The homepage of a retail commerce site has, in the great majority of visitor sessions, approximately six and a half seconds in which to demonstrate that the visitor is in the right place. The figure is approximate; the underlying behavioural pattern from which it is derived is, in my experience of analysing visitor session recordings across more than a hundred retail sites, considerably more consistent than the variation in the precise number would suggest. The visitor arrives, scans the visible page, performs an internal assessment of whether the site is one they wish to engage with, and either commits to a deeper interaction or abandons. The six-and-a-half-second figure is, in a sense, a slightly arbitrary point in the distribution of dwell times before the abandonment decision; the underlying decision itself is, on the data, decisive for whether the visit produces any commercial outcome.
The implication for the design and engineering of the homepage is one that the trade literature has been articulating, in various forms, for some years. The implication is canonical; the specific behavioural data that supports it is, in my observation, less widely circulated than the recommendations that are derived from it. This post is an attempt to articulate the data with sufficient specificity to inform the design and engineering decisions that the recommendations imply.
What the visitor actually does in the first 6.5 seconds
The session-recording data, when examined at the level of individual visitor behaviour, shows a recognisable pattern that repeats across the great majority of sessions. The visitor’s mouse cursor, in the first second following the page load, is approximately stationary; the visitor’s attention is, on the eye-tracking data that has been collected for a smaller subset of the sessions, distributed across the upper half of the visible page in a pattern that approximates the well-documented F-shaped reading pattern of web pages more generally.
The cursor movement, when it begins, is in the second to third second of the session and is typically a small movement towards the upper-left or upper-centre of the visible page; the movement is exploratory rather than directed at any specific element. The visitor’s first identifiable directed action — a click, a scroll, a hover that lasts longer than the cursor’s general exploration — occurs, on the median session, between the fourth and seventh second; the precise timing is sensitive to the design of the page and to the visitor’s prior context, but the median is, with surprising consistency, in this range.
The decision to abandon is, on the same data, taken before the visitor’s first directed action in approximately twenty-eight per cent of the sessions that abandon. That is to say, of the visitors who do not engage further with the site, more than a quarter do so without ever interacting with the page beyond passive observation. The implication is that the page’s design must, in those cases, communicate sufficient signal to retain the visitor on the basis of a passive read alone; any signal that requires the visitor to interact in order to discover it is, in a non-trivial fraction of cases, never discovered.
What the visitor is assessing during the 6.5 seconds
The qualitative interview data, conducted alongside the behavioural analysis, suggests that the visitor’s internal assessment during the period covers approximately three questions. The questions are not articulated by the visitor in any explicit form; the visitor experiences the assessment as an immediate impression rather than as a sequence of considerations. The questions, however, are recoverable from the patterns of subsequent behaviour and are, in summary, the following.
The first question is approximately “Am I in the right place for what I am looking for?” The visitor arrived on the site for some reason — a search engine result, a social-media link, a marketing email, a direct URL entry — and the homepage either confirms that the site contains what the visitor expected to find or it does not. The principal failure mode of this question is the homepage that, by virtue of being too generic in its presentation, fails to confirm any specific expectation; the visitor cannot tell whether the site sells what they were looking for and abandons rather than discover.
The second question is approximately “Is this site credible enough to transact with?” The visitor’s assessment of credibility is rapid and is informed by a combination of visual signals (the design’s apparent professionalism), behavioural signals (the page’s loading behaviour, the absence of broken elements), and trust signals (recognisable brand or vendor relationships, customer testimonials, industry credentials). The principal failure mode here is the homepage that is fundamentally credible but presents itself in a way that triggers the visitor’s distrust signals — slow loading, layout shifts during the initial render, intrusive overlays — independently of the substantive credibility of the site itself.
The third question is approximately “Is the experience of being here pleasant enough to continue?” The visitor’s assessment of pleasantness is also rapid and is informed by a combination of factors that are more subjective than the credibility assessment but no less consequential. The factors include the visual quality of the design, the perceived speed of the interactions, the cognitive load imposed by the visible navigation, and the absence of immediate friction (intrusive ads, modal overlays, cookie banners that interrupt the visitor’s first interaction). The failure mode here is the homepage that is technically functional but visibly unpleasant to engage with.
What this implies for engineering
The engineering implications of the data are, in my view, considerably more demanding than the trade literature’s typical framing acknowledges. The implications are demanding because they require the homepage to perform on three substantively different dimensions — informational specificity, technical credibility, and experiential pleasantness — within a temporal window that is too short for any individual dimension to be addressed by a sequence of progressive disclosures.
The informational specificity dimension requires the homepage to communicate, in the visible above-the-fold area, sufficient signal about the site’s category and offering to confirm the visitor’s expectation. The implication is that the upper part of the homepage is not the place for general brand messaging; the brand messaging may be retained, but it must coexist with concrete signal about the actual catalogue. A homepage that uses the upper half exclusively for an aspirational brand statement, with the catalogue signal pushed below the fold, is a homepage whose informational specificity will be, for a non-trivial fraction of visitors, insufficient to retain them.
The technical credibility dimension requires that the homepage’s loading behaviour be substantially better than the typical retail-site baseline. Layout shifts during the initial render are, on the data, perhaps the most decisive of the technical signals; a homepage that visibly shifts during loading is interpreted by visitors as a signal of broader technical instability and produces measurable abandonment regardless of the credibility of the underlying brand. The cumulative-layout-shift score (CLS) is, in this sense, less of a technical metric than a trust metric, and its optimisation is correspondingly consequential.
The experiential pleasantness dimension requires the absence of immediate friction. The cookie banner, in particular, is a frequent contributor to the abandonment within the first six seconds; visitors who arrive on a page and are immediately presented with a modal blocking the underlying content are, on the data, considerably more likely to abandon than visitors whose first interaction is permitted to proceed without the modal. The implementation pattern that has, in my experience, addressed this is the cookie banner that appears as a non-blocking element at the bottom of the viewport, with the consent decision available but not required for the visitor’s initial exploration to proceed.
The performance work that supports the 6.5 seconds
The performance dimension of the assessment is, on the data, particularly consequential. A homepage that takes more than approximately four seconds to render its visible content has, by hypothesis, less than two and a half seconds of remaining time within the assessment window for the visitor to perform the substantive evaluation; the available time is, in many such cases, insufficient for the evaluation to complete and the visitor abandons before the page has communicated whatever it would have communicated had the rendering been faster.
The implication is that performance work on the homepage specifically — as distinct from performance work on the broader site — is one of the highest-leverage interventions available to a retailer whose homepage is performing below the assessment threshold. The interventions are the standard ones (lazy loading, deferred third-party scripts, server-side rendering of critical content); the homepage benefits from them disproportionately to the rest of the site because the temporal margin is tighter on the homepage than on subsequent pages, where the visitor has, in some sense, already committed to the session and is more tolerant of additional latency. (For a more detailed account of the performance work itself, see A Core Web Vitals diary.)
An advisory close
The homepage’s six and a half seconds are, in some sense, a window of compressed decision in which the visitor performs an assessment that is more demanding than the visitor’s conscious experience of the moment would suggest. The retailer’s task is not to make the assessment easier — the visitor’s assessment will be made regardless — but to ensure that the homepage’s substance and presentation are sufficient to pass the assessment when it is conducted. The retailers who treat the homepage as the page on which all of the conditions that determine the visit’s outcome are simultaneously decided are, in my experience, the retailers whose homepages produce the most reliable conversion outcomes.
It is recommended that any retailer auditing the homepage do so with the six-and-a-half-second window in mind, and with an explicit assessment of whether the homepage’s visible content is sufficient to communicate the substance of the site, the credibility of the brand, and the pleasantness of the experience within the window the visitor will actually grant it.