TikTok has rapidly evolved since its 2018 launch. The video-sharing app rose to prominence in Canada after taking over the user base and platform architecture from predecessor app Music.ly. The pandemic only accelerated TikTok’s popularity as bored, isolated youth turned to the app to pass the time.
As the previous app was largely used by preteens and younger teens as a creative tool enabling lip syncs and advanced editing techniques with little technical knowledge required (Savic, 2021), this is reflected in its current usage nationally. As of 2020, “the number of online Canadians on TikTok is relatively small (15%), those who do use the platform visit it regularly (63% daily)” (Gruzd & Mai, 2020, p. 4) which places its daily use above that of Snapchat, Twitter, Pinterest, Reddit and LinkedIn. Moreover, “Adoption largely skews towards younger age groups, as 55% of those aged 18–24 have an account on the platform” (Gruzd & Mai, 2020, p. 16). In other words, TikTok is a popular app amongst young Canadians with high instances of daily use. Its popularity is growing with older demographics, but remains a stalwart for young users.
TikTok is known for its short, looping videos of various genres. On their official site, much focus is placed on expressing creativity and using unique effects to create entertaining and inspiring content. In their blog posts, TikTok emphasizes that content that does well on the platform has the following:
- potential to appeal to a broad audience of various age groups
- appears genuine (i.e. not attempting to use shortcuts to visibility such as spamming hashtags),
- optimized for TikTok’s formal vertical video format, and exceeding 5 seconds in length (5 tips for creators, 2019).
This means that TikTok has carved out a reputation for showcasing dynamic, surprising videos that have the potential to inspire imitation.
Revenue Streams
TikTok earned an estimated one billion USD in 2020 (Wang et al., 2020). Revenue relies t on both venture capital funding and advertising (Zhang, 2021), advertisements are both stylistically characteristic of TikTok (using trending sounds, featuring popular creators) on the For You page which appear between user-generated content or through sponsored hashtags and filters which companies pay to have placed on the now-defunct Discover page, which users are encouraged to incorporate in their content.
TikTok has branched into its own currency, Diamonds, as a new revenue stream. Starting in 2021 users could convert currency into coins which can be used to purchase “Diamonds” to award to creators for both Live Streams and short videos. Importantly, coins have no monetary value per the Virtual Items Policy in the Terms of Service and can only be purchased by users above the age of majority in the country they reside. This new form of in-app currency is quickly becoming an important revenue stream for TikTok as the platform may keep over 50% of the monetary value of the gift purchase before they are sent to creators (Tidy, 2019; Lucas, 2022). Unfortunately, there is little information to be found on the extent of this revenue stream from official sources, much less TikTok themselves.
Screens
Personalized recommendations
The For You page (FYP), which is the landing page when a user opens the app, is undoubtedly the most important screen in the app. This is an endless scrolling algorithmic feed which serves one video at a time to the user (Fig. 1; Fig. 2). What sets TikTok apart from other apps as it mixes together:
- videos from creators one follows,
- trending videos, videos from the locality the user is in; and, novel content which might lie outside the user’s preferences but allows them to discover new genres.
TikTok positions its recommendations as entertainment media rather than social media. While a sister feed, known as the “Following” feed (Fig. 3), shows videos solely from creators that a user follows, is also found on the home page, very little focus is placed on this screen in the official literature despite the apparent algorithmic curation at work on the feed.
Since TikTok operates on the creative reuse of audio tracks in videos, thus earning its reputation as a lip-sync and dance app, the sounds page is another key screen for discoverability on the platform (see figures above). This page aggregates all videos using a given sound, however, there is no official information on how the videos are organized on the screen. They appear to be sorted by popularity based upon total views, though this cannot be confirmed in any official capacity. Importantly, the sounds page also identifies the origin of the sound by placing the video marked “original” in the top left of the video grid. A similar page exists for visual effects with identical layout, however without the “original” designation (see figures below).
Trending or popular recommendations
The Discover page, where the search bar is found, is a key reflection of the current trends occurring on the app but is woefully under-reported in the official blog. The layout resembles a vertical scrollable list, featuring videos laid out horizontally (Fig. 8). At the top of the screen, above the trending topics, is a banner carousel which features events, programming and other content TikTok wishes to highlight. The Discover feed itself appears to be a mix of algorithmically and human curated content, updating daily to reflect changing popularity metrics. Per an official representative, the purpose of the Discover page is to highlight trending videos, hashtags, effects or sounds within the app. Similarly, the banners on the carousel are programmed by TikTok and can change daily. These banners typically highlight and promote things like scheduled live streams, trending hashtags, and, occasionally, sponsored brand campaigns. They also used to highlight events and commemorations, such as National Indigenous History Month, Pride Month, and Black History Month. Lastly, these banners have also been used to promote public service announcements.
On the For You page the option to like, comment, share (with multiple embedded options such as in-app direct message, iMessage, Facebook Messenger, and Instagram) appear on each video. These interactions are important indicators of interest for the content recommendation algorithm which is discussed further below. The sounds and effects pages, described above, can be accessed from any video that employs them through the spinning record icon in the right hand corner of every video, or through the scrolling sound title along the bottom of the screen, or the yellow wand icon with the name of the effect. Importantly, both the sounds and effects screens contain a red pulsing “use this sound/effect” button at the bottom of the page which incites the user to add their own video to the page, and underlines the focus TikTok places on content creation, rather than simply consumption.
While the sounds and effects pages are typically accessed through the buttons available on the video where they are used, the other key screens such as FYP and Discover are navigable from the (always visible) toolbar at the bottom of the screen. Here the user has the option of navigating to Home (which encompasses the For You and Following feeds); Discover, where search and trends are found; Inbox where users can message mutual followers and view notifications, and Profile where the user will find the videos they have created, liked, and saved.
All the screens on TikTok optimize for mobile devices, particularly mobile phones. While the app does offer a desktop version, the experience of using TikTok is meant to utilize the tactile dimension of scrolling on a touchscreen mobile device. Additionally, the built-in logic of reproducibility (incitements to “try this effect” or “use this sound”) when viewing effect and sounds pages assumes that the user will be inclined to make their own version of a trending video, which requires a smartphone to accomplish.
Algorithmic recommendations
The For You page recommendation algorithm is undoubtedly the key algorithmic recommendation aspect of TikTok. The company themselves state that the For You feed is one of the defining feature of the platform and that, “part of the magic of TikTok is that there’s no one For You feed – while different people may come upon some of the same standout videos, each person’s feed is unique and tailored to that specific individual.” (How TikTok recommends videos #ForYou, 2020)
For You recommendations process a number of inputs sent by the user’s past and current use habits. Signals include:
- categories of interest specified at sign-up;
- categories of stated non-interest;
- likes;
- shares;
- accounts followed;
- comments posted; and,
- the type of content that users themselves create (How TikTok recommends videos #ForYou, 2020).
To a lesser extent the algorithm accounts for factors that are innate to the user including video information and the user’s device, network, and account settings. However, since users do not actively express these preferences, TikTok claims that these are less influential in recommendation. In their words: “A strong indicator of interest, such as whether a user finishes watching a longer video from beginning to end, would receive greater weight than a weak indicator, such as whether the video’s viewer and creator are both in the same country. Videos are then ranked to determine the likelihood of a user’s interest in a piece of content, and delivered to each unique For You feed.” (How TikTok recommends content #ForYou, 2020). Interestingly, in a blog aimed at making the For You algorithm legible to creators, rather than content consumers, video watch time is the only significant interaction metric mentioned when discussing how content is recommended.
From a technical perspective, TikTok has revealed during their virtual tour of the European Transparency and Accountability Centre, that the process of recommending content occurs in three steps.
- Step one is “Recall” in which a swath of videos (approximately 2000-3000) are selected at random.
- In step two, known as “Predict”, the user’s preferences, which have been signaled through their past interactions are weighed against the videos selected and narrowed down to several hundred possibilities.
- In the final step, which they dub “Sort”, eight top videos which are both safe and popular are delivered to the user’s feed. Unfortunately, the mechanism of refinement between Predict and Sort steps remains opaque, as does the criteria by which the initial thousands of videos are selected.
TikTok repeatedly emphasizes that their recommendation algorithms aid users in discovering more of what they like (How TikTok recommends content For You, 2020; What is the ‘For You’ feed, 2020). In this sense, a new user whose behaviour and interests have not yet been analyzed by the For You algorithm, TikTok employs funneling techniques, to refine recommendations. In other words, when a user first joins they are served content with broad appeal and high engagement which is then narrowed down as they interact with the content served to them. Similarly, searching for specific content (videos, sounds) is important for content discovery long-term as interactions can help shape future content recommendations as historic search data presumably impacts the type of videos to appear on the For You page.
TikTok is focused not only on serving content that users have already expressed interest in, but also enabling the discovery of new types of videos. This focus is what they refer to as “interrupting repetitive patterns”, which they explain they ensure by not playing videos from the same sound or creator repeatedly and “diversifying recommendations” by serving content from categories outside a user’s typical interests. As of December 2021, TikTok has committed to expanding efforts to interrupt repetitive patterns especially surrounding videos which, when viewed in clusters, can be harmful but might otherwise be fine viewed one-off. This includes videos which may not violate the Community Guidelines, such as dieting content, but can impact users negatively when constantly served to their For You feeds.
Experiences
Given that the app opens on the For You page and automatically begins playing a video, the suggested modality of use is highly passive and fluid. The endless scroll feature ensures that a user could spend hours consuming content without ever having to leave the For You page. Moreover, when scrolling the For You page from certain devices, the device’s toolbar (including time, date, wifi connection) is blocked out. This implies a desire to immerse the user in the For You page, offering few distractions from the perpetual lived present of the feed. Indeed, while on the FYP, the date the videos were posted is absent from view, but becomes visible when watched from a creator’s profile. These atemporal strategies are reminiscent of strategies utilized by casinos to keep gamblers engaged and, moreover, illustrate the desire to create closed loops of use whereby an equilibrium is reached through repetitive motion (McKelvey & Hunt, 2019), in this case: scrolling videos. Given how effortless it is to consume an unending stream of videos from the For You page is, the experience of discoverability on TikTok can be characterized by “gorking out,” whereby decision fatigue is overcome by the simple mechanism of continuous scrolling.
A note about methods
This memo has been developed by studying the information that TikTok makes available on their official blog, aimed at both users (who might seek to better understand how content is being recommended to them) and at creators (who may wish to optimize their videos and posting style to fit the system of algorithmic recommendation). Importantly, considering our Canadian context, little of this official literature is available in French. As of November 2021, TikTok’s official site and access to menus for Resources, Company, and Legal are available in French however this represents a diminished list of options from the full English site menu. Moreover, the information about content recommendation that is available is fairly basic. When “pour en savoir plus” is selected from the French Help Centre article about the For You page, the user is redirected to the “How TikTok recommends content #ForYou” (2020) article in English on the English site. This demonstrates that TikTok has not prioritized making their recommendation system legible to French-only readers, and is rather leaning on the assumption that access to English equivalents will be sufficient.
In the following sections, this memo will elucidate how TikTok serves video content using an algorithmic recommendation system, contextualized within the model of discoverability set out by McKelvey and Hunt (2019). This model views platforms as having unique conditions of content discovery, dependent on their differing interfaces and algorithms. This notion is elaborated through the interlocking concepts of surrounds, vectors, and experiences. In their words, “surrounds refer to the ways that platforms arrange choices on or between screens. Vectors refer to the interactive pathways we take through data, guided by software. Taken together, surrounds and vectors shape the contours of how users find and consume pieces of content, which influences their experiences of online platform flow.” (McKelvey & Hunt, 2019, p. 2)
References
5 tips for TikTok creators. (2019, August 16). Newsroom | TikTok. https://newsroom.tiktok.com/en-us/5-tips-for-tiktok-creators
An update on our work to safeguard and diversify recommendations. (2021, December 16). Newsroom | TikTok. https://newsroom.tiktok.com/en-us/an-update-on-our-work-to-safeguard-and-diversify-recommendations
Discover more of what you love on TikTok. (2021, August 31). Newsroom | TikTok. https://newsroom.tiktok.com/en-ca/discover-more-of-what-you-love-on-tiktokca
George, A. (2022, May 6). TikTok continues rolling out its Friends tab, replacing Discover. Digital Trends. https://www.digitaltrends.com/social-media/tiktok-continues-rolling-out-its-friends-tab-replacing-discover/
Gruzd, A., & Mai, P. (2020). The State of Social Media in Canada 2020 (p. 21) [Data set]. Ryerson University Social Media Lab. https://doi.org/10.5683/SP2/XIW8EW
How TikTok recommends videos #ForYou. (2020, June 18). Newsroom | TikTok. https://newsroom.tiktok.com/en-us/how-tiktok-recommends-videos-for-you
Lucas, J. (2022, March 30). How TikTok heartthrob William White’s thirsty fandom turned toxic. Input. https://www.inputmag.com/culture/will-white-tiktok-fandom-gen-x-middle-age-women-drama
McKelvey, F., & Hunt, R. (2019). Discoverability: Toward a Definition of Content Discovery Through Platforms. Social Media + Society, 5(1). http://dx.doi.org.lib-ezproxy.concordia.ca/10.1177/2056305118819188
Savic, M. (2021). Research Perspectives on TikTok & Its Legacy Apps| From Musical.ly to TikTok: Social Construction of 2020’s Most Downloaded Short-Video App. International Journal of Communication, 15(0), 22.
Smith, B. (2021, December 6). How TikTok Reads Your Mind. The New York Times. https://www.nytimes.com/2021/12/05/business/media/tiktok-algorithm.html
Tidy, J. (2019, July 2). TikTok’s young fans ‘exploited’ for digital gifts. BBC News. https://www.bbc.com/news/technology-48725515
What is the ‘For You’ feed? (2020, December 24). TikTok Creator Portal. https://www.tiktok.com/creators/creator-portal/en-us/tiktok-creation-essentials/whats-the-for-you-page-and-how-do-i-get-there/
Wang, E., Wu, K., & Zhu, J. (2020, July 29). Exclusive: ByteDance investors value TikTok at $50 billion in takeover bid – sources. Reuters. https://www.reuters.com/article/us-bytedance-tiktok-exclusive-idUSKCN24U1M9
Zhang, Z. (2021). Infrastructuralization of Tik Tok: Transformation, power relationships, and platformization of video entertainment in China. Media, Culture & Society, 43(2), 219–236. https://doi.org/10.1177/0163443720939452
We acknowledge the financial support of the Government of Canada.
