Need cloud computing? Get started now

Akamai Launches Early Hints to Further Boost User Experience and SEO

Akamai Wave Blue

Written by

Tim Vereecke, Robin Marx, and Debayan Roy

July 09, 2024

Tim Vereecke

Written by

Tim Vereecke

Tim Vereecke loves speeding up websites and has spent the past 15+ years exploring the technical and business aspects of web performance. He is a Web Performance Architect at Akamai and also runs scalemates.com: the largest (and fastest!) scale modeling website on the planet.

Web Performance Architect and Developer Champion Robin Marx posing outside

Written by

Robin Marx

Dr. Robin Marx is a Senior Web Performance Specialist in Akamai's presales organization. He helps customers with performance audits and provides tailored recommendations for improvement. Since doing his PhD on protocols like QUIC and HTTP/3, Robin has been a thought leader in the field, explaining complex technical topics in approachable blog posts and conference presentations. Robin also has a somewhat unusual hobby as, on weekends, he likes to hit people with longswords.

Debayan Roy

Written by

Debayan Roy

Deb Roy is a Product Manager at Akamai. He loves traveling, trekking, and making new connections with people who have interests in the field of cloud security and artificial intelligence. He has experience with both startups and enterprises, in roles spanning sales, technical consulting, and account management. As a product manager, Deb is focused on solving problems around performance, user experience, and engagement.

This is an updated version of the original blog post that published on October 10, 2023.

We're thrilled to announce the global launch of Early Hints, a powerful tool designed to supercharge your website's performance.
We're thrilled to announce the global launch of Early Hints, a powerful tool designed to supercharge your website's performance.

We're thrilled to announce the global launch of Early Hints, a powerful tool designed to supercharge your website's performance. Early Hints empowers your website optimization efforts with the ability to tune Core Web Vitals (CWV) like never before. 

It’s no secret that a fast website is a prerequisite for a great user experience. Since Google incorporated CWV as a search engine optimization (SEO) ranking factor, performance has been top of mind for many organizations. 

Serving HTML from cache remains the best way to ensure solid performance — however, consistently serving content directly from the edge cache may pose challenges, including: 

  • Uncacheable content

  • Long-tail content

  • Volatile content

  • Personalized content

  • A/B and multivariate testing

  • Frequent deploys/cache purging

This is where Early Hints come into play. Akamai Early Hints offers a complementary solution to HTML caching, while providing web developers with increased control to maximize performance.

What is Early Hints?

Early Hints employs the HTTP 103 status code that allows the Akamai edge server to send preliminary HTTP headers ahead of a final response (e.g., 200, 301, 404). 

The payload of this HTTP 103 response is a list of resource hints that: 

  • Suggest the preloading of critical subresources (e.g., CSS, fonts, logo, etc.) 

  • Encourage preconnecting to additional domains

Why do we need Early Hints?

If the HTML page cannot be cached at the edge, the edge server must fetch it from the origin. This can introduce delays (tens or hundreds of milliseconds) during which the connection between the browser and the edge is idle, despite having the potential to be used (Figure 1).

This can introduce delays (tens or hundreds of milliseconds) during which the connection between the browser and the edge is idle, despite having the potential to be used (Figure 1). Fig. 1: Example of a WebPageTest waterfall that illustrates the behavior of a page without Early Hints enabled on Akamai

Early Hints allows the browser to download key resources during this idle time, while waiting for the HTML content to arrive.

Early Hints example

Figure 2 depicts the same uncached search results page from Figure 1, but now with Early Hints enabled on Akamai.

Figure 2 depicts the same uncached search results page from Figure 1, but now with Early Hints enabled on Akamai. Fig. 2: Example of a WebPageTest waterfall that illustrates the behavior of a page with Early Hints enabled on Akamai

The first byte of the actual search result page (HTTP 200 OK) comes in at the 130 ms mark. Well before this, at approximately 40 ms, the HTTP 103 response arrives and kicks off loading the main font, an SVG image, two JavaScript files, and the mPulse RUM library.

Consequently, when the browser receives the HTML for the page, it can immediately start rendering the content for the end user without having to wait for the font and JavaScript to be downloaded. In this example, the First Contentful Paint (FCP) shifts from approximately 400 ms to 300 ms.

Performance benefits

The performance benefits of this process include:

Browser support

Although Chrome was the first browser to support Early Hints, other browsers have since joined, including Safari, Firefox, and Opera. (See caniuse.com to verify the latest browser support.)

Define what to send to the browser

How do you define what to send to the browser as Early Hints? The hints are really nothing more than a list of resources/URLs to preload or preconnect to, conveyed as HTTP Link headers. So, how do you determine which Link headers to send? Several conceptual approaches are possible, including:

  • Set Early Hints yourself at the origin and let the CDN follow your lead

  • Configure them manually at the CDN 

  • Calculate them dynamically at the CDN, using tools like Akamai EdgeWorkers and Akamai EdgeKV

  • Have the CDN infer them from mPulse RUM data using machine learning capabilities

Akamai’s approach

Each of these approaches has distinct trade-offs in terms of flexibility and ease of use. While some competitors mainly follow the origin's lead, Akamai chose resolutely for full control at the edge to: 

  • Optimize performance

    • No performance penalties for first hits (long-tail content, expired content)

    • A/B and multivariate testing proof

    • Reduced risk of hinting outdated content

  • Ensure easy setup and maintenance

    • No origin changes required

    • Granular and programmatic control

    • Allows edge-aware logic using connection behavior, referrer, and cookies

    • No cache-purging complexities

    • EdgeWorkers and Property Manager compatibility

  • Allow the choice between

    • Full control for performance-savvy customers (available now)

    • Automated control for quick wins (available soon)

Positive customer feedback

Performance-focused customers have been highly engaged in prototype testing Early Hints, and their feedback has been extremely positive:

  • “We see a 300 ms LCP improvement in real user monitoring (RUM) at the P75.”

  • “We love the flexibility in deciding what to hint.”

  • “We are already playing with it; super easy and it just works!”

  • “After rolling out Early Hints, we measured 30% improvements in LCP.”

Our customers spoke and we listened

Thanks to a very successful beta implementation, our customers helped us to make the production version of Early Hints easier to use without compromising flexibility. Some changes we made include:

  • Adding a new Early Hints Property Manager behavior with variable support (Figure 3)

  • Adding support for more delivery and security solutions

  • Adding support for HTTP/3 on top of HTTP/2

  • Removing the need to implement manual browser checks

  • Removing the need to validate the presence of the sec-fetch-mode=navigate request header

  • Updating documentation on Akamai TechDocs

Updating EdgeWorkers code examples on GitHub

Adding a new Early Hints Property Manager behavior with variable support (Figure 3) Fig. 3: The new Early Hints Property Manager behavior with variable support

The future of Early Hints

Akamai is dedicated to providing our customers with the best possible performance, and we look forward to further enhancing our Early Hints offering.

We plan to make this offering more effective and even easier to use through an automated, machine-learning–based version, called Intelligent Hints. Akamai will automatically be able to analyze your web pages and implement the most appropriate logic to early hint the most relevant resources that impact and improve CWV like LCP.

Get started today

You can get started today. Akamai TechDocs explains how to add the Early Hints behavior to your Akamai delivery configurations. If you’d like additional help with figuring out which hints you should send, reach out to our performance experts via your account team.



Akamai Wave Blue

Written by

Tim Vereecke, Robin Marx, and Debayan Roy

July 09, 2024

Tim Vereecke

Written by

Tim Vereecke

Tim Vereecke loves speeding up websites and has spent the past 15+ years exploring the technical and business aspects of web performance. He is a Web Performance Architect at Akamai and also runs scalemates.com: the largest (and fastest!) scale modeling website on the planet.

Web Performance Architect and Developer Champion Robin Marx posing outside

Written by

Robin Marx

Dr. Robin Marx is a Senior Web Performance Specialist in Akamai's presales organization. He helps customers with performance audits and provides tailored recommendations for improvement. Since doing his PhD on protocols like QUIC and HTTP/3, Robin has been a thought leader in the field, explaining complex technical topics in approachable blog posts and conference presentations. Robin also has a somewhat unusual hobby as, on weekends, he likes to hit people with longswords.

Debayan Roy

Written by

Debayan Roy

Deb Roy is a Product Manager at Akamai. He loves traveling, trekking, and making new connections with people who have interests in the field of cloud security and artificial intelligence. He has experience with both startups and enterprises, in roles spanning sales, technical consulting, and account management. As a product manager, Deb is focused on solving problems around performance, user experience, and engagement.