Optimizing the seller listing onboarding process for a leading e-commerce platform

Most businesses have a website presence in 2023. More than a quarter of businesses conduct their activities online, and more than two fifths of small businesses are planning to invest in improving their website performance. These statistics mean that with more and more businesses moving online, there has been an increase in the number of sellers on e-commerce platforms. As a result, e-commerce businesses need to find ways to onboard new sellers quickly and efficiently without sacrificing quality. Turnaround time and quality are crucial success factors for e-commerce businesses. Process optimization can enable onboarding new sellers and their product listings quickly and accurately.
Find out how Netscribes used probability analysis to help a leading e-commerce platform reduce the time spent on onboarding seller listings while maintaining quality.

Challenges

Turnaround time and quality are two uncompromising success factors every business is eager to ace. A leading e-commerce platform with a similar goal approached Netscribes. It wanted to reduce the time spent on onboarding each seller’s listings across various product lines while keeping quality intact. Large volumes of SKUs and a continuous stream of product descriptions had to be vetted every day, to ensure they were factually correct and aligned with the guidelines of the e-commerce platform.

Approach

To help them achieve this, Netscribes conducted a probability analysis on two fronts:

Listings

Since varying SKUs/ products had different guidelines to be met, each listing was scored for its likelihood of being inaccurate, and also for their likelihood of meeting every guideline separately.
The scoring was performed by processing the input stream of listings through an ensemble of Regression and Machine Learning models for higher prediction accuracy.
This framework helped us predict:

• Which listing is likely to be incorrect?
• What guidelines are they failing to meet?

Agents
The agents vetting these listings were profiled and scored based on their accuracy and speed in completing their tasks.

Results delivered

• The scores deduced from each listing and each agent’s performance were analyzed to optimize task allocations. This entailed mapping listings that demanded a higher degree of scrutiny and diligence, to agents with superior performance track-record in terms of accuracy and speed.
• Each agent was equipped with incorrect probability scorecards across each product listing, allowing them to optimize their time and devote the right levels of diligence to products that required it.