About the customer
Leading BNPL player with over $10 billion in annual online sales
World's top-3 "Buy Now, Pay Later" (BNPL) provider, with a yearly handling of over $10BN in online sales. With a strong grip on the e-commerce market's financial aspects, customer decided to expand into the marketplace business by facilitating the end-to-end buying cycle within their app.
Company Goal
Provide buyers with a safe and informative in-app experience
BNPL platform was looking to expand their services to facilitate customer product reviews. Their goal was to ensure that buyers could make an educated, informed purchasing decision right within their app — all while maintaining safety and cleanliness on their platform. The idea was to only keep relevant, product-focused reviews on the platform while filtering out the ones that focused on delivery, customer service, or any other topic. Expected volumes were set at over 100K reviews per month.
Solution
State-of-the-art Hybrid AI solution, tailored to customer's needs
Customer teamed up with Membrace to tackle this tricky challenge. Membrace provided the following services:
- Relevance Validation
- Sentiment Analysis
- Review Moderation
- Relevance Validation
- Sentiment Analysis
- Review Moderation
Hybrid AI solution was developed for the customer, which combines Machine Learning and Crowdsourcing for fast, scalable and cost-effective results of high quality. Reviews are checked for relevance to the product being described and compliance with the service rules: absence of rude and explicit language, mention of personal data, advertising appeals and clickbait mentions.
The current design allows 90% of the data to be checked using ML models, and cases requiring manual validation are sent for additional crowd verification and the resulting verdicts are used to train algorithms, which allows it to adapt to new trends at the current moment in time.
Opinion
“Helped to moderate reviews with a peak load of 250,000 objects per day”
Membrace solutions allowed for a comprehensive evaluation of each single review, helped [us] visualize the overall buyers' sentiment on popular products and ensure the trust & safety of their buyers across their ecosystem of products.
The daily flow of data for moderation averages 3,000 objects per day, with a peak load of 250,000 objects per day.
Data is supplied in English, German, Spanish, Swedish, Danish and other languages (in total data is supplied in 51 languages)
The daily flow of data for moderation averages 3,000 objects per day, with a peak load of 250,000 objects per day.
Data is supplied in English, German, Spanish, Swedish, Danish and other languages (in total data is supplied in 51 languages)
Results
Ensured the trust and safety of buyers across their ecosystem of products
With Membrace, the client is able to provide shoppers with only the most relevant information they need to make purchase decisions, while removing distractions and unhelpful and unsuitable content.
3,000 average
Daily processed objects
250,000
Peak daily workload
51 languages
number of project-supported languages
98%
Solution accuracy