How to Write PRDs for Product Managers
Defining Goals and Success Metrics
Goals Are Not the Same as Metrics
This distinction trips up even experienced product managers, and it matters enough to address directly.
A goal is the outcome you are trying to achieve. A metric is how you measure whether you achieved it.
- Goals are directional and qualitative. Help new users understand the value of completing their profile. Reduce the friction in the payment flow. Make it easier for teams to collaborate on shared documents. These statements describe an intended change in the world. They do not, by themselves, tell you whether the change happened.
- Metrics are specific and quantitative. They convert the goal into something measurable. Profile completion rate among users in their first seven days. Payment flow drop-off rate at the card entry step. Number of documents shared per active team per week. These are the numbers you will look at after launch to determine whether the feature worked.
Both are required. Goals without metrics cannot be evaluated. Metrics without goals can be optimised in ways that miss the point entirely.
Writing Goals That Are Honest
Goals in PRDs are often written to sound impressive rather than to be useful. Delight users with a seamless experience is not a goal. It is a feeling. Becoming the market leader in collaborative productivity is a strategic ambition that a single feature cannot deliver. Both types of goal-writing share the same failure: they cannot be evaluated.
A useful goal is one that describes a specific, observable change in user behaviour or experience that the feature is designed to produce.
Ask yourself: if this feature succeeds, what will be different about how users experience this product? What will they be able to do that they cannot do now? What will they stop struggling with? What will they do more of, or less of, or differently?
The answers to those questions, written clearly and specifically, are your goals.
The Four Properties of a Good Success Metric
Not every number is a useful success metric. A success metric that does not genuinely reflect whether the feature achieved its goal is worse than no metric at all, because it creates false confidence.
A good success metric is specific, measurable, attributable, and time-bound.
Specific means the metric measures exactly what the goal intends, not a proxy that is merely correlated with it. If the goal is to help users understand the value of profile completion, the metric should measure profile completion rate or downstream outcomes associated with profile completion, not general engagement or session length.
Measurable means the data needed to compute the metric actually exists or will be collected. A metric that requires data the team cannot access is not a metric. It is a wish.
Attributable means you can reasonably attribute changes in the metric to the feature rather than to external factors. This is where A/B testing and proper experiment design become necessary. A metric that moves for reasons unrelated to the feature tells you nothing useful about whether the feature worked.
Time-bound means you know when you will evaluate the metric. Immediately after launch? After 30 days? After 90 days? The right timeframe depends on the metric. Conversion rate changes may be visible within days. Retention impacts may take months to manifest. Setting the evaluation timeframe in advance prevents the motivated reasoning that leads teams to keep measuring until they see a result they like.
Primary and Secondary Metrics
Not all metrics carry equal weight. A well-structured success measurement framework distinguishes between primary and secondary metrics.
Primary metrics are the ones that directly measure whether the feature achieved its core goal. They are the metrics that would cause the team to conclude the feature succeeded or failed. A feature typically has one, or at most two, primary metrics.
Secondary metrics provide additional context and help the team understand the shape of the impact. They also serve as guardrail metrics: measures of things you do not want to harm in the pursuit of improving the primary metric. If improving the primary metric comes at the cost of a significant degradation in a secondary metric, the feature has not truly succeeded.
Defining guardrail metrics before launch is one of the marks of a mature evaluation approach. It prevents the selective interpretation of results that leads teams to celebrate wins that came at a hidden cost.
Remember this: Goals describe the intended change. Metrics measure whether the change happened. A good success metric is specific, measurable, attributable, and time-bound. Every primary metric needs secondary and guardrail metrics to prevent optimising for the wrong thing.