Between 2022 and 2025 I built four social-platform widgets for retail clients, in approximately the following order: Facebook, Instagram, Pinterest, and most recently TikTok. The widgets were superficially similar — each pulled a feed of social content from the platform’s public API, displayed the content in a grid on the retailer’s site, and offered the visitor a route from a social post to a corresponding product page on the retailer’s catalogue. The widgets were also, on the conversion data, dramatically different in their commercial outcomes; the TikTok widget produced approximately seven times the conversion uplift of the Pinterest widget and approximately fifteen times the conversion uplift of the Facebook widget, against deployments of broadly comparable scale and visibility.
This post is an account of what the differences in conversion outcomes had in common, what they had to do with the structural properties of each platform’s discovery loop, and what the differences imply for retailers considering which social integrations to invest in. The TikTok widget is the one in the title because it is the one that produced the strongest commercial result, but the lessons are, in some sense, lessons about the limits of social commerce more broadly rather than about TikTok specifically.
What the four platforms have in common, structurally
Each of the four platforms operates a feed-based content delivery model in which a stream of posts is presented to the user, and the user is permitted to interact with each post via a small set of fixed actions (like, comment, share, save). Each of the four platforms also operates an advertising and creator-economy infrastructure that incentivises the creation of content by parties other than the platform itself. Each of the four, in some form, permits the surfacing of commercial content alongside organic content, with varying degrees of explicitness and visitor consent.
The structural similarities are sufficient that the technical implementation of the four widgets was, in essentially every respect, the same. Each widget consumed a public API endpoint to fetch the most recent posts from a specified account or hashtag; each parsed the response into a normalised data structure; each rendered the data into a grid on the retailer’s site with a small set of visual customisations to fit the host design. The implementation cost of each widget was, after the first, considerably lower than the previous one; the engineering work was largely a matter of adapting the previous widget to the new platform’s API conventions.
Where they diverge — the discovery loop
The structural difference that, on the conversion data, accounts for almost all of the variation in commercial outcome is the structure of the discovery loop on each platform. The discovery loop is the mechanism by which a user on the platform encounters new content from accounts they do not already follow; the structure of the loop determines, more than any other variable I have been able to identify, whether content posted to the platform produces visitor traffic that converts on the retailer’s site.
Facebook’s discovery loop is, in 2026, essentially closed. Content is delivered to users predominantly from the accounts they already follow; new accounts are surfaced primarily through paid advertising, friend suggestions, or direct linkage from external sites. A retailer’s organic content on Facebook is, in practical terms, only seen by the retailer’s existing followers, and the existing followers are by definition already aware of the retailer. The Facebook widget, as a consequence, served principally as a way of displaying content the visitor had likely already seen, to a visitor who had likely already engaged with the retailer in some other way. The conversion uplift was real but small, and was concentrated almost entirely in the visitors who arrived on the site without prior knowledge of the retailer’s social presence.
Instagram’s discovery loop is more open than Facebook’s, principally via the Explore feed and the hashtag-based search; users do, in non-trivial numbers, encounter new accounts through the platform. The retailer’s content is therefore seen by a wider population than Facebook’s, including a meaningful fraction of users who are not yet customers. The Instagram widget, on the conversion data, performed measurably better than the Facebook widget; the uplift was approximately three times that of the Facebook widget against comparable visibility on the host site.
Pinterest’s discovery loop is more open again, and is structured around saved boards rather than feeds. Users encounter content largely on the basis of their previously stated interests, which the platform infers from the content they have saved; the platform’s commercial behaviour is, in 2026, considerably more aligned with retail discovery than the other platforms, and the Pinterest widget performed correspondingly better. The conversion uplift on the deployments I shipped was approximately twice that of the Instagram widget and approximately six times that of the Facebook widget.
TikTok’s discovery loop is structurally different from all three. The dominant content surface is the For You feed, which is populated almost entirely from accounts the user does not follow; the recommendation algorithm is sufficiently aggressive in surfacing new content that the user’s experience of the platform is, more than the other three, an experience of perpetual discovery rather than of consuming content from a fixed set of accounts. The implication for retail content is decisive. A retailer who posts a single video that performs well on the TikTok algorithm reaches a population that is, in expectation, several orders of magnitude larger than the equivalent population on Facebook; the visitor traffic produced from that reach, when surfaced via the widget on the retailer’s site, converts at rates that no other platform’s widget approaches.
The video format itself, and what it implies
The conversion uplift from the TikTok widget is not, in my analysis, attributable to the discovery loop alone; the video format itself plays a non-trivial role. A photograph, of the kind that dominates Instagram and Pinterest, communicates a single moment of the product; a video communicates a sequence of moments and, if produced competently, a demonstration of the product’s use in a context the visitor can recognise as their own. The visitor’s confidence in the product’s suitability for their use, after seeing a competent video, is on the conversion data measurably higher than the confidence after seeing a competent photograph; the difference compounds with the discovery-loop advantage to produce the conversion gap that the data shows.
The video format imposes a corresponding cost on the retailer’s content production. A single TikTok video that performs well requires considerably more production effort than a single Instagram photograph; the retailer’s content team must produce the videos at a sufficient cadence to maintain the platform’s algorithmic favour, and the production cost is, in the deployments I shipped, the principal limit on the volume of content the retailer was able to sustain. (For more on the relationship between content production and conversion outcomes, see UGC Scaler: industrialising user-generated video.)
What the widget does not solve
The widget surfaces existing social content on the retailer’s site; it does not, on its own, address the question of what content the retailer should produce, the question of how to produce it at the cadence the platforms require, or the question of how to convert the social-platform attention into a transactional relationship that survives the visitor’s eventual disengagement from the platform. These questions are, in my observation, the questions that determine whether the retailer’s social investment produces a long-run commercial outcome; the widget contributes to the conversion of platform attention into site visits but does not, on its own, address the broader strategic problem.
An advisory close
The procurement decision that retailers most often face, when the social-commerce discussion comes up, is which of the major platforms to invest in. The dominant procurement narrative is that the retailer should be present on all of them; the conversion data, on the deployments I have shipped, suggests instead that the retailer should be present on the platforms whose discovery loops are structurally favourable to retail and should treat presence on the others as an optional supplement rather than a requirement. For most UK retail catalogues in 2026, the discovery-loop ranking is, in my observation, TikTok first, Pinterest second, Instagram third and Facebook a distant fourth; the procurement budget, where it must be allocated, should reflect that ranking rather than the equal-weight distribution that the agency proposals tend to default to.
It is recommended, finally, that the widget itself be treated as a component of the broader social investment rather than as the investment itself. The widget surfaces the content the retailer is producing; the production of the content remains the larger and more consequential half of the work. (For a related view on the demographic distribution of social-platform engagement, see Why Gen Z bounces off your filters and millennials don’t.)