
	Half‐short 1
	Half‐short 2
Thickness	Very fine	Thick
	Fine	Very thick
Volume	Low	Abundant
	Medium	Very abundant

FIGURE 2.

Re‐distribution of participants across two categories for each hair characteristic.

Selection of tested products

Product Environmental Footprint guidance suggests that a virtual representative product should be chosen for tests—a non‐existing product based on a combination of existing technologies [4]. Instead, we argue that the procedure we followed is more rigorous: We tested 10 shampoos and 10 conditioners available in the market to account for their effect alongside hair characteristics in water rinsability. Each of the products was provided to the five hairdressers in undistinguishable white bottles and were identified by blind codes.

RESULTS

Hairdresser training

The first step was to check whether the hairdressers' training was successful (i.e. no statistically significant differences in rinsability between hairdressers). The mean water used and SD by each hairdresser and measurement for both shampoos and conditioners (see Table 2) indicate that hairdressers used a lower volume of water the second time. Moreover, as also reflected in Figure 3, there was some degree of variability between hairdressers and between measurements per hairdresser.

TABLE 2.

Mean and SD per measurement and aggregated.

Hairdresser	Measurement	Mean (L)	SD	Mean (litres)	SD
1	1	6.78	1.88	7.98	2.15
2	1	8.55	2.53
3	1	7.92	2.25
4	1	9.03	2.28
5	1	7.63	1.70
1	2	6.65	2.93	7.33	2.39
2	2	7.50	3.58
3	2	7.73	1.87
4	2	7.71	2.31
5	2	7.07	1.36
1	1 & 2	6.72	2.35	7.66	2.28
2	1 & 2	8.03	3.01
3	1 & 2	7.83	1.98
4	1 & 2	8.38	2.29
5	1 & 2	7.35	1.50

FIGURE 3.

Mean variability in water consumed per hairdresser, product and measurement.

To check for statistical significance, we fitted a multilevel model, with a level 2 cluster consisting of the five hairdressers, and a level 1 unit consisting of the measurement occasion. In other words, multilevel modelling considers that measurements were grouped by different hairdressers, recognizing that each might have a unique pattern that could affect the volume of water, despite the training received.

Next, we fitted in R (version 4.2.3 [2023‐03‐15]) a null multilevel, unconditional random intercept model because it answers the question of how much variability the cluster‐level (i.e. hairdressers) effects account for. This initial model did not yet consider the type of shampoo or conditioner; instead, it measured how much variation in the volume of water could be explained by differences between hairdressers. Thus, we specified the outcome (i.e. volume of water) predicted by only the intercept, which could randomly vary between groups, in our case, the hairdressers.

The model indicated no significant differences in water consumption between hairdressers. We extended the model by adding measurement time and the type of products used (shampoos and conditioners). Even after accounting for these factors (through a random intercept model), the variance for hairdressers was 0 and 0.06, respectively, still showing no significant difference between hairdressers. Moreover, the model indicated no significant difference in average water consumption between the first and second measurement, suggesting that the volume of water used remained stable.

We assessed the training to be valid, as no statistically significant differences were registered in the average volume of water that each of the five hairdressers used to rinse each of the products (i.e. shampoo/conditioner) according to their level of rinsability (i.e. low/medium/high) in each repetition (i.e. first/second). In other words, (1) there was a good ranking between the three products, (2) we detected no hairdresser effect and (3) the pilot demonstrated a high degree of replicability. However, with the small sample size of five hairdressers, it is possible that this model may not have enough power to detect variability, even if it exists. For this reason, we included hairdressers as a potential independent variable in the next analysis where we measured the effects of various products and hair characteristics on rinsability.

Rinsability of shampoos and conditioners

Once the training successfully concluded, the next step was to develop the experiment. The total sample contains up to 10 observations per shampoo and 10 per conditioner from each participant, as some participants did not test every product. Because we had multiple measurements from the same participants, we could expect the observations to be closely related to each other, violating the independence assumption of residuals (i.e. measurements taken on the same subject are likely to be more like each other than to measurements taken on different subjects). As we aimed to analyse the effect of the participants' hair characteristics on the volume of rinsing water while accounting for repeated measurements from participants, a multilevel model was used for the analysis, as it explicitly models this dependency, with a cluster and unit level.

Multilevel models include both fixed effects (e.g. the shampoo/conditioner type, volume of water during wetting, water temperature and hair characteristics) and random effects (e.g. individual differences among participants and hairdressers). The random effects also account for the within‐subject correlation in the data (i.e. different measurements from the same participant), thereby dealing with the non‐independence of observations. We chose to fit a random intercept model to allow each participant's baseline water use (i.e. average) to differ. This model assumes that the relationship between the predictor variables (i.e. hair length, dryness, thickness, volume, curliness, and damage, type of shampoo/conditioner, wetting and water temperature) and the outcome variable (i.e. water used for rinsing the shampoo/conditioner) is the same for all participants. In other words, if two participants had identical hair characteristics, used the same shampoo/conditioner and rinsed under the same conditions (i.e. temperature and wetting), we assumed they would use the same amount of water. We also assumed that each hair characteristic independently influenced water usage, meaning the impact of one characteristic (e.g. length) would not depend on another (e.g. curliness). Regarding the hairdressers, although we did not find statistically significant differences between them, we still allowed for variations between them, acknowledging that human differences could remain despite the training.

To understand our data visually, we started by displaying the distribution of the volume of water consumed to rinse the shampoos and conditioners, finding they were normally distributed (Figure 4). The descriptive statistics are indicated in Table 3. Figure 5 shows the mean and standard error variability in water rinsability for shampoos and conditioners for each participant, with the dashed red line indicating the overall mean. Figures 6, 7, 8, 9, 10, 11 show the effect of each hair's characteristic on water used. Last, Figure 12 presents the boxplots showing water use for shampoos and conditioners per participant, again highlighting the overall mean with a red dashed line.

FIGURE 4.

Histogram of the litres of water used in rinsing shampoos and conditioners (n = 1448 and 1444, respectively).

TABLE 3.

Descriptive statistics for litres of water used in rinsing shampoos and conditioners.

	Min.	1st Qu.	Median	Mean	3rd Qu.	Max.	SD
Shampoos	1.500	6.000	6.900	7.143	8.200	15.700	1.786
Conditioners	2.200	5.200	6.200	6.398	7.500	14.600	1.775

FIGURE 5.

Mean, ordered from low to high, and standard error volume required per participant across shampoos (above) and conditioners (below). The red, dashed line represents the overall mean for participants.

FIGURE 6.

Dryness effect on rinsability.

FIGURE 7.

Thickness effect on rinsability.

FIGURE 8.

Length effect on rinsability.

FIGURE 9.

Volume effect on rinsability.

FIGURE 10.

Curliness effect on rinsability.

FIGURE 11.

Damage effect on rinsability.

FIGURE 12.

Boxplots for participants' effect on shampoo and conditioner rinsability.

Null model‐ shampoos

We first fitted the null model, the simplest possible multilevel model, which focused on how much water was used with participants during shampoo rinsing, without considering other factors. It was used, on average, 7.15 L of water (SE = 0.07, t = 107.8, df = 144.54, p < 0.001). The random effects analysis revealed significant variations in water used between participants, with a variance of 0.36 (SD = 0.60). The residual (unexplained) variance was large (2.83, SD = 1.68), reflecting differences within each participant's repeated measurement. The scaled residuals exhibited a range from −3.61 to 4.47, indicating some deviation from normality but no extreme outliers.

To understand the variability due to differences between participants, we calculated the intraclass correlation coefficient (ICC). The ICC was 0.112, indicating that 11.2% of the variance in water consumption can be attributed to differences between participants. The substantial amount of unexplained variability was a consequence of other factors, such as hair characteristics, type of shampoo, water temperature, wetting and/or others. Thus, the next step was to include these other predictors to find their impact on the outcome variable by fitting a random intercept model.

Null model—conditioners

We conducted a similar analysis for conditioners, starting with the simplest model. It was required, on average, 6.40 L (SE = 0.05, t = 132, df = 145.78, p < 0.001) to rinse conditioners. The variation between participants was quite small, with a variance of 0.03 (SD = 0.17), and a large residual variance of 3.12 (SD = 1.77). The scaled residuals exhibited a range from −2.34 to 4.59, indicating some deviation from normality but no extreme outliers.

The low ICC value of 0.009 indicates that 0.9% of the variance in water consumption could be attributed to differences between participants. Thus, as with shampoos, the next step was to include the other predictors to find their impact on the outcome variable by fitting a random intercept model.

Random intercept model—shampoos

We fitted the next two random intercept models to examine the effect of shampoo type, wetting and water temperature on water used before considering the characteristics of participant's hair. Model 1 SHP included random effects for participants and hairdressers, while model 2 SHP for participants only. The aim was to analyse whether including for hairdressers' effect improved the model. The models were formulated as follows:

$$\begin{array}{c}\mathsf{\text{Model}}\mathsf{1}\mathsf{SHP}:\text{litre}{\mathrm{s}}_{\mathit{ijk}}={\beta }_0+{\beta }_1\left(\text{shampo}{\mathrm{o}}_{\mathit{ijk}}\right)+{\beta }_2\left(\text{wettin}{\mathrm{g}}_{\mathit{ijk}}\right)\hfill \\ +{\beta }_3\left(\text{temperatur}{\mathrm{e}}_{\mathit{ijk}}\right)+u_{0j}+v_{0k}+e_{\mathit{ijk}},\hfill \end{array}$$

where: ${\text{litres}}_{\mathit{ijk}}$: water volume for observation $i$ from participant $j$ and hairdresser $k$. ${\beta }_0$: fixed intercept. ${\beta }_1,{\beta }_2,{\beta }_3$: fixed effects for shampoo, wetting and temperature, respectively. $u_{0j}$: random intercept for participant $j$, with $u_{0j}\sim N\left(0,{\sigma }_u^2\right)$. $v_{0k}$: random intercept for hairdresser $k$, with $v_{0k}\sim N\left(0,{\mathrm{\sigma }}_v^{2.}\right)$. $e_{\mathit{ijk}}$: residual error, $e_{\mathit{ijk}}\sim N\left(0,{\sigma }_e^2\right)$.

$$\begin{array}{c}\mathsf{\text{Model}}\mathsf{2}\mathsf{SHP}:{\mathsf{\text{litres}}}_{\mathit{ij}}={\beta }_0+{\beta }_1\left({\mathsf{\text{shampoo}}}_{\mathit{ij}}\right)+{\beta }_2\left({\mathsf{\text{wetting}}}_{\mathit{ij}}\right)\hfill \\ +{\beta }_3\left({\mathsf{\text{temperature}}}_{\mathit{ij}}\right)+u_{0j}+e_{\mathit{ij}},\hfill \end{array}$$

$\text{where}:{\text{litres}}_{\mathit{ij}}$: water volume for observation $i$ from participant $j$. ${\beta }_0$: fixed intercept. ${\beta }_1,{\beta }_2,{\beta }_3$: fixed effects for shampoo, wetting and temperature, respectively. $u_{0j}$: random intercept for participant $j$, with $u_{0j}\sim N\left(0,{\mathrm{\sigma }}_u^2\right)$. $e_{\mathit{ij}}$: residual error, $e_{\mathit{ij}}\sim N\left(0,{\mathrm{\sigma }}_e^2\right)$.

After comparing both models (see Table 4), model 1 SHP revealed a better fit to the data (ΔAIC = 50), as indicated by the Likelihood Ratio Test (χ² = 51.97, df = 1, p < 0.001). Therefore, we decided to include hairdressers as random effects, in addition to participants, in the next model, where hair characteristics were added as predictors.

TABLE 4.

Random intercept models comparison for shampoos (model 1 SHP).

Model	AIC	BIC	Log‐likelihood	Deviance
1 SHP	5609.2	5688.4	−2789.6	5579.2
2 SHP	5659.2	5733.1	−2815.6	5631.2

Building on Model 1 SHP, we included the hair characteristics, formulated as follows:

$$\begin{array}{c}\mathsf{\text{Model}}\mathsf{3}\mathsf{SHP}:{\mathsf{\text{litres}}}_{\mathit{ijk}}={\beta }_0+{\beta }_1\left({\mathsf{\text{shampoo}}}_{\mathit{ijk}}\right)+{\beta }_2\left({\mathsf{\text{wetting}}}_{\mathit{ijk}}\right)\hfill \\ +{\beta }_3\left({\mathsf{\text{temperature}}}_{\mathit{ijk}}\right)+{\beta }_4\left({\mathsf{\text{length}}}_{\mathit{ijk}}\right)+{\beta }_5\left({\mathsf{\text{dryness}}}_{\mathit{ijk}}\right)\hfill \\ +{\beta }_6\left({\mathsf{\text{thickness}}}_{\mathit{ijk}}\right)+{\beta }_7\left({\mathsf{\text{volume}}}_{\mathit{ijk}}\right)+{\beta }_8\left({\mathsf{\text{curliness}}}_{\mathit{ijk}}\right)\hfill \\ +{\beta }_9\left({\mathsf{\text{damage}}}_{\mathit{ijk}}\right)+u_{0j}+v_{0k}+e_{\mathit{ijk}},\hfill \end{array}$$

$\text{where}:{\text{litres}}_{\mathit{ijk}}$: water volume for observation $i$ from participant $j$ and hairdresser $k$. ${\beta }_0$: fixed intercept. ${\beta }_1\mathit{\text{to}}{\beta }_9$: fixed effects for shampoo, wetting, temperature and each of the hair characteristics, respectively. $u_{0j}$: random intercept for participant $j$, with $u_{0j}\sim N\left(0,{\sigma }_u^2\right)$. $v_{0k}$: random intercept for hairdresser $k$, with $v_{0k}\sim N\left(0,{\sigma }_v^2\right)$. $e_{\mathit{ijk}}$: residual error, $e_{\mathit{ijk}}\sim N\left(0,{\sigma }_e^2\right)$.

As expected, the model 3 SHP showed a better fit to the data compared to the model 1 (ΔAIC = 42.9), as indicated by the Likelihood Ratio Test (χ ² = 54.91, df = 6, p < 0.001) (see Table 5).

TABLE 5.

Random intercept models comparison for shampoos (model 3 SHP).

Model	AIC	BIC	Log‐likelihood	Deviance
1 SHP	5609.2	5688.4	−2789.6	5579.2
3 SHP	5566.3	5677.2	−2762.2	5524.3

The final random intercept model applied to shampoos (see Table 6) showed the effect of all predictors (shampoo type, wetting, temperature and the six hair characteristics) on water rinsability. Specifically, compared to Shampoo 1 (used as the reference), all shampoos except Shampoo 2 significantly reduced the amount of water used. Shampoo 5 had the highest decrease in water volume, with approximately 0.94 litres compared to Shampoo 1. However, further analysis showed that the effect of Shampoo 5 was not statistically different from shampoos 3, 4, 6 and 9, which performed similarly well in reducing water usage during rinsing.

TABLE 6.

Final random intercept model for shampoos.

Factors	Estimate	SE	t‐value	p‐value
(Intercept)	0.20811	1.39963	0.14869	0.88182
SHP2	−0.24365	0.18841	−1.29318	0.19618
SHP3	−0.82289	0.19356	−4.25129	0.00002***
SHP4	−0.57048	0.19381	−2.94349	0.00330**
SHP5	−0.93621	0.19182	−4.88067	<0.0001***
SHP6	−0.83658	0.19428	−4.30601	0.00002***
SHP7	−0.42329	0.19622	−2.15721	0.03117*
SHP8	−0.53326	0.19263	−2.76829	0.00571**
SHP9	−0.83390	0.19254	−4.33094	0.00002***
SHP10	−0.51148	0.19373	−2.64020	0.00838**
Wetting	0.16015	0.03013	5.31525	<0.0001***
Temperature	0.16415	0.03688	4.45117	0.00001***
Length—High	0.42284	0.10390	4.06991	0.00008***
Dryness—High	0.20405	0.10521	1.93956	0.05454
Thickness—High	−0.01829	0.10261	−0.17829	0.85876
Volume—High	0.49332	0.10662	4.62709	0.00001***
Curliness—High	0.19644	0.10709	1.83426	0.06883
Damage—High	0.14037	0.10268	1.36698	0.17391

Note: Significant p‐values are denoted as follows: ***p < 0.001, **p < 0.01 and *p < 0.05.

Also, the model indicated that increases in the amount of water used for wetting and the water temperature significantly increased water used during rinsing by approximately 0.16 L.

Hair length and volume significantly contributed to rinsing water use. Participants with longer hair (i.e. from half long to very long) required 0.42 L more water than those with shorter hair (i.e. from very, very short to half short 2). Similarly, participants with more hair volume (i.e. abundant or very abundant) also required 0.49 L more than those with less hair volume (i.e. low or medium).

Random intercept models—conditioners

As with shampoos, we found the best model had to include all the predictors as fixed effects and both participants and hairdressers as random effects. The model was formulated as follows:

$$\begin{array}{c}\mathsf{\text{Model}}\mathsf{3}\mathsf{\text{COND}}:{\mathsf{\text{litres}}}_{\mathit{ijk}}={\beta }_0+{\beta }_1\left({\mathsf{\text{conditioner}}}_{\mathit{ijk}}\right)\hfill \\ +{\beta }_2\left({\mathsf{\text{wetting}}}_{\mathit{ijk}}\right)+{\beta }_3\left({\mathsf{\text{temperature}}}_{\mathit{ijk}}\right)+{\beta }_4\left({\mathsf{\text{length}}}_{\mathit{ijk}}\right)\hfill \\ +{\beta }_5\left({\mathsf{\text{dryness}}}_{\mathit{ijk}}\right)+{\beta }_6\left({\mathsf{\text{thickness}}}_{\mathit{ijk}}\right)+{\beta }_7\left({\mathsf{\text{volume}}}_{\mathit{ijk}}\right)\hfill \\ +{\beta }_8\left({\mathsf{\text{curliness}}}_{\mathit{ijk}}\right)+{\beta }_9\left({\mathsf{\text{damage}}}_{\mathit{ijk}}\right)+u_{0j}+v_{0k}+e_{\mathit{ijk}},\hfill \end{array}$$

where ${\text{litres}}_{\mathit{ijk}}$: water volume for observation $i$ from participant $j$ and hairdresser $k$. ${\beta }_0$: fixed intercept. ${\beta }_1\mathit{\text{to}}{\beta }_9$: fixed effects for conditioner, wetting, temperature and each of the hair characteristics, respectively. $u_{0j}$: random intercept for participant $j$, with $u_{0j}\sim N\left(0,{\sigma }_u^2\right)$. $v_{0k}$: random intercept for hairdresser $k$, with $v_{0k}\sim N\left(0,{\sigma }_v^2\right)$. $e_{\mathit{ijk}}$: residual error, $e_{\mathit{ijk}}\sim N\left(0,{\sigma }_e^2\right)$.

The results (see Table 7) indicated that all conditioners significantly reduced the water used during rinsing, compared with Conditioner number 1, serving as the referent. Conditioner 3 had the largest effect with 3.37 L less than Conditioner 1. Moreover, its effect was statistically significantly different from any of the other conditioners.

TABLE 7.

Final random intercept model for conditioners.

Factors	Estimate	SE	t‐value	p‐value
(Intercept)	0.87363	1.17451	0.74383	0.45711
CON2	−0.47313	0.16655	−2.84071	0.00457**
CON3	−3.37478	0.16951	−19.90940	<0.0001***
CON4	−2.10888	0.16901	−12.47758	<0.0001***
CON5	−2.26157	0.16738	−13.51127	<0.0001***
CON6	−2.01834	0.16877	−11.95902	<0.0001***
CON7	−0.80631	0.16848	−4.78579	<0.0001***
CON8	−2.10819	0.16762	−12.57685	<0.0001***
CON9	−1.36482	0.17007	−8.02519	<0.0001***
CON10	−1.09579	0.16873	−6.49442	<0.0001***
Wetting	0.05496	0.02833	1.93969	0.05264
Temperature	0.17982	0.03147	5.71321	<0.0001***
Length—High	0.24104	0.09450	2.55061	0.01180*
Dryness—High	−0.14830	0.09495	−1.56186	0.12061
Thickness—High	0.01165	0.09266	0.12577	0.90009
Volume—High	0.30579	0.09576	3.19335	0.00174**
Curliness—High	0.06577	0.09690	0.67871	0.49844
Damage—High	0.06468	0.09247	0.69950	0.48542

Note: Significant p‐values are denoted as follows: ***p < 0.001, **p < 0.01 and *p < 0.05.

The wetting process did not significantly affect the volume of water used for rinsing the conditioners. However, as with shampoos, each one‐unit increase in water temperature significantly increased the water use during rinsing by approximately 0.18 L.

Similar to the shampoo analysis, hair length and volume significantly influenced the volume of water used during conditioner rinsing. Specifically, participants with longer hair (i.e. from half long to very long) required approximately 0.24 more L of water than those with shorter hair (i.e. from very, very short to half short 2). Similarly, participants with higher hair volume (i.e. abundant or very abundant) required approximately 0.31 more L than those with lower hair volume (i.e. low or medium).

Multilevel diagnostics (i.e. residuals, outliers, normality, homoscedasticity and multicollinearity) were performed to ensure that the models met the assumptions.

DISCUSSION AND CONCLUSIONS

Water footprint assessment of the product life cycle for cosmetic products is necessary to promote consumer transparency and to guide manufacturers' efforts to reduce usage [2, 3]. However, the development of Product Environmental Footprint Category Rules for shampoos guesstimated the amount of water used for shampooing as 15 L per shower, which was considered by Golsteijn et al. [4] to be a potential overestimation. Earlier data from AFNOR [1] had suggested that 7 L are necessary to rinse a shampoo, consistent with our findings showing that, on average, across all conditions, 7.14 L are needed to rinse a shampoo. We also found that 6.39 L are needed to rinse a hair conditioner. This data may be useful to inform the criteria used in the EU ecolabel for rinse‐off cosmetic products [27].

We have learned in this study that hair characteristics have an impact on the ‘useful water’ needed to remove a product from hair. Having long and voluminous hair requires more water than low hair length and low hair volume to rinse shampoos and hair conditioners. Hair thickness, curliness, nature and damage do not significantly affect the water required for rinsing shampoos and hair conditioners. Further research on water usage with haircare products should consider hair type for any robust inter‐study comparison.

We suggest that standardized data on product rinsability is essential in the development of water footprint methods for shampoos and hair conditioners, and we propose industry‐wide experimentation with and subsequent adoption of this methodology.

This agile methodology can be replicated by other cosmetics companies to measure the minimum amount of water needed to rinse hair care products and to inform the development of more water‐efficient products. From the 55 L found to be used in the average shower in France [1], we now know that only about a third of that amount is necessary water—the so‐called ‘useful water’, and the rest is used for pleasure, since a recent study found that 4.77 L are needed for heating the water and wetting the body, and 4.15 L for rinsing the shower gel off the full body [3], to which we need to add the 7.14 L for rinsing shampoo and 6.39 L for hair conditioner from our own findings. This is important because the environmental life cycle assessment of hair care products nowadays shows that the greatest environmental impact is associated with showering time, rather than conception, production, packaging, distribution or disposal of the products themselves.

This study has been conducted in lab situations with a standardized Sensory Evaluation methodology (shower installation, controlled gesture and controlled quantity of product), allowing us to provide objective hair care technologies and hair characteristics comparison. Considering water hardness, other internal studies seemed to demonstrate an impact on rinsability. Hence, we decided to fix this parameter, but it could be interesting to focus on this criterion in another study. Furthermore, it would be interesting to explore the use of this methodology in different countries and contexts to compare sets of data.

All those investigations could create a robust, industry‐wide, objective knowledge base to enable the creation of consumer‐centric strategies to encourage the reduction of water used under the shower.

To further understand water usage, we would benefit from measuring the consistency of these findings with a real‐life approach considering the potential influence of product dosage and water quality.

FUNDING INFORMATION

This work was supported by the L'Oréal Research and Innovation Department.

CONFLICT OF INTEREST STATEMENT

There is no conflict of interest.

Julie D, Pablo P‐D, Xavier F, David M. Understanding the water consumption associated with the use of hair care products: The impact of six hair characteristics on rinsing shampoos and conditioners. Int J Cosmet Sci. 2025;47:1026–1042. 10.1111/ics.13082