Netflix spends more than $150 million every year to improve and maintain their content recommendation engine.

At the core of this engine is an algorithm that recommends movies and television titles to users based on their location, time of day, viewing history, and thousands of other parameters.

The company employs around 800 product engineers, with more than 300 of those dedicated soley to keeping the content recommendation engine running.

This revelation came at an ACM conference talk on recommendation engines yesterday, where the company’s Chief Product Officer Neil Hunt told a shocked audience just how much the company spends trying to keep users watching for longer.

Hunt explained that a typical Netflix user will only browse for “one or two minutes,” considering between “20 to 50” shows before giving up on the platform entirely. They have one chance to get things right with users, which is why the company invests so much in ensuring the best match of content is recommended to each person at the first touch.

Surprisingly, a key part of this strategy includes recommending niche content to users. This helps ensure the long tail of all possible user interests is well represented while creating opportunities for new or less-well-known producers to have their work seen on the platform. “There are no bad shows, just shows with small audiences,” Hunt added.

He also told conference attendees that continuing to improve recommendations by even small increments can drastically reduce the number of users leaving the service, which protects the company’s ever-increasing revenues.

Hunt also revealed that Netflix has already drawn the ire of French regulators despite only launching in the country last month. France’s Council of State, a regulatory body that advises the government on legal matters, asked to review the recommendation algorithms used by the service to ensure that French and European content is well positioned in comparison to American content.

“We want to figure out what French people want to watch most and then tweak the catalogue to fit that,” Netflix’s Chief Content Officer said, telling regulators that the service will become more attuned with the wishes of French users as more users sign up to the service and begin watching shows. The ultimate aim of the recommendation engine, Hunt says, is to help democratize culture by the will of the people, and not the wishes of government bodies.