Case Study – Product Matching

The people and businesses we work with do some pretty groundbreaking stuff. Naturally, this sometimes means we need to omit a client’s identity or commercially sensitive details to respect confidentiality. In this case study, the client’s name has been removed for this reason.

Cogent works with a rapidly growing startup that is building a disruptive online marketplace platform. The two-sided marketplace connects suppliers with consumer facing businesses.

The startup has experienced rapid growth by developing a SaaS solution that deeply integrates with suppliers’ businesses, building on existing relationships with their customers who place orders on the platform.

The Opportunity

Every supplier has their own list of products that might have hundreds or thousands of individual items. The startup also maintains their own list of products across industries to maintain data quality and provide a unified ordering experience for their customers. One step in the new supplier onboarding process is mapping the supplier’s product list into the startup’s list, and right now this is done manually.

Manually mapping products requires expert knowledge since products can have different names across suppliers based on their region, linguistic or technical backgrounds. Further compounding this problem are different systems of measurement for unit quantities (e.g. metric vs imperial). This is an essential part of the onboarding process, but means that bringing a new supplier on can take several weeks.

“It’s a lot of work to take on a new supplier because every new supplier who joins has a good understanding of their products and their customers. But a supplier might have hundreds or thousands of products, and every product from that supplier needs to be matched or added to the thousands of products available in our list.”

With international expansion, the range of suppliers and products joining the platform continues to expand. Meanwhile, the startup’s development team is focused on rapidly iterating and delivering new platform features.

Cogent’s goal was to figure out whether AI could simplify and speed up customer onboarding by designing a product concept that was technically feasible, desired by users and commercially viable.

AI facilitated onboarding would mean a more seamless experience for new suppliers, less manual work for the onboarding team, and shorter lead times getting suppliers active on the platform.

Process

Our Automate service provided a good framework for nutting out these chunky problems. We used it to dig deep into the startup’s business and the way supplier onboarding data is handled right now.

As is most often the case, we started with the people. Through onsite observation and interviews with staff at their offices, we learned how the startup onboards their suppliers. We began to understand the interfaces with other people and systems, the physical environment where people worked, what was working well, and where the key problems were. We listened to the staff describe what they needed and wanted and used that to develop a product concept and a proof of concept (PoC) – a prototype.

By understanding the people first, we were able to design a solution that was empathetic to the existing team, their processes, and systems. Using this vital insight and the existing data that the startup already has about their suppliers and their own product lists, we trained a purpose-built machine learning algorithm to perform product mapping.

Our PoC shows promise and the next step will be to continue to generalise our algorithm and test how it performs for new products from suppliers that it has never seen before.

Outcome

“If the AI could help sort even half of the products in a supplier’s original list, and we did that pretty accurately, then that would save an immense amount of time. Out of four weeks, for a big supplier, it cuts that work to two weeks. And it’s not just about reducing the amount of work. It also means the supplier can come onto their platform sooner and they can start selling faster.”

In short, we’re on the right track, which is the whole purpose of Automate. We proved that AI could help facilitate onboarding new suppliers and that it was worth the business investing in it to do so. Currently, conversations are continuing to figure out how to take the algorithm and evolve it for inclusion it into their product roadmap.

“We’re working together towards a goal that we’re both happy with. Cogent does all the heavy lifting when it comes to the AI work until we are happy with the result. And we’ll obviously be helping them to inform them during that process.”

If you’re interested in anything you’ve seen here and have questions about how to apply AI or machine learning to features or products effectively, reach out to us at Cogent.

Find out more about AI at Cogent here.