Designing for the infamous size & fit problem in e-commerce
Different brands have different size standards. It's difficult to go through the size chart every time.
The reason why I prefer offline shopping over online is that I can’t get the perfect size or touch and feel of it while shopping online.
Help the users look good by assisting them to make an informed decision on size and fit for the product being ordered on Myntra and thereby also reducing returns due to size and fit issues, increasing conversion and positively impacting the user experience.
Assist the users in identifying and choosing the perfect apparel size. Provide different touch points to collect, analyse and use the data that support size recommendations and thereby increasing the buying confidence.
We have done in-depth research sessions with the identified set of users to understand their behaviour while they buy online. Also quantitative data about the return reasons, previous order history, order exchange data, etc. gave us a better way forward towards the right solution.
Here are some of the major insights from the research
People check product fit & quality offline and order online for better discount
Different brands have different size & fit measures
Material plays a part in deciding the fit of the apparels
Most of the users order for their friends and family as well through Myntra
Users body shape also matters a lot in getting the correct fit between different brand sizes
Some of the existing solutions were explored during the initial research phase. A brief categorisation of third party solutions are given here.
With the help of data & research, we found that most of our customers have multiple size profiles. ie, users shop not only for themselves but also for their friends and family. So the solution should support specific recommendations for multiple size profiles in a single account.
Understanding our capabilities to identify product and users vectors were challenging. I had to collaborate with engineering and the operations team constantly to optimise the experience.
Experimenting with 3rd party solutions and then improving for in-house recommendation engine was challenging and time consuming.
Finding healthy balance between the regular order flow and size recommendation was tricky.
The idea is to build a recommendation engine that learns about the users, understand each product sizes and provide exact size & fit recommendation for the customers. We brainstormed about various use cases which includes; single size profile, multiple profile management, standardising size scale across brands, touch points to add additional size data, finding the right real-estate for recommendation, etc.
Stakeholders from product, engineering and design came together to brainstorm on the problem statement. I got different perspectives of the problem during these sessions and was able to iterate on my designs quickly.
The project was heavily focused on data science and engineering solutions to determine and predict size & fit. User information like order history, gender, age, etc. and product information like measurements, attributes, material, etc. were fed into the model.
Personalised Size Recommendation (PSR) feature on Myntra accurately predicts the best fit size for a customer. The recommendation comes just below the size selector on any product page and it prompts user to select the correct size.
For the customer who orders for multiple people, we introduced size profiling and managing system. For single profile users we were able to recommend the sizes without any further user inputs. For multi-profile users we have nudges on "order confirmation" and "my orders" page to tag the product to related profile.
We decided to standardise the sizes and create Myntra standard scale. The newly introduced scale will reduce the inconsistency of sizes between different brands, article types, material and other dependent attributes. Size charts of all the article types have been optimised based on the newly introduced scale.
From the learning of v1.0 of Myntra fit assist we took a step ahead. MFA v2.0 will cater to different user segments(logged out & new users), explore 3rd party services, span to more article types, increase adoption with incentivisation, improve browsing results, etc.