Published: December 9, 2025
Monrif is one of Italy's leading publishing groups, operating high-traffic news brands such as Il Resto del Carlino, Il Giorno, La Nazione, and QN Quotidiano.net. With millions of users accessing dozens of articles per session, optimizing navigation speed and page load time is essential for reader retention and engagement.
In early 2025, Monrif began a performance initiative focused on improving Largest Contentful Paint (LCP) by adopting an effective prerendering strategy with the Speculation Rules API. By identifying the most-visited article pages and applying prerendering for those paths, Monrif achieved up to a 17.9% improvement in LCP—a measurable improvement in performance and user-friendly news experiences. Furthermore, this translated into a notable uplift in user engagement, with peaks of +8.9% in some segments.
Encouraged by the success of prerendering, the team also implemented support for the back/forward cache (bfcache). This allowed returning visitors to experience near-instant load times when navigating back to previously visited pages, further contributing to a seamless and responsive browsing experience.
Where we started
Before the rollout, internal testing and Chrome User Experience Report (CrUX) data showed that many Monrif pages—especially article and homepage views—had suboptimal LCP scores.
Before optimization efforts began prior to March 2025, the 75th percentile LCP for homepage views across several titles ranged from 2.333 seconds to 2.419 seconds. While technically within the "good" threshold (2.5 seconds or less), there was still headroom for improving user-perceived performance during high-frequency navigation patterns.
How we improved LCP
To reduce latency during navigations, Monrif implemented speculative prerendering on all desktop and mobile domains using the Speculation Rules API. The strategy was designed to balance performance impact with technical safety, allowing prerendering only when link elements matched specific selectors and conditions.
{
"prerender": [
{
"where": {
"and": [
{
"not": {
"selector_matches": "[rel~=nofollow]"
}
},
{
"not": {
"selector_matches": ".no-prerender"
}
},
{
"selector_matches": "a"
}
]
},
"eagerness": "moderate"
}
]
}
The eagerness level was set to moderate for any link (with certain exclusions). This setting speculates on hover (on desktop) and based viewport-heuristics (on mobile). This approach allowed the team to control prerendering scope to just the links most likely needed and so minimizing resource usage.
The results
To evaluate the impact of prerendering, we focused our analysis on three major Monrif publications—Il Resto del Carlino, Il Giorno and La Nazione—using GA4 data and CrUX metrics.
Although prerendering was enabled on both desktop and mobile, incidence was significantly higher on desktop, since "moderate" allowed all more links to be in-scope when hovered. For example, on La Nazione, 13.9% of web (desktop) views triggered prerendering, compared to only 2.9% on mobile. As a result, performance improvements were more visible and consistent on desktop platforms, which became the main scope of this study.
By late May 2025, LCP performance had improved significantly across all three publications, as seen in CrUX desktop data:
These improvements closely mirrored the peaks in prerender activity. The following table shows the highest prerender occurrences—both by volume and by percentage increase over their average—for each site on desktop:
Across all titles, the lowest LCP values were observed within greatly overlapping timeframes as the highest prerendering events, particularly between mid-May and early June 2025. This synchronicity suggests a strong correlation between speculative prerendering and real-world performance gains.
The engagement effect
In parallel with LCP gains, Monrif recorded a measurable increase in user engagement during the prerendering rollout. Analyzing data from GA4 across Il Resto del Carlino, Il Giorno, and La Nazione, engagement rate improved significantly in weeks where prerendering was most active.
Here's a breakdown of the correlation between prerender intensity, LCP improvement, and engagement uplift on desktop:
While multiple factors influence engagement, the close temporal correlation suggests that better-perceived performance with prerendering leads to deeper, more sustained user interaction.
Advertising impact
While Monrif's primary goal was to improve user experience and performance, the initiative also appears to have had positive effects on advertising metrics.
In particular, viewability rates in the open market—a key indicator of ad visibility and inventory quality—showed a modest but consistent improvement during the prerendering rollout. Across the three core publications, the average viewability improved by 1.03%, closely following the weeks of highest prerendering activity. As with the performance and engagement analysis, these figures focus on desktop traffic to ensure consistency across datasets.
This improvement, though small in absolute terms, is meaningful: Even marginal gains in viewability can enhance the perceived quality of inventory, increase fill rates, and improve advertiser trust.
What's particularly notable is how this ties back to the performance enhancements. By reducing LCP and improving navigation speed, Monrif helped pages load faster and kept users more engaged. This in turn increases the likelihood that ads remain in the viewport longer, and are more likely to be seen.
Improving instant navigation with bfcache
Along with prerendering, Monrif worked to unlock support for the browser's back/forward cache (bfcache)—a performance mechanism that enables instant page restores when users navigate using the browser's back or forward buttons.
To ensure compatibility with bfcache, developers addressed several common blockers:
- Removed
unloadevent listeners as these are unreliable and soon to be deprecated. - Adopted a strict
Permissions-Policyheader to prevent reintroduction ofunloadevent listeners (particularly by third parties). - Implemented
pagehideevent listeners to safely close connections (for example, IndexedDB, WebSocket). - Avoided the use of
Cache-Control: no-storeunless strictly necessary, since that can block bfcache in many cases.
While the team did not yet have sufficient data to isolate bfcache impact with the same statistical rigor used for prerendering, early testing and Chrome DevTools traces confirmed that compatible navigations were fully restored from memory—delivering sub-second load times when navigating back to previously viewed articles.
Conclusion
By integrating speculative prerendering, Monrif significantly improved LCP on its news websites—achieving up to a 17.9% reduction in load time. These enhancements not only improved perceived performance but also translated into measurable gains in user engagement across its core titles.
Furthermore, this case suggests that user-centric performance optimizations can generate downstream monetization benefits, by improving conditions that affect how and when ads are seen.
This case study demonstrates how modern navigation optimizations like prerendering can enhance Core Web Vitals at scale—even on content-rich platforms—and translate into measurable improvements in user experience and business outcomes.