Critical metrics every product manager must track

Quantitative data is all the rage these days: numbers aren’t just useful, they make us look competent. Better yet, they comfort us and give a sense of control. Data has become the opium of the modern tech companies. Want to look good in a meeting? Just open up a chart and say loudly: “I’ve got the data!” Do that often enough, and promotion is yours. That is, if you don’t drive your company to the ground. Because we can collect so much data about virtually everything people do in our software, we cripple ourselves with information that doesn’t matter. Making the wrong decisions has never been easier.

This post is an attempt to cut through the noise and save time for those who decide which metrics are best to gauge the health of their product or company. The metrics I list here belong to one of the three major categories: user engagement, business, and customer service.

User engagement metrics

1. Number of sessions per user

This metric is a good starting point for user engagement and behavior analysis. Besides, there’s nothing simpler than tracing how often your users log in or open your app.

Tip: Avoid averages and take medians over means, as they are less sensitive to outliers and give more robust statistics.

2. Session duration for a cohort, over time

For every new cohort of users, research how much time they spent interacting with your product. Again, take medians instead of means.

Tip: Compare your average session duration for churned and retained users for some eye-opening data. (E.g., those who churned within 30 days and those who stayed.)

3. Number of key user actions per session

Start with selecting user actions that matter most (e.g., clicks on the “Like” button), then trace them over time and for different cohorts of users.

Tip: Compare the differences in this metric between churned and retained customers.

Pro tip: Always use cohort analysis and time normalization for user engagement metrics. This will help you expose the differences in user behavior and their evolution over time.

Business metrics

1. LTV, or Customer Lifetime Value

LTV is the amount of net profit you will generate from a customer before they churn or stop paying.

Common mistake: Many companies end up calculating Customer Lifetime Revenue, instead of LTV. Calculate LTV correctly by excluding the cost of servicing a customer: salaries of the support team, the cost of running your servers, and so on.

Tip: If you want to start a boardroom fight, ask your investors about the correct way to calculate LTV. :)


2. CAC, or Customer Acquisition Cost

Take all your acquisition marketing costs, then divide by the number of paying customers acquired over a period of time.

Another way to compute CAC is to include salaries of your acquisition marketers, commissions of the sales reps, etc. into the calculation. This will make CAC partially account for the burn rate. LTV and CAC are the most important metrics of startup unit economics, and they are usually considered together as LTV/CAC ratio.

Benchmark: For a healthy SaaS startup, LTV/CAC ratio should be 3 and above.

3. ARPA, or Average Revenue per Account

ARPA is the worth of monthly “contract” with the customer or user. To put it simply, this is how much money an average paying customer gives you every month. If your pricing is stable, ARPA growth means your team is doing something right to provide customers with more value and generate revenue in return.

Tip: Try increasing your ARPA without changes in product or pricing. Often, better support, marketing, and branding can do exactly that.

4. MRR, or Monthly Recurring Revenue

MRR is self-explanatory, and this is simply your company’s revenue per month. Stable, and preferably non-linear, MRR growth is the best way to make your investors happy.

Pro tip: Make your investors happy.

5. Logo churn and Revenue churn rates

Logo churn rate, is the percentage of paying accounts your product loses per month. There are two ways to calculate it: per cohort and aggregated. Typically, cohort churn will yield higher numbers, so internally you should definitely use it instead. Revenue churn rate is the percentage of revenue you company loses per month due to churns or downgrades.

Tip: Pay less attention to logo churn and more attention to revenue churn. All in all, serving fewer customers and getting to pay more is better than serving a bunch and getting peanuts.

Benchmark: For a SaaS startup, your monthly logo churn rate should be below 2%. Your revenue churn should be negative.

6. Retention rate

Contrary to popular belief, retention rate isn’t as useful of a metric as churn rate. It’s good to be aware of it but it’s far less actionable, compared to other metrics.

Customer Service

1. Number of incoming support tickets

This metric is rather obvious: the more problems your users have with your product, the more tickets they submit. If you’ve just released something, and the number of tickets spiked, this might be a sign of a problem.

2. Net Promoter Score

NPS is a proxy metric for customer satisfaction. However, before relying on NPS in your decision-making, you must know that it’s not a scientifically-backed metric. For instance, a meta-analysis by Keiningham, et al. found no support for the claim that Net Promoter Score is a reliable indicator of a company’s ability to grow (A Longitudinal Examination of Net Promoter and Firm Revenue Growth, Journal of Marketing, 2007, American Marketing Association).

Tip: Having high NPS is meaningless in itself. Instead watching it grow, try reaching out to promoters and asking them for reviews. You’ll be surprised by results.

Congratulations if you made it this far! Metrics above only scratch the surface of data-aware product management. If you want to use the data right, you will want to experiment with different types of analysis for different features of your product, in different stages of its life cycle. There’s only one rule to be aware of at all times:

The fact that something has a number attached to it, doesn’t always make it true, valid, or relevant.

This post is an expansion of my old and outdated answer on Quora (which also got quoted on UserVoice). Current version of this post was previously published on Medium.