DeveloperWeek 2020

In today’s world, time is more valuable than ever before and customers demand instantaneous results and solutions. With hundreds of thousands of local service professionals (pros) and millions of customers across the U.S., Thumbtack’s engineering team recognized this demand and made the decision to move their search engine from a manual model to an “instant” model in order to connect their pros with customers in an efficient manner and in effect, scale their product.

So how do you shift from a manual model where pro project quotes took hours or even days to provide, to an instant one that provides quotes in real-time? To start, the technology requirements for both their customers and pros needed to be identified. In Thumbtack’s case, the three requirements are broken down into matching, latency, and load. In order to fulfill matching needs, a new matching system had to be built that incorporated geographical preference, availability preference, and job preference across all categories for pros in addition to the needs of the customers. Latency had to shift to milliseconds rather than minutes, and the load matching system and latency has to scale for 50-60 of queries per second from millions of customers.

In this talk, Neha Agarwal, director of engineering at Thumbtack, will break down the changes she and her team made to build a new instant model from the ground up to scale their product. She’ll go through the different categories like job preferences, ranking, and quotes sheets, as well as what technology changes had to be made in order to create a real-time search engine that makes effortless matchmaking between customers and pros.