Picking the Right Metrics

The actual job of a product manager is to pick the right metrics — not to track every number that feels important. Each metric you choose must give real meaningful insight that guides your next decision. Otherwise, you drown in noise.

Picking good metrics is not easy. It takes practice to spot the signal from the noise. Most PMs start with too many metrics and end up confused, overwhelmed, or misled. The trap is thinking every metric matters equally.

Consider a scenario: you’ve been tasked with improving the search experience on your app. What does that even mean? Search results are not displayed correctly? Results are irrelevant or plain wrong? Query suggestions don’t appear? Real-time search fails? There is an endless barrage of possible issues.

The actual job is to start with the end in mind — what user behavior or business outcome would show your search improvement is working? Then work backward to find the metrics that correlate with that outcome and can pinpoint root causes.

The trap of "all metrics matter"

When you look at your analytics dashboard, you see dozens or hundreds of metrics. Clicks, impressions, bounce rates, session lengths, conversions, time on page, filter usage, query counts, error rates... the list goes on.

It is tempting to track them all. Each metric feels like a lever you can pull. But this is a trap.

Not all metrics are equal. Most metrics are vanity metrics or noise. They don’t move the needle on your key outcomes or tell you what to do next.

You need to identify the “suspects” — the small set of metrics that are most likely to explain why your search experience is poor. Then you can focus your analysis and experimentation there.

Triangulating the root cause with metrics

Let’s walk through an example.

You discover that your app’s conversion rate from the search results page to the product details page is 10%. The industry average is 20%. That is a significant gap. You want to improve it.

What could explain this low conversion? Possible culprits include:

• Product pictures are not exciting
• Prices are too high
• Too many products on the page
• Search results are irrelevant to the query
• Filters are missing or poorly defined
• Filters are too few or not working

You cannot investigate all these at once. You need to triangulate.

If product pictures were terrible, you would notice immediately. If pictures are already good but users don’t click, improving pictures might not move the needle much.

Prices could be the issue — but only if your products are price sensitive. You can check this by plotting clicks against price. If cheaper products consistently get more clicks, price is a suspect.

You can run experiments — for example, offer discounts on selected products that rarely get clicks despite appearing in search results. If clicks shoot up, you have evidence price sensitivity is a driver.

Notice the careful constraints in this experiment. You are not discounting all products — only those that:

• Have low click rates
• Appear frequently in search results

This keeps the experiment focused on the search experience problem.

Too many products on the page is unlikely to cause a 10% drop in conversion. It typically causes users to drop out after a few clicks, not immediately.

Search results being irrelevant is a big suspect. How do you test that?

Imagine a user session where the search results are poor. The user searches, scrolls, clicks a product if it looks relevant, then either they are not satisfied or want more options, so they tweak the query and search again. This repeats.

If you look at the number of queries per session plotted against the number of products clicked, you might see a high number of queries but very few clicks. That signals poor search relevance.

You can define a metric:
Ratio of queries per session to number of clicks — the higher this ratio, the worse the search results.

This is how you carefully select metrics that matter and derive insight.

Someone might argue: users don’t click because they don’t like the products shown, not because search is bad. That is valid. You then define a different metric to assess product attractiveness.

The moral: picking good metrics requires meticulous thinking. They can make or break your business.

Metrics are hypotheses, not facts

Every metric you track is a hypothesis about what matters to your users or the business. You must treat metrics as starting points for investigation, not as absolute truths.

For example, if you see low clicks from search results, you hypothesize it is due to price sensitivity. You design an experiment to test that hypothesis (discount some products) and observe results.

If the experiment disproves your hypothesis, you try another — maybe search relevance is the issue.

This cycle of hypothesis → measurement → experiment → learning is the core of data-driven product management.

Business metrics vs KPIs

In practice, you will encounter two categories:

• Business metrics: Quantifiable measures used to track business processes, like website traffic, session duration, or number of sign-ups. These tell you what is happening.
• Key Performance Indicators (KPIs): Metrics that track critical areas of performance tied directly to business goals, like conversion rate, retention rate, or revenue growth. These tell you how well you are doing against targets.

The line between the two blurs in practice. Your job is to focus on KPIs that align with your product goals and use other metrics as supporting signals.

Context is everything when choosing metrics

Metrics have no meaning in isolation. Their value depends on context:

• What is your product’s core value?
• What user behavior signals success?
• What business outcomes do you want to impact?
• What baseline or industry benchmarks exist?
• What constraints or user segments matter?

For example, a 10% conversion rate on search results might be bad in a category with low-price, high-volume products but acceptable if your products are luxury or high-value items.

You must understand your product context deeply to pick the right metrics.

The practice of metric selection

Talvinder often says: “A pragmatic product leader is constantly measuring. He doesn’t do anything which cannot be measured or has some means of measuring the impact.”

You should bake metrics into your product design and experiments from the start.

Some practical guidelines:

• Pick a small number of metrics (3-5) that matter most for your current objective.
• Define metrics clearly and unambiguously.
• Ensure metrics are actionable — they should guide decisions or experiments.
• Avoid vanity metrics that look good but don’t influence outcomes.
• Continuously revisit and refine your metrics as you learn more.
• Use a combination of qualitative and quantitative data to interpret metrics.

Experiment design with metrics

Metrics are central to experiments that validate product changes.

When you run an experiment, define:

• The metric you will measure to assess impact
• The segment of users you will test on
• The time period for measurement
• The hypothesis you are testing

For example, if you hypothesize that discounts increase clicks on low-performing products in search results, your experiment might be:

• Metric: Click-through rate (CTR) on discounted products
• Segment: Products with under 5% CTR appearing in top 10 search results
• Time: 2 weeks
• Hypothesis: Discounting these products will increase CTR by at least 20%

This focused approach avoids noise and helps you learn efficiently.

Indian market context: the search problem

In India, search experience is especially critical. Users often have low attention spans and high expectations for relevance.

Indian e-commerce companies like Flipkart and Meesho invest heavily in search relevance and filtering because poor search kills conversion.

When you pick metrics, consider local factors like:

• Regional language queries and code-switching
• Mobile device constraints
• Price sensitivity and discount culture
• Diverse user literacy and digital skills

Metrics that work in Silicon Valley may not apply as-is to Indian users. You must adapt.

Summary: The honest truth about metrics

• You cannot track everything. Focus on a few key metrics that matter.
• Good metrics require thoughtful definition and context.
• Metrics are hypotheses to be tested, not facts to be accepted blindly.
• Experiments anchored in metrics are how you learn and improve.
• The trap is confusing volume of data for insight.

Your job is to pick the right metrics, interpret them carefully, and drive decisions that move your product forward.

Test yourself: Diagnosing search conversion issues

Where to go next

• Understand how to translate metrics into product decisions: Product Analytics Fundamentals
• Learn to design effective experiments: Experimentation and A/B Testing
• Master user research methods to complement metrics: User Research Methods
• Explore how to build product vision and strategy from data: Product Vision and Strategy
• Deepen your understanding of KPIs in SaaS and e-commerce: KPIs for SaaS and E-commerce