The Four Steps to Product Discovery
How teams approach product discovery varies and depends on several factors ranging from product roadmap, organizational bureaucratic structures, experimentation models and rigidity to the founding mission amongst other things.
So how should you approach product discovery as a platform product manager? These steps shared below are a combination of various steps already identified by companies like Mixpanel, Toptal, NFX, Google, Amazon, and Patreon.
- Understanding the Stage of a Startup: understanding what stage you’re in as a startup is crucial to understanding what growth levers to pull and when. A typical startup life cycle is the progression of its phases over time and is most commonly divided into five stages: launch, growth, shake-out, maturity, and decline.
Knowing where you are in this cycle is important in creating a platform that spans multiple product lines and enables you to make strategic decisions and trade-offs across different products.
2. Challenging assumptions: there is a very thin line between business ignorance and business intelligence. Typically product decisions are composed of a myriad of expectations that you can extrapolate from past experience, but for new products, this way of planning is often ineffective and misleading. Early-stage startups struggle to build experimentation into their product decisions as a result of maintaining a superior assertion that doesn’t fit changing needs of users.
There is much knowledge behind your assumptions especially when you are building products for millions of users that are influenced by different social and cultural factors. Using a “one size fits all” blanket is the bane of disaster for product teams. With enough empirical research, you can shatter your assumptions and in turn make good product decisions.
3. Conducting empirical market and user research — innovative solutions often start with a genuine market need. Successful startups don’t just stumble on a gold mine, it takes years of insightful data to identify trends of behaviors and unspoken needs. Capturing effective feedback from your customers, however, is just the starting point. To identify truly actionable data with an on-market horizon of 3 to 5 years requires a more thoughtful approach.
Creating a ‘fail-fast’ experimentation approach requires a large amount of data, without data it is difficult to imagine the future. To help you better conduct an effective research as a platform product manager its important that you take a collaboration first approach with business, engineering, and design teams.
Understanding changing consumer needs and preferences, and industry shifts can be approached from a two-sided spectrum as primary and secondary research. With an exploratory primary approach, you can gather lots of open-ended data from many customers to better understand a problem or opportunity. The challenge here is how do you qualify and tailor a mix of observational and unstructured data. You can contextualize your findings within a band of what is technologically and commercially practicable.
Once you understand the larger market issues or opportunities, you can use a secondary research approach to gather information about certain markets and competitors that could guide your execution.
4. Build a custom experimentation model— so now you have data that your competitors probably don’t have access to. Running multiple experiments better enlightens your decision-making process and it’s important when you have access to multiple nodes of conflicting data. It provides the framework to conduct commercial experiments necessary to verify such early business judgments at scale.
Experiments allow you to uncover surprises that could redirect or refine your research. It also brings the technical feasibility of an idea into focus very early in the product decision process.
A scientific approach to creating an experimentation platform emphasizes an iterative learning process, practically when conducting experiments it’s better to model your system against a “deduction and induction” model.
Deduction is a process of going from an idea or theory to a hypothesis to actual observations/data that can be used to test the hypothesis, while induction is the process of generalizing from specific observations/data to new hypotheses or ideas. Experimentation plays a critical role in collecting data to test hypotheses and enabling the deduction-induction iterations as part of the scientific method. —
The quasi and causal inference framework used by Netflix is a functional example of an experimentation platform as a powerful way to let data guide your decision-making. One area Netflix actively applies experiment to is changes to the quality of streaming experience. In such situations, they run a quasi-experiment and apply causal inference techniques to determine the impact of the change.
Building this type of model requires effective and open collaboration across product and engineering. As a platform product manager, you are required to foster such an environment where data and experimentation are cardinal pillars in your decision-making process.