How do you plan to spend your traffic?

For product managers and performance marketers who get asked whether it worked.

Earlycall checks an experiment before you run it — can your traffic answer the question at all, and what does being wrong cost you? Free, no signup

A real week: 13.7M visits spent on a question that had no answer in it

Is yours any different? Two numbers will tell you.

Visitors or sessions a week

How many convert today (%)

The design review five questions · nothing stored

A product change, a landing page, an ad creative, an email — anything you will be asked to justify afterwards.

What do I get at the end?

A one-page decision record: the question in plain words, the metric that decides it, how long it needs, what you agreed to do at each outcome, and what would change your mind — written down before the data arrives, which is what makes it worth anything afterwards. Copy it into a ticket, or send it to whoever asks why you shipped.

Read a complete example →

Step 1What are you trying to move? What would you call this piece of work? However you'd refer to it in a standup. And what are you trying to find out?

Step 2Where could you move it?

One is fine. Several, and you find out which of them your traffic can actually answer a question about.

+ Add a safety-net number (optional)

The business number that must not get worse A safety net, not the decider — catching harm needs far less traffic than proving a gain.

Its current rate

Where do I find this?

GA4: Reports → Engagement → Conversions, divided by sessions for the same period. Shopify: Analytics → Conversion rate. Or whatever number your weekly report already quotes.

Total visitors, per week

Where do I find this?

GA4: Reports → Acquisition → Traffic acquisition, set the range to Last 7 days and read Sessions. Site-wide, not just the page you are changing.

When the test ends, which number decides it? The most expensive mistake in experiment design, and the easiest to fix: a change only a slice of visitors ever see, judged on a number everyone contributes to.

Step 3How big a change would you make? Be honest — this is the single thing teams overestimate most.

And what kind of product is this? It decides which corpus we compare you against, and it matters more than it sounds: headline tests on viral content move the needle roughly 60× harder than e-commerce checkouts. Using the wrong one is how a hopeless test gets told it looks healthy.

How much do you expect it to move the metric? (optional) Your own history beats our corpus. Leave it blank and we use a band from other people's experiments — and on most bets, filling this in is the single change that turns "fix it first" into "run it".

How many versions will you run against the current one? Counting the current version. Two means a straight A/B.

Step 4What happens if you’re wrong? If it turns out badly, how long to undo it? Who gets hurt if it’s bad? How quickly would you notice a problem? Which is bigger — the upside or the downside?

Step 5Your bar

Smallest improvement worth shipping Below this you wouldn’t bother, so proving it isn’t worth traffic. %

You need an answer within days

Nothing is stored.