Hurry up⚠️ Only 20 tickets left in this pool WaysAwardsSubmit your WaysAwards project — enter the 2026 nominations

Personalisation

Building for the personalisation spectrum

September 4, 2023 10:00 AM

Long Lecture

🇬🇧 English

CMYK

About

Abstract:

Simple sites can enjoy simple architectures and hosting strategies. But what happens when we need to offer more complex, dynamic experiences to our users? Or when we need to offer localisation or personalisation? Things can get difficult when we need to build and serve these sites with confidence.

Modern rendering models and workflows make this far more approachable than before. And technologies like Edge Functions allow us to enrich our sites with more advanced features for our users.

In this talk we’ll learn about popular approaches to rendering our sites, and see how we can choose between them. And we’ll explore a practical example using Edge Functions to deliver personalised experiences to our users without compromising on performance or robustness.

Presentation outline:

1. Understanding the challenge

• The challenges when building web sites with personalisation

• The Personalisation Spectrum (Universal, Individual and the stages between)

• Traditional and popular approaches

• Rendering patterns that aim to help

2. Seeking a solution

• Edge-side-rendering and how it complements other rendering

• Desirable properties for sites that scale

• How to design your site for robust and scalable personalisation

‍

3. Learning by example

• Exploring a practical example of delivering localised and personalised content

• Reviewing the key takeaways and concepts

• Next steps for learning and experimenting

Talk by

Photo of Phil Hawksworth, Principal Developer Experience Engineer at Netlify

Phil Hawksworth

Principal Developer Experience Engineer at Netlify

Watch the full talk

Watch this WaysConf session, then continue with related talks or explore the current programme.

Explore the 2026 agenda More WaysConf recordings

From the recording

Talk in brief

In his WaysConf 2023 presentation, Phil Hawksworth addresses the complexities of web rendering and content personalization, advocating for pragmatic architecture guided by the web's rule of least power. Rather than viewing personalization as a binary feature requiring heavy monolithic software, Hawksworth introduces the personalization spectrum—ranging from universal global content and localized regional experiences to audience segmentation and fully individualized views. He maps these levels to appropriate rendering strategies, comparing static site generation, client-side rendering, server-side rendering, edge-side rendering, and incremental methods like distributed persistent rendering and incremental static regeneration. Through a practical experiment archiving 24,000 social media posts, Hawksworth demonstrates how combining static pre-generation, on-demand edge builders, and edge search functions delivers high performance and fault tolerance. Ultimately, he stresses that teams must clarify content availability and business requirements before choosing technical stacks.

Key takeaways

01
Personalisation Operates on a Spectrum
Personalization is not an all-or-nothing requirement but spans from universal content to localized, segmented, and fully individual experiences. Teams should implement only the level of personalization that their available data and content strategy can support.
Watch from 7:20
02
Adhere to the Rule of Least Power
Developers should choose the simplest architecture and rendering method capable of satisfying user requirements. Starting with static assets and layering on edge functions or dynamic scripts minimizes infrastructure complexity and runtime risk.
Watch from 4:49
03
Server-Side Does Not Mean Server-Full
Server-side rendering no longer requires maintaining dedicated physical or virtual origin servers. Modern edge runtimes and serverless functions allow request-time computation close to users on content delivery networks.
Watch from 20:24
04
Separate Content Data from View Templates
Abstracting content into structured data decoupled from presentation layers enables flexible rendering choices across build time, request time, and on-demand execution. This separation allows single data sources to power static pages, edge-rendered searches, and dynamic routes.
Watch from 32:13
05
Let Content Availability Dictate Stack Choices
Advanced personalization frameworks are useless without the underlying content variants and user insights required to power them. Technical architectures must be selected only after establishing what content actually exists and needs to be delivered.
Watch from 44:40

Video chapters

2:24Defining Rendering and Web Architecture Evolution
Phil Hawksworth traces how web rendering evolved from simple static HTML serving to complex client-side scripts and modern multi-runtime architectures.
7:00Introducing the Personalisation Spectrum
Hawksworth outlines the personalization spectrum across four stages: universal global content, localized experiences, segmented user cohorts, and fully individualized views.
13:41Comparing Rendering Strategies Across Time and Location
The speaker breaks down static site generation, client-side rendering, server-side rendering, and edge-side rendering based on where and when computation occurs.
21:03Demystifying Incremental and Deferred Rendering
Hawksworth clarifies incremental acronyms such as distributed persistent rendering, on-demand builders, deferred static generation, and incremental static regeneration.
26:20Demonstrating Multi-Method Rendering with a Personal Archive
A case study archiving 24,000 tweets shows how static pre-generation, on-demand building, and edge functions combine into a single performant site.
32:49Mapping Rendering Techniques to Personalisation Levels
Hawksworth explains how to select appropriate technical strategies for localized pages, cookie-based audience segmentation, and authenticated user views.
38:29Managing Architecture Choices and Project Expectations
The session concludes with key architectural guidelines and an interactive Q&A discussing cache management, client expectations, and future web stack trends.

Read edited transcript highlights

These concise notes were edited from automatic captions and checked against the talk structure. They are not a verbatim transcript.

Categorizing Personalisation Requirements

When enterprise clients request a personalized website, they often express an ambition rather than a clear technical requirement. True personalization exists on a broad spectrum defined by technology capabilities, available user data, and content readiness. On the far left sits universal global content, which is the easiest to produce and maintain. Moving right introduces localized content for specific regions, segmented experiences targeting defined user cohorts through cookies or arrival times, and finally fully individualized views based on user profiles. Building further to the right increases system complexity and content demands, making it critical to identify the minimum level of personalization necessary.

Watch from 7:15

Evaluating Modern Web Rendering Models

Rendering techniques should be evaluated not only by where execution takes place, such as on the server, client, or network edge, but also by when it happens. Static site generation performs computation at build time long before a user arrives, delivering optimal speed and server simplicity. Client-side rendering handles rendering in the browser using JavaScript after downloading assets, which offers high interactivity but carries performance risks if overused. Meanwhile, server-side rendering and edge-side rendering compute responses at request time, with edge execution running serverless scripts directly on content delivery networks close to end users.

Watch from 13:49

Clarifying Incremental Rendering Terminology

Incremental rendering approaches solve the problem of long build times on massive websites by deferring page generation until requested. Distributed persistent rendering, deferred static generation, and on-demand builders share a core concept: non-essential pages are excluded from the initial build and compiled via serverless functions upon the first user request, after which the resulting static HTML is persisted for future visitors. In contrast, incremental static regeneration serves a pre-rendered scaffold while asynchronously revalidating and updating cached content in the background following user interactions.

Watch from 21:19

Hybrid Architecture for Large Scale Content

A real-world experiment archiving 24,000 individual social media posts illustrates how multiple rendering strategies can be combined into a coherent architecture. By decoupling content into a clean structured JSON file, static site generation builds the primary site pages and index lists in under two seconds. Distributed persistent rendering handles individual post URLs on demand through serverless builders, persisting generated HTML as pages are accessed. Additionally, edge-side functions perform real-time string matching against the dataset to serve instant search results, completely avoiding heavy client-side JavaScript execution.

Watch from 26:41

Pragmatic Frameworks for Personalised Content

Technical solutions for personalized experiences should be layered incrementally based on requirements. Universal global content can be pre-rendered entirely at build time, while localized versions can use static builds combined with location-based CDN routing. For segmented experiences, such as displaying onboarding tutorials only to first-time visitors, edge functions can evaluate client cookies and conditionally modify or filter static views. Fully personalized views, such as authenticated user dashboards, combine static shell layouts with client-side rendering to fetch private user data securely from backend application programming interfaces.

Watch from 32:59

Explore WaysConf 2026

View the agenda

Get tickets