The catalogue filter, as a discovery interface, has been the dominant UX pattern for retail commerce for sufficiently long that its dominance is rarely questioned in the procurement conversations I am part of. The pattern is canonical; the implementations are widely available; the platforms that retailers operate on typically present the filter interface as a default that requires explicit effort to remove rather than as a choice that requires explicit justification to include. The position is reasonable as a starting point; it is also, on the behavioural data I have collected across more than one hundred retail sites, decisively wrong for a substantial fraction of the audiences the filter is being deployed against.
The fraction in question is, in approximate terms, the under-25 cohort. The cohort’s behavioural pattern on retail catalogues differs from the older cohorts in ways that are sufficiently large to have specific implications for the design of the discovery interface; the implications, in my observation, are rarely incorporated into the procurement decisions that determine the interface’s design. This post is an account of the divergence, the specific data that supports the analysis, and the design recommendations that follow.
The behavioural divergence in some specific terms
The data on which the analysis rests was collected across the sites of a UK retail portfolio operating in fashion, beauty, homewares, garden equipment, and footwear, over the period of approximately eighteen months from January 2023 through June 2024. The behavioural variables included the channel of catalogue exploration (search interface, category browsing, filter use), the depth of exploration measured in product-page views per session, and the conversion outcome of each session. The demographic variables included age band, with the under-25 cohort identified through a combination of self-reported data, account age, and demographic modelling supplied by the sites’ analytics platforms.
The dominant pattern, replicated across the categories with very few exceptions, was that visitors under 25 accomplished the great majority of their catalogue exploration via the site’s search interface. The proportion of search-driven exploration in this cohort exceeded sixty per cent across the categories examined; the corresponding proportion of filter-driven exploration was below fifteen per cent. For visitors over 45, the pattern was approximately the inverse: search-driven exploration accounted for less than twenty per cent of the cohort’s catalogue interactions, while filter-driven exploration accounted for more than fifty per cent.
The mid-range bands (25 to 44) showed a more even distribution between the two modes, with the cross-over occurring approximately at age 35. The cross-over is not, on the data, an abrupt transition; the proportion of search-driven exploration declined approximately monotonically with age, with no identifiable discontinuity at any specific age. The implication is that the under-25 cohort is not a homogenous population using a different mode of exploration; rather, the cohort sits at one extreme of a continuous distribution, with the older cohorts at the other.
Why the divergence exists
The mechanism by which the divergence emerges is, on the qualitative interview data conducted alongside the behavioural analysis, approximately the following. The under-25 cohort has, by virtue of their entire pre-retail-relevant lifetimes having occurred in a search-dominated digital environment, internalised search as the default mode of information access. The cohort’s expectation, when arriving on a retail catalogue, is that the catalogue will be queryable via a search box that will resolve their natural-language description of what they are looking for; the absence of a competent search interface is, for this cohort, experienced as a failure of the catalogue rather than as an alternative information architecture that the cohort should adapt to.
The over-45 cohort has, by contrast, more direct experience of pre-search information architectures — the printed catalogue, the department store’s physical aisle structure, the early web’s directory-based navigation — and the cohort’s mental model of catalogue exploration is, accordingly, more category-based than search-based. The filter interface is, for this cohort, a digital analogue of the physical category structure that the cohort already understands, and the cohort uses it competently and willingly. The search interface, for this cohort, is treated as an alternative for cases where the category structure has failed rather than as the primary mode of access.
The mechanism is not, in any abstract sense, surprising; the consequence for retail commerce is, however, considerably more direct than the abstract observation would suggest. The retailer whose catalogue’s primary discovery interface is the filter is, for the under-25 cohort, presenting an interface that the cohort does not, in any meaningful sense, use. The cohort either bounces from the catalogue or, in the better case, locates the search interface and uses it; in either case, the filter’s substantial visual real estate has produced no contribution to the cohort’s experience of the catalogue.
The design implication
The implication for the retailer’s catalogue design depends, decisively, on the demographic composition of the audience the catalogue is serving. For a retailer whose audience is concentrated in the under-25 cohort — a fashion brand targeting Gen Z, a beauty brand whose social acquisition is concentrated on TikTok, a footwear brand operating in the streetwear category — the filter interface is, in operational terms, a feature that the audience is not using and that the catalogue’s interface design is allocating disproportionate visual weight to. The recommendation is to invert the typical visual hierarchy: the search interface should be prominent, persistent, and immediately accessible; the filter interface should be available but visually demoted. (For more on the broader behavioural patterns this analysis is part of, see The Demographic Behaviour project.)
For a retailer whose audience is concentrated in the over-45 cohort, the implication is the inverse. The filter interface remains the appropriate primary discovery surface; the search interface should be available but visually demoted. The recommendation is, in this case, the conventional one; the data simply confirms that the convention is appropriate for the audience in question.
For a retailer whose audience is mixed — which includes, in practice, the great majority of UK retail catalogues — the implication is more nuanced. The recommendation that has emerged from the analysis is to present both interfaces with broadly equivalent visual prominence and to allow the visitor’s first interaction to determine which interface dominates the subsequent experience. The pattern requires more design work than the single-mode interfaces, and is not, in my observation, available as an off-the-shelf component on the dominant retail platforms; the additional work is, on the conversion data, recovered comfortably for retailers whose audience is genuinely mixed across the age distribution.
The implementation specifics that matter
The implementation of the search interface, when it is being used as a primary discovery surface for the under-25 cohort, is consequential in ways that the standard search implementations frequently fail to address.
The first is the search interface’s tolerance for natural-language queries. The under-25 cohort’s queries are, on the query log data I have access to, more frequently natural-language phrasings (“waterproof boots for festivals”, “skincare for combination skin”, “duvet for hot sleepers”) than they are product names or model numbers. The search implementation must therefore handle the natural-language queries competently; a search that returns no results for a natural-language query is, for the cohort, a failure of the catalogue. (For the engineering work that this requires, see AI search is overhyped: auto-complete is underbuilt.)
The second is the search interface’s responsiveness. The under-25 cohort’s expectation, set by the platforms the cohort uses outside retail, is that the search interface will return results as the visitor types rather than after the visitor submits a complete query. A search that requires a submission step is, for the cohort, slower than the cohort expects, and the slower search produces measurably higher abandonment rates than the type-ahead search.
The third is the search interface’s typo tolerance. The under-25 cohort’s queries are typed quickly, on mobile devices, with frequent typographical errors that the cohort does not stop to correct; the search must therefore tolerate typographical errors competently or lose the query. The off-the-shelf search implementations on most retail platforms are, in this respect, considerably less competent than the cohort expects; the gap between expectation and capability is decisive for whether the search interface produces useful conversion outcomes.
An advisory close
The catalogue’s discovery interface is, for an audience that is not homogenous in age, one of the most consequential design decisions the retailer makes; the decision is, in the procurement conversations I have, frequently made on the basis of the platform’s defaults rather than on the basis of the audience’s actual behavioural patterns. The defaults are appropriate for some audiences and inappropriate for others; the retailer’s first task is to determine which case applies to the catalogue in question.
It is recommended that any retailer auditing the catalogue’s discovery interface do so with reference to the demographic composition of the audience the existing analytics reveal, rather than with reference to the audience the retailer imagines is being served. The two are, in my experience, frequently different, and the difference is decisive for whether the discovery interface is calibrated to the audience that is actually visiting the site.