Skip to main content

When you have to kill your product

One of the toughest moments is when you have to kill your product 💀—when you see that your product is failing. 

In this article, I am sharing my piece of advice about reading early on the signs, potential challenges that I have encountered in the AI product development, and last but not least, mitigation strategies!

🔎 Reading the signs

A few signs that indicate your product should be removed from the market:

  • Low performance in primary KPIs: Measuring performance from day one is crucial, especially if your product is costly. When ROI is significantly low and your initial impact estimates are far off—for example, when technical costs are much higher than the value the product brings back to the company—it’s a strong indicator that the product isn’t viable. Post release analysis is absolutely necessary to indentify potential improvements before deciding to retire the product.

  • The company is changing direction: Strategic decisions beyond your control can influence the survival of your product. The company may develop another product that becomes a new priority or industry standard, creating conflicts that lead to your product being deprioritized (it happens!).

  • Customer feedback is not positive: Beyond KPIs, qualitative feedback is equally important. It helps you understand the customer experience and gather ideas for future improvements.

💣 Challenges while building the product

Uncertainty is even higher when building AI products. AI development lifecycles can be longer than traditional software development, meaning it takes more time to confirm that your product works. For instance, in a team working in sprints, after two weeks, an AI team is less likely to have a meaningful increment ready, compared to a team developing a front-end feature for a web service.

There’s a high likelihood that a model won’t perform at the expected level, or that multiple iterations will be needed to achieve the required performance. This is especially critical for products that impact human lives, where high performance is essential.

Side note: I usually explain to stakeholders, new to the AI space, that we might need a few model iterations to get where we want, by showing an image of the Russian dolls, called babushka/matryoshka. We will get eventually to the biggest doll - highest increment.  


🌞 Mitigations

Failing fast and controlling the damage should always be top of mind.

To minimize risks, it’s essential to complete an opportunity assessment before starting development to estimate the potential value. Run a few scenarios—what I enjoy doing with analysts is calculating expected outcomes across three versions: optimistic, neutral, and worst-case. Then, clearly define the requirements in the PRD.

During execution, aim for small, valuable increments. Deliver value step by step. Launch experiments or PoCs. This is your reminder to narrow the scope and fully leverage Agile methodologies.

The technical team needs to challenge and provide feedback. What might seem like a small increment to you may require significant time, and you may end up building a full product instead of an MVP (been there, done that). Along with the tech lead and delivery lead—if you’re lucky enough to have one—work on a delivery plan to ensure key milestones are timeboxed and that you can deliver on time and assess early.

This is my piece of advice. It’s time to embrace the learnings and move on! 💖



Labels:
Location: Ελλάδα

Comments

Post a Comment

Popular posts from this blog

Welcome to my blog

Hi folks! I am Danai Aristeridou! I’m a strategic AI Product Owner based in Thessaloniki, Greece, with a background in data science and product management. I’m passionate about building smart, user-focused products that solve real-world problems. I love working at the intersection of tech and business — turning complex challenges into impactful, easy-to-use solutions. I created this blog to share ideas, insights, and lessons from my journey in AI and product development. Let’s connect and grow together. 😊 👉Here you can view my detailed CV
Post a Comment
Read more

Bridging the Gap: Guiding Stakeholders into the Age of AI

Bridging the gap: Guiding stakeholders into the age of AI Transitioning to the AI era isn’t always easy—especially in more traditional, old-fashioned industries. People are often reluctant to change or may believe that “if something works, why bother fixing it?” But the product lifecycle tells another story. At some point, every product will need a drastic revamp, a re-introduction to the market, or a complete phase-out. Let me walk you through my journey, and then we’ll dive into the practical aspects—which are, honestly, much more interesting! Who Am I?  👋 Hi folks! I'm Danai Aristeridou—a Product girl who landed (somewhat organically) in AI product management from a data science (DS) background. I still remember my early days in a hands-on DS role. Stakeholders struggled to understand what we were doing. There was a lack of trust and a need for control. We had to educate them, build understanding step-by-step, and iteratively move toward our goals. It wasn’t always smooth, but...
Post a Comment
Read more

Common mistakes in product discovery sessions

What I enjoy most about the role of a Product Owner is product discovery. Many of these sessions have ended in failure 😓—but looking back, I now see those failures as invaluable. Connecting the dots helped me realize that each misstep carried a lesson. Out of these experiences, I’d like to share the most common mistakes I made (and I believe many newbies do too). Leading by example in reverse !  💀 Using discovery sessions for validation: Let me explain by using a story.  When I go to yoga, my lovely instructor always says "Use your mind like a beginner. Approach each position as if you’re experiencing it for the first time." . In Greek it's called the αρχάριος νους. Αccording to  Shunryu Suzuki: In the beginner's mind there are many possibilities, but in the expert's mind there are few. I try to bring this principle into discovery sessions today—but it wasn’t always the case. In my early days, I’d tal...
Post a Comment
Read more