Memorisely
Product Management

Define success before you build

Choosing north star and guardrail metrics so you know if a feature worked before anyone writes code.

Zander Whitehurst

7 min read

Most product teams can tell you what they are building. Far fewer can tell you how they will know if it worked. The feature ships, a launch post goes out, and a few weeks later someone asks whether it made a difference. Nobody is quite sure, because nobody decided in advance what difference it was meant to make.

Defining success before you build changes that. It turns a launch from a finish line into an experiment, and it gives the team a clear way to learn whether their work actually mattered.

Why success gets defined too late

Teams tend to leave metrics until the end because building feels more urgent than measuring. There is always a deadline, and setting up tracking feels like something that can happen later.

The problem is that later rarely comes. When it does, the team picks whichever numbers look best. A small rise in page views becomes evidence of success, and nobody mentions that the metric that really mattered barely moved. Without an agreed definition upfront, every launch can be spun as a win, which means the team never really learns anything.

Start with a north star metric

A north star metric is the single number that best captures the value your product delivers to customers. It sits above individual features and gives every team a shared direction.

Good north star metrics share a few qualities:

  • They reflect customer value. They go up when customers get more out of the product, not just when the business extracts more from them.
  • They predict long term success. Growth in the metric should lead to growth in revenue and retention over time.
  • They are easy to understand. Anyone in the company should be able to explain what the number means.

For a collaboration tool, it might be the number of teams actively working together each week. For a learning platform, it might be the number of lessons completed. Individual features should be able to explain how they contribute to that number, even if indirectly.

A north star captures customer value, and every feature should explain how it contributes.

Add guardrail metrics

Chasing a single number can lead you somewhere unhealthy. Push notifications might increase daily active users while quietly driving people to unsubscribe. A more aggressive upgrade prompt might increase conversions while increasing refunds.

Guardrail metrics protect against this. They are the numbers you do not want to get worse while you improve your main target. Common guardrails include:

  • Support ticket volume.
  • Churn or cancellation rate.
  • Page load time and error rates.
  • Customer satisfaction scores.

Decide on your guardrails before launch, alongside your success metric. If a feature hits its target but breaks a guardrail, that is not a win. It is a trade off the team needs to discuss openly.

Guardrails sit alongside the target, so a win on one number cannot hide damage elsewhere.

Leading versus lagging indicators

Some metrics take months to move. Retention, revenue and lifetime value are the numbers that really matter, but they are lagging indicators. By the time they change, it is too late to adjust course quickly.

Leading indicators move sooner and predict the lagging ones. If you know that customers who invite a colleague in their first week are far more likely to stay, then first week invites become a leading indicator for retention.

A useful metrics plan pairs the two:

  1. A lagging indicator that represents the outcome you ultimately care about.
  2. One or two leading indicators that you can observe within days or weeks.
  3. A hypothesis that connects them, such as "if more new users complete setup, more will still be active after ninety days".

Leading indicators let you learn fast. Lagging indicators keep you honest about whether that learning is translating into real results.

Set targets you can defend

A metric without a target is just a number to watch. A target tells the team what good looks like and when to celebrate or rethink.

Targets do not need to be perfect, but they should be grounded. Look at your current baseline, check how similar changes have performed in the past and consider what would make the work worth the effort. A target of "increase activation" is vague. A target of "increase activation from 32% to 38% within six weeks of launch" is something you can actually judge.

A grounded target starts from today's baseline and names a number and a date.

Be honest about uncertainty. For a genuinely new idea, it is fine to set a range or a minimum threshold instead of a precise number. What matters is writing it down before you see the results.

A simple metrics plan

You do not need a complicated document. A short metrics plan at the top of the project brief is usually enough:

Goal: [what we want to change for customers]
Success metric: [the number that tells us it worked]
Baseline: [where it is today]
Target: [where we want it to be, and by when]
Leading indicator: [what we expect to move first]
Guardrails: [what must not get worse]
Review date: [when we will look at the results]

This takes about twenty minutes to write. It saves far more than that in arguments after launch. We build plans like this into every PRD in our AI Product Management Bootcamp.

Use AI to sanity check

AI tools are a great second opinion on a metrics plan. Share your plan and ask the AI to challenge it:

  • Could this metric go up without customers getting more value?
  • What guardrails might be missing?
  • Is the link between the leading and lagging indicator believable?
  • Is the target realistic given the baseline and timeframe?

AI is particularly good at spotting ways a metric could be gamed or misread, because it will happily imagine scenarios the team has not considered. It is less good at knowing what your data actually looks like, so treat its suggestions as questions to investigate, not answers to adopt.

AI is good at spotting how a metric could be gamed, as long as you treat it as a question.

Share it with the whole team

A metrics plan only works if designers and engineers know about it. When the whole team understands what success looks like, they make better decisions in the small moments that never reach a planning meeting.

A designer choosing between two layouts can ask which one is more likely to move the success metric. An engineer can flag that a technical shortcut might affect a guardrail like load time. Shared metrics turn the whole team into decision makers, not just the PM.

I learned this the hard way on a project where we spent a quarter rebuilding an onboarding flow. The design was beautiful and the launch went smoothly. When I asked how we would measure it, three people gave me three different answers. One was looking at completion rate, one at time spent and one at upgrades. All three numbers moved in different directions, and we spent a fortnight arguing about whether it had worked. On the next project, the metrics plan came before the first design file was opened, and the review took fifteen minutes.

Review it honestly

When the review date arrives, look at the numbers you agreed on, not the ones that happen to look best. If the target was hit, great. If it was missed, that is valuable too. It tells you something about your customers that you did not know before.

The teams that improve fastest are not the ones that always hit their targets. They are the ones that define success clearly, measure it honestly and use what they learn to make the next decision better.

Keep reading

View all
View all

Optimize your AI for better experiences, not AI slop.

View all

AI Prototyping Skills

Prototype with AI using clean UI patterns and guardrails so flows feel product-grade, not generic template slop.

$99

AI UI Motion Skills

Motion and micro-interaction patterns that make AI-built UI feel designed, polished, and trustworthy in real products.

$99

Fix AI Slop Skills

Spot AI slop early and fix type, spacing, and components so your product looks intentional, not like a template.

$99

AI Responsive Design Skills

Responsive layout patterns for rails, sidebars, and breakpoints so AI-generated UI adapts cleanly from narrow to wide.

$99

AI Dark Mode Skills

Dark theme tokens, contrast, and surfaces that keep AI-generated UI readable and product-grade, not flat gray on black.

$99

AI Color Token Skills

Neutral and primary ramps with clear steps so AI palettes stay systematic, not a random set of accent hex codes.

$99
View all