TikTok: No Random Walk, i.e. This is Why Every App is Adding Stories Now
The shortest-form mobile video social network hooks viewers by predicting away the burden of entertainment overchoice and connecting the world to an endless scroll of unique storytellers
National security hot button. Gen Z darling. 800 million MAUs. A lightning rod for controversy. Perhaps misunderstood. Boardroom intrigue. I love TikTok. I rarely have time for long-form entertainment these days. I can learn something new, have a laugh, get a slice of feel-good, and connect with everyday people, 15 seconds at a time. It’s become a good joke on Tech Twitter that every app now has stories. But why? What’s driving this “…weird social obligation…” as Gizmodo recently asked? Does shortest-form (read: not Quibi) video have major entertainment platforms in its cross-hairs, or the internet as a whole? Social platforms are being put on notice.
TikTok has a lot going on. For this exercise, I’ve narrowed the scope to the main endless swipe content feature and customers in the U.S. I’ve excluded content creation, discover/search, messaging/inbox, push notifications, and live streaming. Prioritize and focus, right?
Let’s conduct a product breakdown: I’ll walk you through TikTok’s customer journey, review two customers (viewers) and their personas, and talk about TikTok’s core value proposition. I’ll build on that to discuss where TikTok needs improvement and envision the next big feature.
I’ll try to keep it pithy.
Start with the customer: journey and empathy mapping
How it works
First, let’s define a touchpoint as any interaction that can alter how a customer feels about a product. That gives us four major touchpoints on the main content page. Let’s discuss and walk through each one. Try it for yourself to better follow along. Let’s also call our customers viewers.
The major viewer touchpoints are:
- View content
- Swipe for the next video
- Engage with the content
Here’s how it works. Open TikTok. A video plays immediately from your “For You” queue. If you like it, you might watch it again (1). You can also select “Following” to only see videos from creators you’ve followed. If you want to move on, you swipe to see the next video, often within seconds of the current video playing. Videos are vertical and mobile-native (2).
If you like what you saw, use the icons on the right to further engage (3). You can subscribe to the creator (“+”), tap on their avatar to learn who they are, like/heart the video, and comment. You can also share on social media. Each video includes the song playing, a brief description, and relevant hashtags on the bottom.
We can summarize these interactions as view(1 and 2) and engage (3).
Let’s use some design thinking. Our viewers’ needs are informed by feelings. To break this down, I opened TikTok and a mind map to do some stream-of-consciousness empathy mapping. I incorporated what I could see and hear, how I felt, and what I thought about as I swiped. I noted any pain or gain points that stood out. I took things a step further: after getting some feedback from some younger family members, I put on my Daydream VR headset and watched a few TikTok compilations. Immerse yourself!
Feeling and thinking: I hope this video matches my mood. I’m ready to be surprised or see something new. Did My Korean Dad post today? Am I going to get an ad first? I always mis-tap on them…
Needs: relevant entertainment, fast. I may not be able to play audio right now, so text helps. I need a mental pick-me-up.
(2)Swipe for the next video
Feeling and thinking: this is boring…next! What am I looking at right now? How long do I have to swipe before I find something I’ll like? I want something new. I want something from someone familiar.
Needs: figure out my mood and don’t make me scroll too long to find what I want. I need to smile.
Feeling and thinking: I liked that! That creator is amazing, more please! What do other people think? What did I just watch? What song was that? My friends have GOT to see this.
Needs: I want to see more of this type of content, I want to see more of this creator, I want to quickly share this to my friends and family.
These feelings and needs give us insight into our viewers' goals. What are TikTok’s business goals? How closely do the two align?
Key metrics: optimizing the algorithm
TikTok’s mission is to inspire creativity and bring joy. There’s no shortage of the former when anyone can become a creator and viewers can join in on everything from duets to dance challenges. The latter depends on how good the recommendations are and how fast the AI calibrates to overall interests and current mood.
TikTok monetizes its content through a virtual currency that can be used to support creators and relevant ads based on interest. Recommendation quality informs how long your attention can be held and therefore determines how well TikTok can monetize content. Let’s use that lens to evaluate the actions and metrics that are optimized to achieve these goals:
Metrics and actions: partial vs. full watch, # of rewatches, total session time
(2)Swipe for the next video
Metrics and actions: Swipe velocity (time between swipes), swipes by content type/creator, swipes per session
Metrics and actions: likes, follows/subs, re-shares (and to which social media platforms), comments
Viewer personas: more than Gen Z
We can discover who our viewers are by empathizing with them to better understand their entertainment goals and pain points. We should supplement their point of view with data. I’m going to extrapolate the personas from my experience with TikTok and what I’ve learned about the platform’s viewers from marketing commentary.
60% of TikTok’s viewers are 16–24 in the U.S. But that’s not the whole story. The customer base is diverse. Just ask the seniors of TikTok. We can learn more by considering two ends of the spectrum and putting ourselves in their shoes:
If we had more time, we might interview users and observe how they use TikTok and other content platforms to distill clearer viewer personas. There are many, many more. Ideally, our personas are built based on a combination of data and direct field research.
Predicting away overchoice
TikTok eliminates the burden of entertainment overchoice. Netflix has an estimated 36,000 hours, or over four years of documentaries, shows, and movies. In 2018, the catalog began to interfere with the content platform giant’s acquisition process. This problem has many permutations: paradox of choice, FOMO, analysis paralysis, and decision fatigue.
TikTok picks your content for you and predicts what you’ll enjoy. There are no thumbnails or categories that you must choose from. No need to search, no need to mull anything over. Open and go. Less friction and fewer chances to drop out of the adoption funnel or abandon the app. TikTok’s algorithm is the decision support system. That makes it easier to simplify the UI.
Good content matters, too. A powerful recommendation model is nothing without a variety of entertainment. We all need a brief respite from the world especially this year. Swiping for fun is a welcome change from doomscrolling.
Bytedance is tight-lipped about how its algorithm works. I agree with pundits: the engine likely also factors in location, geography, time of day, and perceived demographics and psychographics into what gets shown next. I suspect the algorithm is using real-time training data for continuous updates. Changes are likely deployed multiple times per day to support viral video detection, dance challenge trends, and better results from cold starts (i.e., no prior information about the user). I hope they have a rockstar DevOps team, fwiw.
Did you try it out? How long did it take before you were shown videos about topics that interest you? Did you begin seeing more content that you liked as you swiped?
Not all rosy: dark pattern advertising
I spent more time on TikTok than usual over Thanksgiving. I thought about what step in the process gave me the most friction and detracted from feeling creative or joyful. Where did I land?
Pre-roll takeover ads. TikTok monetizes content in part through advertising. This often includes an ad that plays before your first video and mid-rolls intermittently throughout your session. The pre-roll ads have a dark pattern problem: swiping away takes you to the click-through site.
Pre-roll ads are an inevitable part of the viewer experience. TikTok should optimize every portion of the journey, even ads. At best, it’s poor design without empathy for the viewer and their expectations. Long-term, the ads could be perceived as deceptive and hurt TikTok’s social commerce aims. The company has already faced challenges with its advertising business.
How we could solve it
Let’s break down how best to solve the problem. The swipe UX for pre-roll ads differ from regular content and mid-roll ads. This causes distrust and annoyance. From a business goal perspective, ad click-throughs may be up, but engagement is likely suppressed. From a design perspective, it’s inconsistent and elicits the wrong signals. From an engineer’s point of view, pre-roll ads appear to unnecessarily have separate code mechanics.
If our goal is to make the experience consistent and engaging for viewers, a good solution would be to allow a swipe-down to skip the pre-roll ad. Alternatively, we could require a certain number of seconds before allowing a swipe-down.
Tradeoffs and considerations
So, what are the tradeoffs? It’s possible this is part of A/B testing and others do not experience this issue. As of a few days ago, I no longer see these ads. I’m also amazed at how many ads are not in the style of TikTok. Most appear to be condensed versions of TV and YouTube ads. There are other opportunities to improve the ad process (for consumers and the business alike). From a metrics perspective, we expect click-throughs will go down.
We should monitor overall ad engagement (CPMs, ad pricing, etc.) to track the overall impact. Even if it decreases, TikTok should evaluate how dark patterns support or deter from its overall mission. We could start by testing from groups that tend to engage with our ads before broader rollout/fix.
Mood: the algorithm gets a human touch
TikTok’s feed brings joy by introducing viewers to a variety of content around the topics they love. The recommendation engine needs inputs to get better at showing you what you’re interested in. But what about a more human cue?
TikTok’s recommendation algorithm is good. Scary good. When I need a quick break and don’t want to pick something to watch, this process works well. After all, if you don’t mind where you’re going, any path will take you there. But I have to guide myself to videos that I want to see when I’m in a certain mood. This often drives me to abandon the session. I’m not the only one. How often do you see people leaving reviews complaining about “the algorithm?”
Mood: TikTok’s next big feature?
TikTok could use some initial input from its viewers. We need videos that match our current emotions and don’t waste time. Enter Mood.
From a design standpoint, we need to add this feature while maintaining the other key draws of TikTok. That means using space wisely and not disrupting the UI. Content still plays on the screen, one video at a time. Mood will appear at the top next to Following and For You.
UX will remain largely the same, with one twist. Upon tapping on Mood, an emotion title will display with short previews of top content for that mood or emotion. Swiping right or left scrolls through the moods available. Swiping up or down begins play of videos in that genre.
From an engineering standpoint, our in-app mechanics require some changes. We’ll need to test how adding this feature affects our other metrics and features. We should A/B test this feature first before a full feature ship. We should also test a prototype first before a broader rollout. For example, we could start with 1–2 initial emotions. We can determine which moods would be best through market research into our target personas. We could target a small subset of users of the “Following” section (a close proxy for users that give us a signal of what they want to watch).
We’ll work with Data Science to measure effectiveness over time and ensure we can curate content tied to an emotion. We’ll also assess how this new cue can be leveraged if successful. Can we enrich our ad targeting with these emotional cues? Can we resurface content? How does this user input affect our existing recommendation algorithm? Our overall goal should be to measure increases in the total session time. We expect some users to watch the previews before swiping in. This gives us a proxy for measuring against existing features like the “Following” section we discussed earlier.
Mood supports TikTok’s core mission and its viewers' needs. Our primary goal should be to measure for the key metrics we identified earlier. As we grow, we should ensure the feature stays aligned with the company’s mission and other new features. Mood also gives us a new input for the algorithm: how our viewers feel before swiping through content.
Looking to the future: curated algorithms?
Brad Slingerlend of NSZ (Non-Zero Sum) Capital discussed the concept of a more human approach to algorithms in his newsletter:
“It seems a poor assumption that the big tech platforms have devised the best algorithms to serve their customers. Many of their searches/feeds are optimized for ads, clickbait, or addictive hooks (just don’t tell Zuckerberg that), or to reward a set of user behaviors that a platform’s machine learning routines have deemed useful… Expanded broadly to all types of information/media consumption, perhaps folks would prefer to have human curators — content jockeys — whose selections they prefer over those of a purely computational algorithm (with questionable and/or oblique intentions)…Jack Dorsey expressed excitement over the possibility of third-party Twitter feed algorithms (first suggested by Stephen Wolfram), and seemed to indicate that the best algorithms might come from outside the company. All this dialog makes me thoroughly intrigued by the idea of having a smorgasbord of algorithms guided by a diversity of human views for myriad use cases and consumers…an entire economy of influencers and corporations could emerge… In an ever-changing landscape of disruptive innovation, no company should have the hubris to believe they have a monopoly on the best algorithm for anything.”