Case study

AI-Powered Content Recommendations for the Engineering Community

<1s

API latency for recommendations

1.3M+

community members impacted

real-time

content personalisation

The upshot

DesignSpark is a massive community hub for engineers, but with thousands of articles, users struggled to find the right technical content. Datasparq built a real-time recommendation engine and a semantic "metadata factory" on Google Cloud to increase community engagement and drive product discovery.

AI and data solutions development partner

The opportunity

DesignSpark is a content-rich site, but high volume often leads to "search fatigue." RS Group needed a way to help users discover new content effectively, which in turn drives users to learn, interact, and ultimately purchase components from the main RS site.

The solution

Datasparq designed a bespoke recommendation engine built entirely on Google Cloud Platform (GCP). Unlike a static "Related Articles" plugin, this system uses various algorithms to provide article recommendations based on an individual's unique reading history and the interests of their peers.

We integrated Google Analytics 4 (GA4) with BigQuery to capture live user activity. By streaming behavioural data into a centralised warehouse, we created a "digital breadcrumb trail" of what engineers were researching, from PCB design to industrial automation.

To recommend content, the system first had to "understand" it. We built a data service that continuously analyses thousands of technical articles, extracting semantic meaning and creating an ontological model of the content library.

We applied topic extraction and classification models using a Postgres Cloud SQL instance. This allowed the system to group articles by complex engineering themes, ensuring that a user interested in "IoT sensors" wasn't recommended unrelated "3D printing" content.

We deployed the recommendation engine using Google Cloud Run. This enabled us to serve personalised "Next Read" suggestions via an API with sub-second latency, ensuring the user experience remained fast and fluid.

1. Proof of concept

First testing with a supermarket depot demonstrated capabilities beyond established industry tools, impressing transport leadership with performance that surpassed existing SaaS solutions.

2. First deployment

Applied to a major retailer's largest and most complex depot, exploring delivery types and time window flexing. This phase delivered substantial annual savings that continue to benefit operations today.

3. Extended capabilities

Extended to a more challenging network with multiple depots, handling unionised workforces, two-person deliveries and a wider variety of vehicle types. This phase demonstrated the optimiser's adaptability to different operational constraints.

4. Enterprise-ready

Standardised data formats and pipelines to create a system applicable across diverse customer networks, with further optimisation for a home improvement retailer adding wait time and compactification features that doubled performance.

Interactive example

Try it for yourself

Explore how the science works in this PlayML data science notebook

The impact

Launched in January 2022, the platform now engages over a million engineers daily. By surfacing the right technical insights at the right time, RS Group has improved community retention and created a direct path from "learning" to "buying" on the main ecommerce site.

AI-Powered Content Recommendations for the Engineering Community

1. Proof of concept

2. First deployment

3. Extended capabilities

4. Enterprise-ready

Try it for yourself

Explore more of our work

Reducing manufacturing waste and production time by ⅓ with data & AI

Producing a data & AI strategy to unlock £3.5 million in savings

Transforming retail property management through better data

Call us when you're ready