Matching dogsitters all about algorithms

Scott Porad

As CTO, Porad and his team work on algorithms that match dog owners with petsitters. The team uses machine learning algorithms that look at which sitters have the highest satisfaction ratings, book most frequently and have the fewest safety issues. It also factors in geography, because dog owners typically want a sitter that’s in their neighborhood.

Over time the algorithms learn what dog owners look for in a sitter. Not only do the algorithms use this to improve recommendations, but the platform gives advice to sitters who don’t currently rank highly, helping them to improve.

“You can’t rely on intuition in this case,” Porad said. “You have to rely on data because you’re dealing with so many different things.”

To develop and maintain the algorithms, Porad has hired a team with a diverse set of backgrounds. The data science and engineering team consists of 25 people who have degrees in biology and nuclear engineering, in addition to more conventional fields like computer science and statistics.

The benefit of this approach, Porad said, is that you get different perspectives on problems. People from divergent fields will look at problems differently, which can lead to data-driven innovation. “You’re bringing different points of view to the table,” Porad said.

While many analytics managers favor staff who have familiarity with business problems, Porad is hiring people out of academia. What’s important is that they have a scientific mindset. This works at Rover because the nature of the problem it’s working on — matching dog owners with safe and convenient petsitters — lends itself to a scientific approach more so than other analytic pursuits, like data-driven marketing or process automation.

To do analyses, Porad and his team rely on a tool called Periscope, which lets users write SQL queries and build and share data dashboards. The tool is connected to an Amazon Redshift data warehouse, a PostgreSQL database and a Zendesk customer database. More data science-heavy work is done in Python Notebook, a browser-based Python tool.

Porad said he will support other tools if the need arises. For example, the Rover technology team has used the open source R programming language to some extent in the past. But he’s critical of the trend of supporting whatever tool analysts ask for. He tried that approach at previous jobs but found that when everyone uses their own tool it makes it harder for people to work collaboratively.

“I just saw that things got really inflexible,” said Porad, who also worked previously at online pharmacy Drugstore.com. “It really just ties your hands behind your back a lot.”

High stakes for innovating on analytics

There’s a lot at stake for Porad and his team. He said throughout the sharing economy, businesses are still trying to figure out how best to serve customers. It’s particularly challenging because, whether you’re talking about Rover or Uber or Airbnb, the companies don’t interact with customers directly. Instead, customers deal primarily with independent contractors. The platform’s role is to ensure the interactions are smooth and deliver a good experience.

This is why analytics is so important. While a business in the sharing economy might not be able to base decisions on the actual experience of interacting with customers, they at least have the data, which can still tell them a lot. It takes data-driven innovation to capitalize on the data.

“The companies that can figure out what’s happening with their customers better are the companies that are going to win,” Porad said.