Which is the best clothing size recommendation technology for fashion e-commerce?
|
PRIME AI |
The best size recommendation technology is the one with
the ability to match individual customer’s body shape
with any retailer’s unique garment specifications.
When comparing size recommendation tools, many factors have to be considered, including, in order of importance:
• the accuracy of the size recommendation being served;
• the user experience (UX) when interacting with the solution;
• the ease of implementation for the retailer;
• and finally, if there are any other added benefits for the business.
Let’s have a closer look at different technologies and their approach to finding customers’ perfect fit online.
When comparing size recommendation tools, many factors have to be considered, including, in order of importance:
• the accuracy of the size recommendation being served;
• the user experience (UX) when interacting with the solution;
• the ease of implementation for the retailer;
• and finally, if there are any other added benefits for the business.
Let’s have a closer look at different technologies and their approach to finding customers’ perfect fit online.
Size recommendation - Matching customer body shape to
garment specifications at SKU level [PRIME AI
technology]
PRIME AI technology matches garment SKU to customer body
shape. Thanks to artificial intelligence using dedicated
neural networks developed internally for this sole
purpose, the solution is able to identify garment
specifications as well as deviations in manufacturing
process without any input from retailers. The learning
process is based on data gathered from real-time
customer purchases and returns on the retailer’s
website.
Accuracy – Prime AI technology uses machine learning to accurately understand each individual body shape based on weight, height, and a limited set of parameters. It will generate size recommendations with high accuracy and retailers will benefit from true and meaningful reduction in returns that other fit recommendation methods have failed to demonstrate.
PRIME AI size recommendation is not reliant on the accuracy of the retailer’s size charts. The technology uses the retailer’s size charts (or its own models if the retailers cannot provide size charts) as a baseline but quickly evolve them based on sales and returns information captured. As a result, PRIME AI can provide useful feedback to retailers on the actual accuracy of the size charts they hold.
UX – The user interface is fully customisable to best match retailer’s branding as well as website overall look and feel. In addition, PRIME AI is continuously evolving the widget functionalities to reflect customer behaviour and rapid changes seen in mobile technology. The input required from customers is voluntarily kept to a minimum, requiring low effort without overwhelming the user with too many questions, or asking to take pictures of one’s body or wearing special costume.
There is no requirement for customers to create an account with PRIME AI unlike with other similar solutions. Also, PRIME AI will not ask customers to think about other competitor brands when in process of shopping. Therefore, retailers will benefit from higher conversion and lower returns than using brand to brand comparison technology.
Ease of implementation – PRIME AI is able to collect garment specifications and define initial sizing models with no input from the retailer. Multi-brand retailers don’t have to provide size charts or measurements of the garments they are selling. There is no physical handling of the garment involved either. This capability is unique to PRIME AI.
PRIME AI keeps integration to retailer’s e-commerce platform as low effort as possible. The integration does not require complex coding, and even non-technical personnel can be guided effectively to enable personalised clothing fit recommendation. In other words, integration is a quick, easy and low cost process.
Other added benefits – PRIME AI supplies monthly actionable insights and metrics captured by the size recommendation widget. A dedicated account manager paired with data scientists will help retailers to understand their data down to SKU level. Analytics retention is set to 365 days. Also, in the scenario of the recommended size not being available, alternative products will be recommended utilising AI powered recommendation engine, or customers can benefit from back in stock functionality and cross device shopping behaviour tracking at no additional costs.
Ultimately PRIME AI can provide retailers with non-negligible benefits from significant uplift in conversion ratio and noticeable returns reduction that no other fit recommendation method on the market can match.
Below is an overview of other competitor technologies providing size and fit recommendation, offering some insights into how they work and why accuracy level differs.
Accuracy – Prime AI technology uses machine learning to accurately understand each individual body shape based on weight, height, and a limited set of parameters. It will generate size recommendations with high accuracy and retailers will benefit from true and meaningful reduction in returns that other fit recommendation methods have failed to demonstrate.
PRIME AI size recommendation is not reliant on the accuracy of the retailer’s size charts. The technology uses the retailer’s size charts (or its own models if the retailers cannot provide size charts) as a baseline but quickly evolve them based on sales and returns information captured. As a result, PRIME AI can provide useful feedback to retailers on the actual accuracy of the size charts they hold.
UX – The user interface is fully customisable to best match retailer’s branding as well as website overall look and feel. In addition, PRIME AI is continuously evolving the widget functionalities to reflect customer behaviour and rapid changes seen in mobile technology. The input required from customers is voluntarily kept to a minimum, requiring low effort without overwhelming the user with too many questions, or asking to take pictures of one’s body or wearing special costume.
There is no requirement for customers to create an account with PRIME AI unlike with other similar solutions. Also, PRIME AI will not ask customers to think about other competitor brands when in process of shopping. Therefore, retailers will benefit from higher conversion and lower returns than using brand to brand comparison technology.
Ease of implementation – PRIME AI is able to collect garment specifications and define initial sizing models with no input from the retailer. Multi-brand retailers don’t have to provide size charts or measurements of the garments they are selling. There is no physical handling of the garment involved either. This capability is unique to PRIME AI.
PRIME AI keeps integration to retailer’s e-commerce platform as low effort as possible. The integration does not require complex coding, and even non-technical personnel can be guided effectively to enable personalised clothing fit recommendation. In other words, integration is a quick, easy and low cost process.
Other added benefits – PRIME AI supplies monthly actionable insights and metrics captured by the size recommendation widget. A dedicated account manager paired with data scientists will help retailers to understand their data down to SKU level. Analytics retention is set to 365 days. Also, in the scenario of the recommended size not being available, alternative products will be recommended utilising AI powered recommendation engine, or customers can benefit from back in stock functionality and cross device shopping behaviour tracking at no additional costs.
Ultimately PRIME AI can provide retailers with non-negligible benefits from significant uplift in conversion ratio and noticeable returns reduction that no other fit recommendation method on the market can match.
Below is an overview of other competitor technologies providing size and fit recommendation, offering some insights into how they work and why accuracy level differs.
Size recommendation - brand to brand matching [Other
companies]
Such technology is built on the relative comparison of
size charts from different brands. Customers are
required to provide their size for other brands of
clothes they wear and they might potentially know. The
solution generates a recommended size based on the
tabulated relationship between the retailer’s size chart
and the competitor’s size chart. Retailers will see
gains in conversion ratio and (potentially) slightly
lower returns ratio. However, over time the impact to
conversion and returns will fade away due low accuracy
of the method.
Accuracy – The biggest weakness of such method is that size charts themselves are not very precise. In addition, garments measurements in manufacturing process do deviate from the original specifications. That deviation is never reflected in size charts, resulting in the same size chart being displayed for many items in the same category despite many of them fitting differently. Therefore, the size chart comparison approach can only provide approximate recommendation at category level. Considering the lack of accuracy, the impact on returns is really minimal. Transactional information is being used only to collect statistical data.
Consider this example: If the customer enters a height of 200cm, weight of 100kg, and tells the tool that he is wearing size XS in another brand, then the tool, will recommend a best fitting size close to XS. This is clearly wrong and obviously not the best method to generate size recommendation.
UX – It is worth noting that displaying competitors’ names on a given retailer’s website might not be the best marketing strategy. Most of e-commerce experts would have serious concerns mentioning other brand names at the most crucial stage in a customer’s journey, when they are about to finalise their purchase. Despite hard work and significant costs to attract new customers to their website, the solution essentially grants free exposure to competitors on browsing platforms where it is easy and fast to visit other websites.
Asking for brand names and associated sizes also results in more questions, extending the shopping experience, which will eventfully restrict full revenue potential. It is common knowledge that average time spent shopping on site is in decline due to customer shift to mobile platforms as well as the possibility to visit alternative retailers easily.
Providers of this type of technology may offer solutions that do not display other retailers brands to address concerned of competitor exposure. However, this goes against the original foundations of the method and will come at the expense of accuracy.
Ease of implementation – this type of technology usually does not require any complex coding on the retailer’s side. Therefore, implementation is easy and fast as long as retailer has size charts available. This solution cannot be implemented for multi-brand retailers if they can’t provide size charts.
Other added benefits – retailers should consider how reliable will the insights provided be. Understanding of individual SKU is very limited as size charts are created at category level and their accuracy is questionable. Retailers may potentially find some interesting statistical trends, which should be interpreted carefully.
Accuracy – The biggest weakness of such method is that size charts themselves are not very precise. In addition, garments measurements in manufacturing process do deviate from the original specifications. That deviation is never reflected in size charts, resulting in the same size chart being displayed for many items in the same category despite many of them fitting differently. Therefore, the size chart comparison approach can only provide approximate recommendation at category level. Considering the lack of accuracy, the impact on returns is really minimal. Transactional information is being used only to collect statistical data.
Consider this example: If the customer enters a height of 200cm, weight of 100kg, and tells the tool that he is wearing size XS in another brand, then the tool, will recommend a best fitting size close to XS. This is clearly wrong and obviously not the best method to generate size recommendation.
UX – It is worth noting that displaying competitors’ names on a given retailer’s website might not be the best marketing strategy. Most of e-commerce experts would have serious concerns mentioning other brand names at the most crucial stage in a customer’s journey, when they are about to finalise their purchase. Despite hard work and significant costs to attract new customers to their website, the solution essentially grants free exposure to competitors on browsing platforms where it is easy and fast to visit other websites.
Asking for brand names and associated sizes also results in more questions, extending the shopping experience, which will eventfully restrict full revenue potential. It is common knowledge that average time spent shopping on site is in decline due to customer shift to mobile platforms as well as the possibility to visit alternative retailers easily.
Providers of this type of technology may offer solutions that do not display other retailers brands to address concerned of competitor exposure. However, this goes against the original foundations of the method and will come at the expense of accuracy.
Ease of implementation – this type of technology usually does not require any complex coding on the retailer’s side. Therefore, implementation is easy and fast as long as retailer has size charts available. This solution cannot be implemented for multi-brand retailers if they can’t provide size charts.
Other added benefits – retailers should consider how reliable will the insights provided be. Understanding of individual SKU is very limited as size charts are created at category level and their accuracy is questionable. Retailers may potentially find some interesting statistical trends, which should be interpreted carefully.
Size recommendation - matching customer to customer
[other companies]
This technology statistically compares what customers
with same body measurements have bought and returned.
Therefore small, medium and luxury retailers can’t
expect a significant impact and can only anticipate a
very limited reduction in returns if any, due to limited
data being collected to be statistically meaningful.
Accuracy – Customers are being segmented based on their measurements while using returns data to identify low return segments. The technology potentially can be more accurate than the method of comparing size charts. However, it requires a significant number of data points to reach statistically acceptable conclusions. An additional weakness of such approach is that it cannot provide size recommendation based on customer fit preference nor body shape. The tool will also not be able to recommend any size for customers with less common measurements due to lack of enough statistical data to generate a meaningful recommendation. For example, very tall and skinny person.
Most importantly the tool still does not understand individual garment specifications nor individual customer body shape or fit preference. Hence, there will always be a non-negligible portion of customers not getting the right recommendation.
UX – The user interface requires less questions than any other tool, allowing for a quick and efficient process. Some technology providers offer additional steps to give more confidence with regards to accuracy (e.g. they add fit preference selector which makes no difference in many cases due to insufficient data points to create new customer segments).
Recommendation can be confusing for customers. For example, the tool might indicate: “65% of customer like you bought size Small and 35% Medium”. Which shows that there is still a good chance of selecting the wrong size, also limiting retailer’s conversion ratio due to some customers doubting in which bracket they fall (65% or 35%). If customer has less common measurements, there will be no recommendation generated for them at all. Retailers should consider to what percentage of visitors they would be comfortable saying there is no suitable size for them!
Ease of implementation – the technology can be very easy to implement as it also does not require any size charts or garment measurement figures allowing multi brand retailers with significant sales volume to use this tool effectively. However, as mentioned earlier the tool will have very limited impact on returns as large number of people will get inaccurate recommendation. For new brands, there might also be a gestation period where the tool will need to gather enough data to be able to serve statistically correct recommendations.
Other added benefits – Any insights will be statically more valuable and trustworthy than the size chart comparison models. However, there will still be large amount of data in grey areas. In the scenario of people falling into different segments due to their measurements, recommendation will not feel right and they will choose to ignore it or shy away from completing a purchase due to doubts on size. Following any data insights in such scenario may bring unexpected and costly results for the retailer in long run.
Accuracy – Customers are being segmented based on their measurements while using returns data to identify low return segments. The technology potentially can be more accurate than the method of comparing size charts. However, it requires a significant number of data points to reach statistically acceptable conclusions. An additional weakness of such approach is that it cannot provide size recommendation based on customer fit preference nor body shape. The tool will also not be able to recommend any size for customers with less common measurements due to lack of enough statistical data to generate a meaningful recommendation. For example, very tall and skinny person.
Most importantly the tool still does not understand individual garment specifications nor individual customer body shape or fit preference. Hence, there will always be a non-negligible portion of customers not getting the right recommendation.
UX – The user interface requires less questions than any other tool, allowing for a quick and efficient process. Some technology providers offer additional steps to give more confidence with regards to accuracy (e.g. they add fit preference selector which makes no difference in many cases due to insufficient data points to create new customer segments).
Recommendation can be confusing for customers. For example, the tool might indicate: “65% of customer like you bought size Small and 35% Medium”. Which shows that there is still a good chance of selecting the wrong size, also limiting retailer’s conversion ratio due to some customers doubting in which bracket they fall (65% or 35%). If customer has less common measurements, there will be no recommendation generated for them at all. Retailers should consider to what percentage of visitors they would be comfortable saying there is no suitable size for them!
Ease of implementation – the technology can be very easy to implement as it also does not require any size charts or garment measurement figures allowing multi brand retailers with significant sales volume to use this tool effectively. However, as mentioned earlier the tool will have very limited impact on returns as large number of people will get inaccurate recommendation. For new brands, there might also be a gestation period where the tool will need to gather enough data to be able to serve statistically correct recommendations.
Other added benefits – Any insights will be statically more valuable and trustworthy than the size chart comparison models. However, there will still be large amount of data in grey areas. In the scenario of people falling into different segments due to their measurements, recommendation will not feel right and they will choose to ignore it or shy away from completing a purchase due to doubts on size. Following any data insights in such scenario may bring unexpected and costly results for the retailer in long run.
Size recommendation – other methods
Other methods include:
Body scanning technologies with various scanning approach from wearing special costumes to using special cameras, apps and so on. Anything making user journey longer, more complicated or even raising concerns over their privacy will lead to people not using the tool. Despite retailer bearing costly implementation.
Physical measurement of garments, where each items is being measured by hand or dressing artificial mannequins to see how stretchable clothes are. Using such methods significantly slows down supply chain and result in new products staying without size recommendation until they are measured. This is very labour-intensive work bearing big costs in order to be operational and scalable.
Body scanning technologies with various scanning approach from wearing special costumes to using special cameras, apps and so on. Anything making user journey longer, more complicated or even raising concerns over their privacy will lead to people not using the tool. Despite retailer bearing costly implementation.
Physical measurement of garments, where each items is being measured by hand or dressing artificial mannequins to see how stretchable clothes are. Using such methods significantly slows down supply chain and result in new products staying without size recommendation until they are measured. This is very labour-intensive work bearing big costs in order to be operational and scalable.
Conclusion: clothing size recommendation by PRIME AI
Today, PRIME AI offers the most accurate size
recommendation technology on the market, effectively
matching customer body shape down to garment SKU. As a
result, retailers gain competitive advantage from the
data collected, which is being processed and reported
and analysed by the company’s data scientists and
fashion retail experts.
PRIME AI solution is suitable to any retailer regardless of their budget or number of SKUs.
Request a demo now or contact us to find out more!
PRIME AI solution is suitable to any retailer regardless of their budget or number of SKUs.
Request a demo now or contact us to find out more!