US 5,790,935 · Granted 1998-08-04

Hughes' 1998 Patent for Predicting What You'll Want to Watch Tomorrow

Imagine if Netflix could guess which shows you'd want to watch next, then secretly download them to your home box during late-night hours when the internet isn't busy. That's what this patent does—it uses everyone's viewing patterns to predict what each person might like, then pre-loads it locally so the network doesn't get crushed during prime time.

Curated and verified by the IsItPatented Editorial team

The plain-English version

What it protects

The claim covers a system where a central server analyzes viewing preferences across all subscribers using collaborative filtering (essentially, 'people who like X also like Y'), predicts what individual subscribers will want, and automatically stores those recommendations on local devices during off-peak hours. What's protected is the combination of predictive recommendation logic, off-peak pre-loading, and the reduction of peak-hour network traffic through this local-storage strategy.

Why it matters

This patent captured an elegant solution to a real problem in the 1990s: how to deliver on-demand video without overwhelming network infrastructure. By shifting some of the computational and storage burden to the subscriber's home device, the system conserved expensive bandwidth during peak hours. The collaborative filtering approach—predicting user preferences from collective behavior—became foundational to recommendation engines that streaming services later built at scale.

Real-world use

When a streaming service downloads a few suggested shows to your device overnight and they're ready to play instantly the next evening, that's the kind of prediction and pre-caching this patent anticipated.

Original USPTO abstract

A digital information system delivers virtual on-demand information over existing, as well as the next generation, digital transport systems by offloading a portion of the systems' peak bandwidth requirements to the local subscribers. A collaborative filtering system synthesizes the preferences of all of the subscribers and then predicts those items that each subscriber might like, and therefore request. Each subscriber is provided with a local storage device for storing, during off-peak hours, those items recommended by the collaborative filtering system. As a result, only a relatively few subscriber requests must be serviced directly from the central distribution system.

Patent details

Publication number: US 5,790,935
Filing date: 1996-01-30
Grant date: 1998-08-04
Assignee: Hughes Aircraft Company
Inventor(s): PAYTON; DAVID W.
CPC class: H04N21/26241

View original record on Google Patents →

Want to file your own patent?

If you're designing a smart home gadget that learns user habits, check our free patent scanner to see what's already locked down in predictive recommendation territory.

Free patentability scan