The global market for replacement car parts is a whopping US$1.9 trillion and it is overwhelmingly offline. Whether the customer is a garage, an insurance company, a repairer or an enthusiast they'll generally start their search online but end it by calling up a bunch of dealers on the phone to discuss the part.
Why? Well it turns out it's incredibly difficult to know if a given part will fit a given car. There are over 20,000 part manufacturers with 100m different parts between them. Each manufacturer makes multiple different types of cars and even those models have different variants. Together this all forms a huge matching problem.
To make things worse, the parts are often made by a parts manufacturer like Bosch and sold to an OEM like Toyota. Bosch often won't know which make or model of car the part was made for and Toyota won't know that the same part is also being sold to another OEM like Mercedes.
The OEMs would prefer you buy your parts directly from them for around twice the price so they don't mind things being murky.
So today, sellers of car parts must rely on old school methods to determine whether a part will be a fit for a buyer. They employ experts who have been in the industry for 30 years and know that a particular door handle that fits a Prius will also fit a Honda Civic. They also keep records and databases of which parts fit where.
But human memory and incomplete databases can only get you so far. Which is why buyers and sellers still need to talk to each other for most parts transactions. And it's why this entire marketplace is stuck offline.
Which brings us to the enormous opportunity of Partly - to solve the matching problem for car parts and bring this huge industry online.
Partly was born from a successful online marketplace in New Zealand called Allgoods. One day, cofounders Levi Fawcett and Nathan Taylor noticed that sales of car parts were completely different from sales of other products on their marketplace. There was overwhelming online demand for car parts and most of it was coming from outside NZ. People were clearly desperate to buy car parts online and the founders became fascinated when they saw how backward the industry was.
Levi is a mechatronic engineer with a history of attacking obscure problems. Of the five companies he has started one was to build caravans, another involved designing architectural balustrades. After uni he became an early employee at Rocket Lab where he oversaw their first 10 rocket launches as the head of hardware-in-the-loop simulations.
The problem presented by car parts matching was impossible for Levi to ignore. Soon he and Nathan were joined by Tony Austin the former Australian country manager of Amazon and Head of eBay Motors in the UK who had been trying to solve this same problem for some of the large online marketplaces.
It was clear to Levi and the team that the only way to get accurate matching across the entire industry was to attack the problem algorithmically. But traditional graph databases simply could not handle the over 400m linkages required to map which part fits which car. Accuracy was also paramount. As Levi puts it "if it’s not 99% accurate, it’s no good. If we say this part fits on your vehicle, it has to be right.”
So the founders turned to a new type of graph neural network technique that was starting to gain traction at large data companies like Pinterest. This enabled clever sampling and data transformation processes that could solve for both the massive size of the problem and the need for a high level of accuracy.
The traction so far is impressive. Partly has standardised and labelled 30% of car parts and 99% of the world’s vehicles to enable matching, making it a truly global parts database. This progress enabled the team to build software for sellers that enables them to manage their inventories and APIs for marketplaces like eBay who can then tell end users not just what parts are available but which will actually fit their car.
Partly has a long way to go to gather all the data required and enable sales of car parts to move online all across the world. In this they have been aided by the growing "right to repair" movement which has made access to car parts data somewhat easier.
Their partnership with eBay has also shown them that by solving the matching problem they can radically improve the experience of buyers and sellers and the amount of activity on the marketplace itself.
By continuing their mission, Levi and the team are poised to help bring one of the largest offline markets into the 21st century. We couldn't be more excited to be on this mission with them.