The Metrics Worth Tracking in Launch Week

Launch week produces a flood of numbers. Page views, referrers, sign-ups, comments, social impressions, a ranking, perhaps a spike on the server graph. Most of it is easy to measure and weakly connected to whether the business is working. The discipline is to decide, before launch, which handful of metrics will tell you the truth.

This article proposes a small set, explains how to define each so it cannot be gamed by your own optimism, and says what to do with the results.

Principles

1. Decide in advance. Pick the metrics before you see the data. Otherwise you will choose the ones that look best.
2. Prefer ratios to totals. A thousand visits with a tiny activation rate is worse than two hundred with a high one.
3. Segment by source. The same metric means different things from different channels.
4. Define each precisely. Write the event, the window and the denominator.
5. Keep it small. Five to eight numbers fit on one screen.

The core set

1. Qualified visits

Visits from people who look like your buyer. Define "qualified" in terms you can measure: arrival on a product page, time on page past a threshold, a view of pricing or documentation. Total visits include casual browsers and bots.

2. Sign-up rate

Sign-ups divided by qualified visits. Track it by landing page and by source. A low rate suggests a message or friction problem. A very high rate from a single source suggests that audience is a strong fit.

3. Activation rate

The share of new sign-ups who reach your first moment of real value within a defined window, for example completing a first import or sending a first message within twenty-four hours. This is the most important number in launch week, because it tells you whether the product delivers to strangers.

Define the activating event with care. "Logged in twice" is not activation. "Created and shared a first report" might be.

4. Time to first value

Median minutes from sign-up to the activating event. If it is long, the first-use path needs work. Look at the distribution, not only the average, since a few very slow cases can hide a good median.

5. Source quality

For each channel, compare sign-up rate and activation rate. A channel that brings fewer people but a higher activation rate may be worth more effort. Attribution depends on tagging links. Google's campaign URL guidance says to use utm_source, utm_medium and utm_campaign always, to keep values lowercase and consistent and to use unique values for distinct platforms, so that a campaign is not split across spellings.

6. Pipeline signals

For business products, count conversations that could become revenue: demo requests, replies to your outreach, trial-to-contact conversions, pricing page views from known accounts. These lead revenue, so track them from day one.

7. Qualified lead rate

Of the inbound enquiries, how many match your target profile? A launch that attracts the wrong audience can still look healthy on traffic.

8. Page health

Core Web Vitals on your launch pages. Google's guidance on web.dev says a good experience is LCP within 2.5 seconds, INP of 200 milliseconds or less and CLS of 0.1 or less at the 75th percentile of page loads. Monitor them during launch week, since slow or shifting pages lose visitors before any other metric shows it.

What to ignore, or read with caution

Raw page views. They include bots and curiosity.

Social impressions. They measure the platform's distribution, not your result.

Rank on a leaderboard. Useful as a signal of attention, not of business health.

Total sign-ups. Without activation, a vanity number.

Average session duration. Easily distorted.

Comments counts. Look at the content of the comments, not the count.

Define each metric in writing

For each metric, fill in a small spec.

• Name.
• Event or query. What exactly is counted.
• Window. Hours, days.
• Denominator, if a ratio.
• Segment. By source, by plan, by device.
• Owner. Who checks it.
• Target. What would count as good, written before launch.

A shared document keeps everyone honest. If you change a definition mid-week, note it.

Set targets with ranges

Write three numbers for each ratio: a disappointing level, an acceptable level and a strong level. Base them on whatever evidence you have: earlier tests, similar products, your funnel in beta. If you have none, say so and treat launch week as the first measurement. Targets do not predict, but they stop you from reinterpreting results after the fact.

Build one screen

Create a single dashboard with the core set, refreshed at least hourly during launch day. Keep it plain.

• Funnel: qualified visits to sign-ups to activations.
• Source table: visits, sign-up rate, activation rate by channel.
• Time to first value distribution.
• Pipeline signals count.
• Page health indicators.
• Error rate and uptime.

If a number on the screen does not drive a decision, remove it.

Read the funnel from the bottom

Start with activation. If it is low, nothing upstream matters yet. Fix the first-use path. If activation is healthy but sign-up rate is low, work on the page. If both are healthy and visits are low, work on distribution. Reading upward prevents you from pouring traffic into a leaky bucket.

Watch the freemium trap

If your product has a free tier, be careful with sign-up and activation counts. Wikipedia's discussion of freemium notes the central tension of balancing the free tier so it attracts and retains users without reducing the perceived value of the premium tier. A launch can produce many free activations and few conversations about paid use. Track conversion to paid or to a sales conversation separately and keep expectations realistic about timing.

Cohorts

Group sign-ups by day and by source and follow each group over the following weeks. A cohort table shows whether Monday's sign-ups behave differently from Thursday's and whether the contest audience retains differently from the newsletter audience. Cohorts take a few weeks to mature, so mention in your launch review that the retention picture is incomplete.

Qualitative signals

Numbers do not capture everything. Keep a log of:

• Questions that arrive repeatedly.
• Objections raised by prospects.
• Words customers use to describe the product.
• Feature requests and their sources.

These are as valuable as any metric. Tag them by theme and count them.

Guard against self-deception

• Do not move the goalposts. Keep the pre-written targets.
• Beware survivorship. Active users are not all users.
• Beware small samples. Twenty sign-ups cannot support a ratio to two decimal places.
• Check for bots and tests. Exclude internal and automated traffic.
• Check the instrumentation. An event that never fires makes a metric zero.
• Share bad numbers too. They are the useful ones.

After the week

Write a one-page review: the metrics, the targets, the outcomes, the surprises and the three actions that follow. Archive it with your dashboard definitions. When you launch again, you can compare like with like.

On this site

Profiles on this site include analytics context for listed products. The submit page starts an entry, a launch entry currently costs $5, and our live Ahrefs DR is shown (currently 0) so you can judge the referral context. The about-founders page introduces the people behind the project.

A worked example of reading a week

Imagine a team launches an expense-reporting tool. At the end of the week the dashboard shows that two hundred and forty people visited the product page from qualified sources, sixty-one signed up and twenty-two reached the activating event, which they defined as submitting a first expense report within a day. Those figures give a sign-up rate and an activation rate you can compute at a glance. The source table shows that a newsletter brought few visitors but the highest activation, while a launch contest brought many visitors and a low activation rate. Time to first value has a median of eleven minutes, with a long tail caused by people who stopped at the step where they connect a bank account.

The team draws three conclusions. First, the bank connection step should become optional, since it delays value. Second, the newsletter audience is a better fit, so they will invest in similar newsletters. Third, the contest brought attention but not many buyers; they will keep it as an awareness channel, not an acquisition one. None of these conclusions required fancy analysis, only a few well-defined numbers read in the right order.

Instrumentation checklist

Before launch, verify each event fires exactly once per action, with the right properties: source, plan and device. Test with a throwaway account and look for the event in the analytics tool within minutes. Confirm that internal traffic is excluded. Check that consent handling follows your policy where applicable. Record the exact names of events in a shared document so that everyone describes them the same way. A day spent on this before launch saves weeks of argument afterwards.

The takeaway

Define a small set of metrics before launch, centre them on activation, segment by source with consistent tags, set written targets and read the funnel from the bottom up. Then the week gives you information you can act on.