# Relentless Product Improvement > A framework for driving product growth by combining deep user empathy (Jobs to be Done), analytical rigor, and design clarity. Use it when troubleshooting a feature that has plateaued, launching a new "big bet" within a mature product, or diagnosing why users are abandoning a specific workflow. - Skill: `samarv/relentless-product-improvement` (Agent Skill) - Install (CLI): `npx skillmds@latest add samarv/relentless-product-improvement` - Raw SKILL.md: https://api.skillmd.com/api/skills/samarv/relentless-product-improvement/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Marketing & Growth - Author: samarv (https://skillmd.com/u/samarv) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/samarv/relentless-product-improvement --- # Relentless Product Improvement To build products at a global scale, you must embody "productive dissatisfaction"—the refusal to tolerate mediocre user experiences and a commitment to compounding small improvements until they hit a tipping point of success. This framework guides you through the process of understanding why users "hire" your product, diagnosing failures with rigor, and shipping with extreme clarity. ## 1. Deeply Understand the "Big Hire" Move beyond demographics to find the *causation* of why someone uses your product. Every user "hires" a product to do a job. - **Conduct "Interrogations":** Ask users exactly where they were and what they were doing the moment they decided to use your product. - *BAD:* "Do you like this feature?" - *GOOD:* "Where were you? Were you in bed? At work? What was the exact thought that prompted you to open the app?" - **Identify Emotional Jobs:** Recognize that users often hire products for emotional needs (e.g., feeling connected, reducing anxiety) rather than just functional utility. - **Find the "Big Hire" Moment:** Identify the first time a user decides to use the product. The information obtained from that specific moment is more critical than any general feedback. ## 2. Apply Analytical Rigor to Root Causes When metrics drop or a feature fails to gain traction, do not guess. Dissect the data to find the specific point of failure. - **Perform Segmented Root Cause Analysis:** If a core metric is down, investigate if the issue is isolated to a specific region, device, demographic, or use case. - **Measure the "Closed Loop":** For social or communication products, ensure the user's action results in a response. - *Example:* If a user posts a private story but has too few people on their list, the probability of a "reply" is zero, causing the user to abandon the feature. - **Track S-Curves:** Monitor every feature's growth. When a feature hits a plateau (diminishing marginal returns), reallocate resources to first-principles bets rather than minor optimizations. ## 3. Design for Clarity over Cleverness Avoid "differentiating" through unique UI if it compromises the user’s mental model. Standards provide leverage. - **The Door Principle:** A door should communicate whether to push or pull through its design (a handle vs. a flat plate). Apply this to UI. - **Use Global Icons:** Do not reinvent standard symbols. If a feature uses the camera, use a camera icon. Avoid "clever" icons that require a learning curve. - **Leverage Spatial Thinking:** People think of apps spatially. If you add a new feature to an existing product, it must feel coherent but distinctive. - **Complement, Don't Replace:** When building a new format (like AI Mode in Search or Stories in Instagram), design it to expand the product’s utility rather than forcing users to change existing habits. ## 4. Scaling the Breakthrough Once you have internal conviction (the "perfect shot" moment), move from scrappy to robust. - **Move Fast on "Live" Testing:** Use "Trusted Tester" groups of ~500 people to find where the product "sucks" before a public Labs launch. - **Identify the Conviction Moment:** Once you see a small group of users finding intense value, invest resources to build the "best version" rather than staying in a "minimum viable" state too long. --- **Example 1: Reviving a "Flop" Feature** - **Context:** A private sharing feature (Close Friends) is failing to grow. - **Input:** Data shows users are only adding 1-2 people to their list and never getting replies. - **Application:** 1. **Understand People:** The "job" is feeling connected. Two people isn't enough to guarantee a response. 2. **Analytical Rigor:** Identify that users with 20+ friends on their list have high retention. 3. **Design for Clarity:** Change the name from "Favorites" (which implies a very small number) to "Close Friends." Add a clear visual cue (e.g., a green ring) on the *outside* of the feature so users know it's unique before they click. - **Output:** The feature loop closes, replies increase, and retention stabilizes. **Example 2: Complementary Product Expansion** - **Context:** A mature search engine wants to integrate generative AI without breaking the core keyword-search habit. - **Input:** Users are starting to append "AI" to their queries because the standard results aren't answering complex questions. - **Application:** 1. **Identify the Gap:** Recognize that "keyword ease" is failing for multi-sentence, complex intents. 2. **Build a Distinctive Space:** Create an "AI Mode" that is a full-page, conversational experience, rather than just a small box in the existing results. 3. **Design for Clarity:** Name it "AI Mode" so users immediately understand the capability change. - **Output:** Users use the core search for quick utility (phone numbers, prices) and the AI Mode for "expansionary" curiosity (planning, research). --- ## Common Pitfalls to Avoid - **Reinventing the Wheel:** Creating a custom UI for a standard action (like a "sharing" icon) just to be "on-brand." This forces users to think, which leads to abandonment. - **Staying Scrappy Too Long:** Holding onto a tiny team after you've found a breakthrough. This slows down the iteration cycle and allows competitors to get ahead. - **Ignoring the "Causation" Interview:** Building features based on what users *say* they want in a survey rather than what they *did* in the moments leading up to using the product. - **Acceptance of Mediocrity (Habituation):** Becoming like a "grown-up" who ignores the stickers on fruit that puncture the flesh. Always ask "Why does this suck?" and "How can this be better?"