Delivering page speed improvement via server-side tagging for a Multinational Retailer
.jpg)
Client Situation
The time it takes for a page to load is critical – especially for online retailers. Studies (e.g. here and here) have consistently shown poor performance directly lowers conversion rate. Aware that their website speed could be improved, our Client asked us to identify and quickly deliver improved performance.
Loop Horizon identified that optimising the Client's tag implementation was a fast way of driving performance improvement, with limited dependencies that could slow the project down. Critically, full site capability needed to be maintained and there could be no detrimental side effects to adjacent performance metrics.
Loop Contribution
Following an initial review, Loop identified three areas for improvement:
JavaScript optimisation
Loop Horizon identified and removed several redundant tags (including the soon to be deprecated Google Universal Analytics) and cleaning up data-processing. Loop Horizon optimised scripts that had their own bespoke data transformations, aligning them all to one single data model. This significantly reduced the amount of data-processing needed to deliver the same user outcomes.
Server side migration
Loop Horizon also took the Adobe suite of tools and moved them to Adobe's server-side instance, via Adobe’s WebSDK. Adobe Analytics and Adobe Target were moved to a combined single JavaScript library and data stream.
Content providing script optimisation
Loop Horizon reviewed content providing scripts, identifying opportunity to move selected items higher in the load-order, without detrimental impact to functional site performance.
Client Results
Loop Horizon delivered up to a 5.6-second page-load improvement for all pages across the Client website and removed up to 512 kilobytes from the page.
Content providing scripts were moved earlier in the load order to deliver the end-user experience more quickly. Combined with the reduction in JavaScript, this delivered the improvement in LCP (Largest Contentful Paint). The Client had tracked LCP historically, and on the day of the roll-out, there was a persistent step-change in LCP of 11%. Application of statistical analysis confirms a lower, but more robust, improvement of 6%. This is further supported by the improvement in Google’s Lighthouse metrics for LCP moving from ‘Needs Improvement’ to ‘Good’.
A direct A/B deployment wasn’t feasible for this project so P&L impact needed to be inferred. Comparable studies demonstrated improvements on this scale would translate to multi-million pound revenue upside for our client, providing extraordinary returns on an 8 week engagement
.png)
.png)
