CoStar Group reduces costs by 80% using AWS Elemental MediaConvert

This blog is co-authored by Mark Osborn, Principal Software Engineer at CoStar Group.

Seeking greater control over their video transcoding capabilities while reducing costs and enhancing quality for their customers, CoStar Group embarked on a transformative solution. Let’s dive into how CoStar brought their video transcoding and distribution backends in house using Amazon Web Services (AWS)—delivering stunning 4K content to customers while reducing costs by 80 percent.

History

Since 2012, the videographers at CoStar have created bespoke video content for their online marketplaces. From professionally shot videos of apartment buildings, to highlight reels of aerial drone footage, to detailed clips about neighborhood amenities, the video team of CoStar captures it all. Their goal is to provide beautiful, informative, and original content to help their customers make informed decisions.

Until February 2025, CoStar used a third-party software as a service (SaaS) provider for their video transcoding and hosting needs. Their SaaS provided the following functions:

• A place to upload original videos to be transcoded to many different formats and resolutions, with a maximum resolution of 1080p, for playback on any type of device
• JPEG frame extraction at a certain point in the video to serve as a thumbnail
• Creation of a unique identifier for each video enabling it to be played back from a content delivery network (CDN)

In late 2024, CoStar started to explore AWS to see if they could build a home-grown replacement for the third-party service they were using. They set the following goals for the project:

• Be in control of their own destiny by owning and operating their video transcoding and distribution backends
• Increase video resolution to 4K for their customers
• Send the most appropriate resolution and format to their customers to ensure that everyone can view their content
• Make their playback URLs more SEO-friendly
• Reduce costs

With those goals set, it was time to start building.

Building a replacement solution

When building a like-for-like replacement for a pre-existing service, it is possible to break down the individual components that are needed.

Storage

For starters, CoStar needed storage for their source video files in the original format and resolution sourced from the videographers. They also needed a place to put their playback assets, variants of the videos that had been transcoded to different formats and different resolutions.

Amazon Simple Storage Service (Amazon S3) was the perfect fit due to its cost-effectiveness, scalability, and robust performance capabilities. This was also an excellent opportunity to leverage Amazon S3 Intelligent-Tiering storage class to allow lesser-played assets to automatically move to less expensive storage tiers.

Output formats and resolutions

As mentioned earlier, the third-party provider CoStar previously used would transcode their video assets into multiple formats and resolutions. For their replacement system, they needed similar capabilities, so they decided to use four resolutions in three formats:

• Resolutions: 360p, 720p, 1080p, and 4K (2160p)
• Formats: HLS, MP4, and WEBM

CoStar also established the business requirement of not upscaling video, since doing so could make images look blurry and artificial. For example, if the source video was 1080p, then they would not produce a 4K/2160p asset—customers would only receive an output of 1080p or lower.

Initially CoStar made changes to their various products so that they could play back video in Apple HTTP Live Streaming (HLS) format. HLS is a protocol that allows the player to measure its bandwidth during playback to pick the appropriate resolution and bitrate version. This is similar to the experience on platforms like Netflix, where videos may initially appear blurry before sharpening to higher resolution as the stream adjusts to the connection.

With CoStar selecting HLS as the main playback format, it meant that the appropriate resolution assets were automatically sent to customers. Devices with lower bandwidth would receive lower resolution, while devices with higher bandwidth would receive high resolution—up to 4K while ensuring videos started playing back as quickly as possible. MP4 and WEBM formats were also selected for their broad compatibility of playing on devices that don’t support HLS.

Transcoding video

AWS Elemental MediaConvert (MediaConvert) is a file-based video processing service that enables transcoding of video content to many different outputs, making it a perfect fit for the use case of CoStar. They utilized the MediaConvert API to create a transcoding job that converts a source video into three desired formats at their specified resolutions, while confirming they don’t upscale beyond the dimensions of the original video. CoStar utilized the Quality-Defined Variable Bitrate (QVBR) feature of MediaConvert with their Advanced Video Coding (AVC) outputs. QVBR automatically adjusts bitrate based on scene complexity to maintain consistent visual quality while optimizing file size.

CoStar also leveraged the Probe API in MediaConvert to retrieve metadata from their source videos (including resolution, duration, bitrate, and frame rate). They used the duration to extract a JPEG frame at a specific point in the video, calculated as a percentage of the total duration.

In all cases the output assets were placed in a separate S3 bucket using a deterministic key structure to make it possible to locate the assets for a particular video ID. 

Callback

Since some video assets can take longer to transcode than others based on their input resolution or duration, it was important to build this system to be asynchronous. When a transcoding job completes, MediaConvert publishes an event to Amazon EventBridge, which triggers an Amazon Simple Notification Service (Amazon SNS) topic. This topic then delivers messages to an Amazon Simple Queue Service (Amazon SQS) queue, invoking a callback AWS Lambda function. This helped CoStar to quickly integrate with downstream systems to mark a video as being available to show in their products.

Content Delivery Network

For static assets (like images or video) it is important to leverage a CDN to host these files as close to customers as possible. This provides the best customer experience by minimizing latency.

With the transcoded video assets of CoStar located in an S3 bucket, it was effortless to place an Amazon CloudFront distribution and Amazon Route 53 hostname in front—making assets addressable through HTTP URLs. Since the video assets are immutable, CoStar was able to configure Cache-Control HTTP headers appropriately to allow both CloudFront and web browsers to cache video assets for extended periods.

In the end, CloudFront provided a fully managed solution to quickly serve up assets from Amazon S3 without needing to write a single line of runtime code.

Search engine optimization-friendly URLs

One of the main goals CoStar had, was to be able to make their video playback URLs as SEO-friendly as possible. A common industry practice for making SEO-friendly URLs is to place an SEO slug as part of the URL (for example, https://example.com/videos/home-property-address/123456x.mp4). The slug is primarily used for SEO and readability in URLs, while the application typically relies on other identifiers (like IDs) to retrieve the actual content.

CoStar decided to make use of CloudFront Functions to canonicalize their URLs. CoStar could then ignore any slug segment and redirect onto the appropriate Amazon S3 asset. The beauty of this design is that a URL for a particular asset could have many different SEO slugs, but they would all be normalized to the same cache key—sharing the same asset cached by CloudFront.

Figure 1: New video encoding flow architecture using AWS Elemental MediaConvert.

Migration

With their new system built, tested, and deployed to production, CoStar started migrating almost one million pre-existing videos from their old system. This involved re-transcoding all the videos from their original source assets into their desired formats and resolutions.

For this part of the process, the scale and flexibility of MediaConvert made a huge difference. CoStar wrote a one-time migration tool to enqueue details of each source video into an Amazon SQS queue that activated a Lambda function. The Lambda function called the MediaConvert API to create a transcoding job, then waited for the callback through EventBridge to know when it was complete.

CoStar chose to transcode the back catalog over a period of one week. Out of nearly one million source videos, only seven failed to transcode due to corrupt source files. This gave CoStar tremendous confidence in their choice of MediaConvert for their transcoding needs.

Since they were re-transcoding assets from source files, it meant that they could offer 4K resolution outputs for any source videos that were in 4K. For CoStar this resulted in 20 percent of their back catalog becoming automatically “upgraded” to 4K resolution.

Go-live

With the new system up and running and the back catalog re-transcoded, the final step of the transition was to update their video URLs to point to their new CloudFront distribution.

For CoStar, this manifested as a SQL database patch, which they deployed over a period of two days. This gave them the ability to test in production and roll back if needed, due to increased traffic on the new CloudFront distribution, while monitoring analytics closely.

During both the migration and post-migration phases, Technical Account Managers from AWS Enterprise Support provided essential guidance and timely issue resolution.

Figure 2: Transcoded video with thumbnails on the Homes.com website.

Reducing costs

CoStar was thrilled that everything went smoothly, and at the end of February 2025 they had completely moved traffic off the old system. CoStar made use of the AWS Cost Explorer extensively during the migration period and first month of normal usage patterns. It helped them create an accurate idea of what the replacement system cost to run compared with the original third-party costs.

In the end, they reduced costs by 80 percent with their new AWS solution, and achieved all the goals they set for themselves (including offering 4K video to their customers).

Conclusion

CoStar faced the challenge of reducing costs and gaining control over their video transcoding and distribution systems while enhancing video quality. By implementing a comprehensive AWS solution centering around AWS Elemental MediaConvert, CoStar successfully replaced their third-party SaaS provider with a more flexible and powerful in-house system and reduced costs by 80 percent. With this solution, CoStar transcoded nearly one million videos into multiple formats and resolutions, including 4K, while also implementing SEO-friendly URLs and optimized content delivery.

Contact an AWS Representative to know how we can help accelerate your business.

Further reading

• Dynamically configuring job settings with AWS Elemental MediaConvert
• Enhance video efficiency with the bandwidth reduction filter in AWS Elemental MediaConvert
• Automating audio editing and transcoding using AWS

About CoStar Group

CoStar Group is one of the leading providers of real estate information in the world. They are well known for brands like Apartments.com, LoopNet, Homes.com, and Matterport. For almost 40 years, CoStar has provided industry-leading commercial real estate information, analytics, and a news platform that delivers insights to commercial real estate professionals and other industries. CoStar helps them make informed decisions more quickly. They track more than 7.3 million properties and have shot over 30 million professional photographs and one million videos of real estate, neighborhoods, parks, and schools.