Speculation rules tested across three e-commerce sites: results and insights

At Iron/Out, we regularly partner with e-commerce brands to experiment with performance techniques that have the potential to influence both user experience and business outcomes. In this recent series of A/B tests, we assessed speculation rules as a method for improving site speed and engagement. Our focus was on identifying whether faster perceived performance would lead to deeper product exploration and ultimately influence conversion.

What are speculation rules? #

Speculation rules are a browser-native mechanism that enables preloading or prerendering of pages a user is likely to visit next. In this test, we applied them in two ways:

• On desktop: prerendering was triggered when a user hovered over a product tile.
• On mobile: the first four visible product tiles on a listing page were prerendered automatically.

The assumption was that by preloading high-probability next steps, users would experience faster transitions into product detail pages (PDPs), improving perceived performance and reducing friction.

Hypothesis and goals #

Inspired by Google’s Ray-Ban case study, which reported significant improvements in navigation speed through speculative preloading, we wanted to test whether the same approach could influence behaviour deeper in the funnel.

We monitored:

• Largest Contentful Paint (LCP), a key Core Web Vitals metric.
• PLP and PDP bounce and exit rates
• PDP view volume and frequency
• Funnel progression: Add to Bag, Checkout, Purchases
• Revenue per visitor (RPV) and Average Order Value (AOV)

Test design #

Three retail brands ran the test for 23 to 24 days each, across both desktop and mobile. Metrics were segmented by device and analysed for statistical significance.

Results overview #

1. Engagement improved
Users viewed more PDPs, and session depth increased across all brands. Bounce and exit rates dropped on both listing and detail pages. These patterns suggest speculative preloading did reduce user friction during browsing.

2. LCP impact was minimal
Some variants saw slight reductions in LCP, but no change was statistically significant. One test showed no improvement at all. While speculation rules clearly executed, the performance delta was marginal.

3. No movement in conversion behaviour
Despite the engagement improvements, Add to Bag, Basket visits, and Checkouts remained flat. Purchase rates showed no uplift. This indicates that while users browsed more, their intent to purchase did not increase.

4. Revenue and AOV trended down
Two of the three tests showed directional declines in RPV and AOV, particularly on desktop. These changes weren’t statistically significant, but were consistent enough to be noteworthy. Mobile metrics remained more stable.

An important consideration: were the sites already fast enough? #

All three tested sites already had LCP timings well within Google's Core Web Vitals thresholds. From a user experience standpoint, pages were already loading quickly well under the point where latency typically starts to impact behaviour.

This raises a valid question: when a site is already performing at a high level, are further speed improvements still likely to drive measurable conversion gains? It’s possible that these brands were already within the performance “sweet spot” where load time is no longer a barrier to action. If users aren't being slowed down, making things slightly faster may not change what they do.

For teams considering speculation rules, this highlights the importance of performance context. Sites that are struggling to meet CWV thresholds might see more tangible gains. But for high-performing platforms, incremental speed improvements may yield diminishing returns particularly if other conversion drivers (such as content, value proposition, or UX) are not addressed in parallel.

Final assessment #

Each experiment was classified as a learner. While speculative preloading improved engagement and user flow, it did not impact key commercial metrics. The tests offered clear insight into where the limits of performance optimisation might lie when conversion intent is already determined by other factors.

Key takeaways #

Performance supports, but doesn’t guarantee, conversion
Speed improvements must be paired with persuasive UX, clear product value, and effective merchandising to realise the business impact of good website performance. Faster journeys alone don’t create intent.

Deeper engagement doesn’t always mean higher intent
Increased PDP views may reflect easier navigation, but not necessarily stronger interest in purchasing. This distinction matters when interpreting success.

Desktop impact was more variable
Negative trends were more evident on desktop, possibly due to more frequent hover events triggering unnecessary prerendering. Mobile was more stable overall.

Contextual optimisation is essential
Where speculation rules are applied matters. Future testing could prioritise slower templates, edge case journeys, or tailored predictive rules rather than default triggers.

Conclusion #

Speculation rules are a smart technique to explore for enhancing perceived speed, especially in high-traffic browsing flows. But like all performance tools, their value is contextual. When applied to sites already performing well, improvements may be technically sound but commercially neutral.

For web performance specialists, SEOs, CROs, and developers, the key takeaway is this: always assess the performance ceiling before aiming to raise it. If your site is already fast, the next gains may need to come from experience, not milliseconds.

For more on the evolution of performance practices, see how the industry is shifting from developer-first to user-centric metrics.