Exercises: The Algorithm Whisperer

Difficulty Guide: - ⭐ Foundational (5-10 min each) - ⭐⭐ Intermediate (10-20 min each) - ⭐⭐⭐ Challenging (20-40 min each) - ⭐⭐⭐⭐ Advanced/Research (40+ min each)

Part A: Conceptual Understanding ⭐

A.1. Explain the core logic of a recommendation algorithm in your own words. What are the six steps (observe → learn → predict → serve → measure → repeat)?

A.2. What is the difference between an interest graph and a social graph? Which platform primarily uses which, and what does this mean for new creators?

A.3. Describe the distribution funnel. Why does early engagement from the seed audience matter so much for a video's ultimate reach?

A.4. Compare TikTok's primary metric (completion rate) with YouTube's (watch time + CTR). Why does this difference lead to different types of content succeeding on each platform?

A.5. What are the four universal signals that every algorithm rewards? How do they map onto the skills from Chapters 1-6?

A.6. Why do "algorithm hacks" fail long-term? Give at least two reasons from the chapter.

Part B: Applied Analysis ⭐⭐

B.1. Check your analytics on any platform. For your last 10 posts/videos, list: - Completion rate (or average view duration) - Share count - Save count - Follow rate (new followers / views) Which videos performed best on universal signals? Is there a pattern?

B.2. Compare the same topic on TikTok and YouTube. Find a video about a similar topic on each platform. Analyze: - How long is each video? - How is the opening different? - What engagement patterns do you see in comments? - How does the content structure reflect each platform's algorithm priorities?

B.3. Find a creator who posts on multiple platforms (TikTok + YouTube, or TikTok + Instagram). Does the same content perform equally well everywhere? If performance differs, hypothesize why based on each platform's algorithm model.

B.4. The chapter claims Instagram weights DM shares heavily. Test this: look at your Instagram activity over the past week. How many Reels did you send via DM vs. how many you just liked? What does this suggest about which Reels the algorithm should amplify?

B.5. Look at a creator whose views fluctuate significantly from video to video (some get 10x others). Using the distribution funnel, hypothesize what's happening: are some videos failing at the seed stage, the expand stage, or the amplify stage?

Part C: Real-World Application Challenges ⭐⭐-⭐⭐⭐

C.1. The Cross-Platform Audit ⭐⭐ Choose one video idea and write three different openings — one optimized for TikTok (autoplay, first-frame hook), one for YouTube (CTR-optimized title + thumbnail concept), and one for Instagram Reels (visual + caption). Explain why each opening differs based on each platform's entry point.

C.2. The Signal Optimizer ⭐⭐⭐ Take one of your recent videos (or a video you're planning). For each universal signal, identify: - Attention: What keeps the viewer watching? Where might they drop off? - Action: What would motivate them to like, share, save, or comment? - Return: What makes them want to see your NEXT video? - Satisfaction: Will they feel good about having watched this? For each signal with a weakness, propose a specific improvement.

C.3. The Algorithm-Proof Test ⭐⭐⭐ Imagine TikTok made a major algorithm change tomorrow that deprioritized completion rate and instead weighted shares 3x more heavily. How would this change affect: - Your content strategy? - The types of content that succeed on the platform? - The behavior of creators? Now imagine the reverse: shares are deprioritized and completion rate becomes 3x more important. What changes? This thought experiment tests whether your strategy is algorithm-dependent or algorithm-proof.

C.4. The Distribution Funnel Hack ⭐⭐⭐ Most videos fail at Stage 1 (seed audience engagement). Design a strategy to maximize your seed audience's engagement: - Who IS your seed audience? (Followers? Interest-matched strangers?) - What time do they tend to be active? - What type of opening do they respond to most? - How can you ensure they see and engage with the video quickly? Important: frame this as serving your seed audience well, not "gaming" them.

Part D: Synthesis & Critical Thinking ⭐⭐⭐

D.1. The chapter says TikTok's interest graph is "the most democratizing platform for new creators." But critics argue it also means your content is constantly competing against everyone, making it harder to build a stable audience. Is TikTok's model actually better for new creators, or does it just trade one form of disadvantage (lack of followers) for another (constant competition)?

D.2. YouTube's shift from clicks (pre-2012) to watch time (2012-2016) to satisfaction (2016-present) shows the platform learning from its mistakes. Each metric was "gamed" — clickbait for clicks, artificially long videos for watch time. Predict: how will satisfaction be gamed, and what will YouTube's next evolution be?

D.3. The chapter argues that "building for humans first" is the best algorithm strategy. But human attention is itself subject to manipulation (Chapter 4: emotional contagion; Chapter 5: curiosity gaps). Is "building for humans" genuinely different from "building for the algorithm," or are both just different frames for "exploit psychology for engagement"?

D.4. If algorithms reward what viewers watch and engage with, and viewers tend to watch high-arousal negative content (outrage, drama), then algorithms naturally amplify negativity. Should platforms modify their algorithms to deprioritize high-engagement content that creates negative outcomes? Is this censorship, curation, or social responsibility?

Part E: Research & Extension ⭐⭐⭐⭐

E.1. Research the YouTube recommendation algorithm's history. Find blog posts from YouTube's engineering team (often published on the YouTube Official Blog or in research papers) about their algorithm updates. How have their stated goals evolved over time?

E.2. TikTok's algorithm has been described as both "democratizing" and "addictive." Research the debate around TikTok's algorithm design — from technology critics, platform researchers, and TikTok's own statements. What are the strongest arguments on each side?

Solutions

Selected solutions available in appendices/answers-to-selected.md