Tap, Wait, Watch: Deep Link Performance Across TikTok, YouTube, Snapchat & Triller

When it comes to mobile apps, seamless navigation is key to keeping users engaged. Deep linking is crucial in enhancing the user experience by directing them straight to specific in-app content, eliminating unnecessary steps, and reducing friction.  Today users frequently share content across social media platforms and messaging apps, deep linking is more important than ever, as it ensures users land exactly where they expect to, without unnecessary delays. But how well do deep links perform under real-world network constraints?

TestDevLab’s Audio-Video Testing Lab evaluated deep link performance across four popular short-form video apps—TikTok, YouTube, Triller, and Snapchat—under varying network conditions. Our goal? To analyze how efficiently each app handles deep links regarding startup time, buffering, and video quality.

Through rigorous testing, we uncover how these platforms adapt to challenges like slow network speeds and different startup states, providing valuable insights into app reliability and user experience.

What is deep linking?

Deep linking is a powerful technology that enables users to be taken directly to specific content within a mobile app—bypassing the app’s home screen. Whether the link is shared through social media, messaging, email, or the web, it creates a seamless transition by opening the relevant in-app page, rather than forcing users to manually navigate to it.

This approach significantly improves user experience, boosts engagement, and increases conversion by reducing friction between discovery and interaction. For example, clicking on a shared video link can open that exact video inside the app instead of just launching the homepage.

Platforms like Android use App Links (and Apple uses Universal Links) to handle these actions, allowing a single URL to intelligently direct users to either the app or mobile website depending on availability—without prompting the user to choose.

In today’s content-driven environment, where links are constantly exchanged, deep linking is essential for keeping users engaged and delivering a frictionless mobile experience.

Test setup for deep link evaluation

To test and analyze deep link performance in our AV test laboratory, we set up a mobile device environment with a custom network configuration, selected four popular social media apps commonly used for sharing videos, and used a unique video for consistency. 

1. Devices & environment

The following environments were created and devices used:

• Mobile device used: Samsung Galaxy S21, Android 14
• Device used: Apple MacBook Pro 13.3, macOS Sequoia 15.0.1 
• Network conditions: 
  • Bitrate limited to 30 Mbps
  • Bitrate limited to 10 Mbps
  • Bitrate limited to 3 Mbps
  • Bitrate limited to 1 Mbps

Why did we choose these network limitations? Here are some real-world examples of network limitation use cases:

• 30 Mbps bitrate (high-speed connection, optimal conditions). A user is at home, connected to a stable Wi-Fi network, watching high-quality TikTok videos without buffering. The app loads videos instantly, and switching between content is seamless.
• 1 Mbps bitrate (very slow connection, severe buffering, and low quality). A user is traveling in a remote location with a weak signal, trying to watch a Triller video shared by a friend. The video loads extremely slowly, starts in low resolution, and pauses frequently due to buffering.
• 3 Mbps bitrate (low-speed connection, buffering expected). A user is commuting on a train in an area with poor 4G/5G coverage. They open a Snapchat video link from WhatsApp, but the video takes longer to load and plays in lower resolution to compensate for the slow connection.
• 10 Mbps bitrate (moderate-speed connection, minor delays possible). A user is on mobile data (4G) in a coffee shop, streaming YouTube Shorts. The videos load quickly.

2. Apps tested

We tested four popular social media apps for short video forms:

• TikTok, version 38.7.3
• Youtube, version 20.06.36
• Snapchat, version 12.29.0.49
• Triller, version v55.2b54

3. Video uploaded 

To ensure a fair comparison across different apps, we uploaded the same video to all platforms for testing under various deep link scenarios and network conditions. Video complexity plays a crucial role in evaluating performance, as different apps may handle high spatial complexity differently.

Video specifications

• Duration: 15 seconds
• Frame rate (FPS): 30
• Bitrate: 13,969 Kbps
• Resolution: 1080x1920
• Codec: H.264 (High) (avc1 / 0x31637661)
• Color format: YUV 4:2:0 (progressive)

Video complexity

• High spatial complexity – Contains highly detailed scenes, intricate textures, and high contrast.
• Low temporal complexity – Features a static camera and background, with only the subject (a dancing girl) in motion.

Testing methodology

To ensure consistent and objective results, we followed a structured testing approach across multiple deep link scenarios and network conditions. The section below outlines the testing period, scope, test steps, duration, and tools used.

• Testing period: Feb 21- Feb 27, 2025
• Test count: 1,410
• Deep link scenario with 3 variations tested:
  • Cold startup: The app is entirely closed, and tapping the deep link launches the app from scratch, without any preloaded data or cached resources.
  • Opened app: The app is running in the foreground for 5 seconds, even if idle, with certain resources potentially already preloaded.
  • Warm startup: The app is running in the background, and tapping the deep link resumes the app from its previous state, utilizing any preloaded resources.

Below is a table of the steps taken to perform the tests for each deep link scenario.
Note: The precondition for the test is to set the app in a cold start state (clear cache).

• Recordings length:
  • 15 seconds for 30 Mbps, 10 Mbps and 3 Mbps tests
  • 30 seconds for 1 Mbps tests
• Tools used for data processing and collection:
  • ffmpeg
  • Text detection for startup time measurement
  • VMAF - full reference video metric; additionally used in-house built video processing argument to ignore color differences (darker overall for some of apps)
  • Ntrex - in-house built tool for processing network
  • Tempero - in-house built tool for  data collection
  • Viqubox - in-house built tool for audio, video, network processing
  • Video complexity measurement tool - VCA

To avoid any external impact from video link sharing, the video link is opened from a note-taking app on Android. As per the scenario, the user opens a pre-saved note that contains the video link.

Key performance metrics

The following performance metrics were gathered and analyzed:

• Startup + buffer time. How long it takes for the video to load after clicking on a video link from the Notes app plus adding the duration of buffering events during recording time
• Out-of-scope tests. Metric used for poor latency, when video playback does not start within the recording time (either 15 or 30 seconds depending on network limitation). Out-of-scope tests are excluded from the average data calculation as video playback has not started and no performance data has been gathered.
• VMAF. This full reference video quality metric was developed by Netflix. For this particular experiment, an ignore color difference argument is used when processing video quality because most of the apps have a darker overlay at the top or bottom of the screen.
• Startup time. How long it takes for the video to load after clicking on a video link from the Notes app

Results & findings

We observed notable differences in how each app handles deep linking under varying network constraints and startup conditions. The findings highlight key strengths and weaknesses across TikTok, YouTube, Triller, and Snapchat—offering a deeper look into their performance and reliability in real-world scenarios.

1. Deep link with cold startup state scenario 

Scenario: The app is completely closed, and the deep link launches it from scratch.

Example: A user is scrolling through WhatsApp chats and sees a message from a friend: "You HAVE to see this! 🤯🔥" with a TikTok link attached. Out of curiosity, they tap the link, and since they haven’t opened TikTok yet today, the app launches from scratch. Once it opens, it takes them directly to the video their friend sent.

To mimic this scenario, the app cache is cleared, all apps (including the test app) are closed in the background, and no processes are active. The user then opens a note-taking app and taps on the video link.

At 30 Mbps and 10 Mbps, none of the apps experience buffering. However, buffering is required for all apps when testing under 3 Mbps and 1 Mbps conditions.

In the video below, we can see the overall app performance in the deep link cold startup scenario when the network is limited by 30 Mbps. When video playback starts, the video is tinted in green. 

When analyzing the deep link with a cold startup state scenario, we can see that YouTube has the fastest startup time across all network limitations compared to its competitors. However, on a 1 Mbps connection, Snapchat outperforms YouTube in overall startup + buffering time.

With a 1 Mbps limit, YouTube starts playing in 7.1 seconds but buffers for an additional 10 seconds during the 30-second test. In contrast, Snapchat starts slower at 9.36 seconds but buffers only 5.24 seconds throughout the test.

With a 3 Mbps limit, Triller is faster than competitors in terms of startup + buffer time.

Among the competitors, TikTok shows the poorest performance, taking the longest to open the video link. It requires more time to transition to the video, as it always opens the home feed before redirecting to the linked video.

At 1 Mbps, there are only 9 out of 20 TikTok tests that are within the scope meaning that TikTok manages to start the playback within 30 seconds, other 11 tests do not start to play the video during the testing time.

As we can observe in the graph above, YouTube maintains slightly higher video quality than its competitors across all network limitations. However, on a 1 Mbps connection, its average VMAF drops from 81 (observed at 30 & 10 Mbps) to 71.

TikTok, Snapchat, and Triller exhibit similar video quality across 30 Mbps to 3 Mbps tests. However, struggling to balance startup + buffer time performance on 1 Mbps, Snapchat reduces its video quality, with the average VMAF dropping from 73 to 60.

2. Deep link with opened app scenario: casual browsing & link click

Scenario: The app is already open, but the user is redirected to a specific video via a deep link.

Example: A user opens Triller but gets distracted and doesn’t scroll—they just leave the app open. After a few seconds, they switch to WhatsApp to check a message from a friend. The friend has sent a link saying, "You need to watch this dance challenge! 🔥" The user clicks the link, and since Triller is already open, the app doesn’t reload but simply jumps straight to the shared video.

To mimic this scenario, the app cache is cleared first, all apps (including the test app) are closed in the background, and no processes are active. Next, the user opens the app and waits 5 seconds to allow the content to load. Finally, the user opens a note-taking app and taps on the video link.

The deep link with opened app scenario is challenging not only for TikTok but also for Triller. When the network is limited to 30 Mbps, Triller starts faster compared to the deep link with the cold startup scenario. However, at 10 Mbps, Triller's startup time is three times longer, with highly inconsistent performance—ranging from 2.23 seconds to 12.25 seconds.

At 3 Mbps, Triller struggles significantly, with 26 out of 27 tests failing to start video playback—meaning the app does not load the video within 15 seconds. When the user taps on the video link, Triller first transitions to the home feed before navigating to the video link, but the screen remains unloaded, and playback never starts. At 1 Mbps, all 26 tests failed to start video playback and are considered to be out of scope.

In the graphs below, Triller is assigned a value of 30 seconds, representing the maximum test duration for out-of-scope tests.

At 1 Mbps, there are only 10 out of 22 TikTok tests that are within the scope meaning that TikTok manages to start the playback within 30 seconds.

3. Deep link with warm startup steps: Watching a video, switching apps, and returning

Scenario: The app is running in the background, and the deep link brings it back to the foreground.

Example: A user is browsing YouTube and comes across a Travis Scott concert clip from last night’s show. They tap on the video and watch for about 15 seconds before deciding to check something else. They switch apps—maybe they open Notes to jot something down or reply to a WhatsApp message. A few minutes later, their friend drops a YouTube link in a chat saying, "Check out this insane edit of the same concert! 🎥🔥" The user taps the link, and since YouTube was already running in the background, the deep link brings it back up instantly, opening the new video without restarting the app.

Once again, YouTube outperforms competitors on network limitations ranging from 30 Mbps to 3 Mbps when testing the deep link with a warm startup scenario.

However, at 1 Mbps, Snapchat surpasses YouTube in performance—its video playback starts faster than competitors and experiences less buffering.

TikTok continues to lag, showing significantly slower startup times across all deep link scenarios.

Triller also struggles in the deep link with the warm startup scenario. On a 3 Mbps connection, only 12 out of 29 tests were within scope, highlighting extreme performance inconsistency—video playback failed to start within 15 seconds in more than half of the tests. Yet, for successful test cases, the average startup time was 8.39 seconds, with no additional buffering. Interestingly, Triller maintains consistent video quality across all network limitations, as seen in the comparison graph.

At 1 Mbps, only one out of 27 Triller tests successfully started video playback within the 30-second scope. Similarly, TikTok struggles at this speed, with only 6 out of 16 tests completing playback within the scope.

At 1 Mbps, YouTube starts video playback in an average of 7.14 seconds, with an additional 11.30 seconds of buffering during a 30-second test recording. However, analyzing individual test results reveals significant inconsistency in startup times. See the graph below that includes YouTube startup time and extra buffering during the test on 1 Mbps. 

In some cases, YouTube can start playback in as little as 0.95 seconds without any additional buffering—but at the cost of reducing image quality to VMAF 50. Conversely, in other instances, startup time extends to 12.72 seconds, requiring an extra 10.3 seconds of buffering, resulting in a total Startup + Buffer time of 23.05 seconds. In these cases, however, video quality remains high at VMAF 78.

Overall user experience when clicking on a link when the network is limited by 1 Mbps limitation at the deep link warm startup scenario is visible in the example below. The video is tinted in green when the video is playing (after the playback has started) and in red when buffering happens.

4. Deep link scenario impact on app performance

The behavior of video quality under each network constraint is consistent across all three scenarios. An example of video playback quality at 30 Mbps is shown in the table below. YouTube delivers higher video quality compared to competitors, while TikTok, Snapchat, and Triller maintain similar quality levels.

App performance varies significantly across different deep link scenarios. As previously mentioned, Triller struggles to start video playback within the testing time on 1 Mbps and 3 Mbps limitations in the deep link opened app and deep link with warm steps scenarios. However, it manages to start playback in the deep link cold startup scenario. This suggests that Triller performs better when no content is preloaded. The more the user interacts with the app, the more it struggles under challenging network conditions.

The deep link cold startup scenario is less challenging for YouTube than the other two scenarios when analyzing results for 30 Mbps, 10 Mbps, and 3 Mbps tests. However, at 1 Mbps, YouTube's performance remains consistent across all deep link variations, with video playback starting in approximately 7 seconds, followed by an additional buffering period of 9–11 seconds. While YouTube successfully initiates playback quickly, the almost immediate buffering at the start may impact user experience, as the video stops shortly after beginning. Additionally, YouTube slightly reduces its video quality to 1 Mbps compared to higher network conditions.

TikTok consistently takes longer to transition from deep links to video playback, opening the home feed first.

Key findings

• YouTube is the fastest among competitors but experiences initial buffering at 1 Mbps.
• Snapchat shows a similar start time and buffering to YouTube at 1 Mbps. 
• Triller struggles significantly under network constraints, with frequent out-of-scope tests at 3 Mbps and 1 Mbps.
• TikTok consistently takes longer to transition from deep links to video playback, opening the home feed first.

Conclusion 

Our deep link performance analysis of TikTok, YouTube, Triller, and Snapchat under different network conditions reveals key insights into how these apps handle deep linking in real-world scenarios. While YouTube consistently demonstrated strong performance across most tests, Snapchat outperformed in low-bandwidth conditions for deep linking with warm startup, and Triller struggled with video playback reliability, particularly under network constraints.

Are the videos on your app lagging and buffering? We can help! Contact us to learn more about our video quality testing services and discover how we can help you achieve maxiumim user satisfaction.