Product Metrics a Seed-Stage Founder Should Watch

A seed-stage startup should track five numbers: new signups per week, activation rate, time to first value, week-four retention for each signup cohort, and the weekly count of outcomes users accept. Only the first one lives on your website. The other four happen inside the product, after someone creates an account, and most seed founders have none of them on the page they open every Monday.

That gap decides what you fix. At seed you have room for a handful of real bets before the money runs short, and the numbers you look at each week choose those bets for you. If every number sits before signup, every fix will too. These five numbers move the bets past the signup button.

A Monday page with four green arrows and nothing after signup

Picture an AI bookkeeping tool for freelancers and two-person studios. You connect a bank account. It sorts every transaction into a tax category. At the end of each month it shows a screen headed Review March, with a list of anything it was unsure about and one green button at the bottom: Approve month. Pressing it means that month of books is done. That button is the whole reason the product exists.

Every Monday at nine, the founder opens a Notion page called Monday numbers. It has four rows, each pasted in from a different browser tab. MRR, $6,420, from Stripe. Website visitors, 3,180, from Plausible. Signups this week, 212, from a count on the users table. X followers, 1,904. Each row has a small green arrow beside it, because each one went up.

Now read the page again and notice where it ends. Not one row describes anything a person did after making an account. The page cannot say how many of the 212 connected a bank, how many saw a sorted month, or how many ever pressed Approve month. In this example the answer to the last question is 19, and nobody on the team knows it.

Why the numbers you see first all stop at the signup button

Nobody chose this page. The tools chose it. Stripe shows revenue the day you take your first payment. A website analytics script counts visitors the minute it is pasted into the header. X counts your followers for you. Signups is one database query. Every number past signup is different, because someone has to decide what to count and then add an event to the product code before a single data point exists.

So that work waits. In this example there is a ticket called Add product events, opened in July and moved to next sprint three times, because it has no visible deliverable. In standup the founder reads the four arrows aloud. The investor update leads with visitors. When visitors climb and revenue does not, he rewrites the homepage headline, because the homepage is the only thing his page points at. Website conversion is worth knowing, and there is a fair benchmark for it in our post on SaaS website conversion rate. It is still one number, and it ends exactly where your product begins.

This is the quiet trap at seed. The website is measured and the product is not, so all the evidence points at the half that is probably fine. A founder who acts on that evidence is not being careless. He is being sensible about the only data he has.

The five numbers, and the single event behind each one

The test for a seed metric is simple. It has to point at a screen you could change this week. If you cannot name the screen, the number is trivia. Here are the five, using the bookkeeping tool as the example.

• New signups per week. The one website number worth keeping. Count each week on its own, never a running total, and split out the one channel you pay for if you pay for any. The screen behind it is your homepage and your signup form.

• Activation rate. The share of a week's signups who reach the first action that predicts they stay. Here it is approving a first month within two weeks of signing up. Choosing that event is its own job, covered in how to define activation for an AI product. For this page you only need it written down.

• Time to first value. The median hours from account created to that activation event. For the bookkeeping tool, it is the gap between signing up and first pressing Approve month. If that gap is measured in days, ask what the user is waiting for, and whether you made them wait.

• Week-four retention by cohort. Of everyone who signed up in a given week, how many did the core action again in their fourth week. One row per signup week. This is the start of the retention curve an investor will ask to see, and daily active users cannot stand in for it.

• Weekly accepted outcomes. The total count of the thing users came for, accepted by them. Months approved, not transactions sorted. This is your north star by another name, and the north star examples for AI products show why it should count accepted work, not model work.

Four of those five happen after signup. That ratio is the point. At seed, the product is where people decide, so most of your Monday attention belongs there.

What the four green arrows cost him

Every one of those 212 signups was paid for, in ads, in content or in the founder's own hours on LinkedIn. The cost lands whether or not anyone presses Approve month. An AI product pays twice, because every bank that got connected also ran the model over months of transactions. AI product builders average around 52 percent gross margin against 70 to 80 percent for traditional software, so there is less room to absorb signups that go nowhere.

Benchmarks measured on SaaS and product-led companies put activation at 20 to 40 percent for most products. In this example, 19 out of 212 is under one in ten, and the Monday page cannot show it. The same SaaS data says a ten point gain in activation typically drives a 15 to 25 percent rise in free-to-paid conversion. That is the cheapest growth on offer, and it is invisible from a page that ends at signup.

Then there is the clock. Across all sectors, the median gap from seed to Series A is around 616 days, and roughly 70 percent of seed companies never raise one. For AI startups the Series A bar is around 3.5 million in ARR, up from about one million three years earlier (Carta, Q1 2026). A quarter spent rewriting the homepage is a quarter spent against that bar. And when a partner asks for the cohort chart, the founder says he will follow up, then spends a Saturday building it by hand from a database export.

What to stop looking at on Monday

Most lists of seed-stage metrics give you fifteen or twenty: CAC, LTV, net revenue retention, burn multiple, NPS, DAU over MAU. That is bad advice for a company with a few hundred users. Half those numbers are forecasts wearing a metric's clothes. A lifetime value built on four months of data is a guess. An NPS from eleven survey replies is an anecdote. A long list also hides the five that matter in a crowd of numbers you cannot act on.

Take these off the Monday page:

• Follower counts on any platform. They describe your marketing, not your product.

• Total signups and total users. A running total only goes up, so it can never warn you.

• Website visitors, unless you are testing one specific channel this month.

• Page views, time on site and bounce rate, which belong in a homepage review, not a weekly product check.

• Tokens, messages and model runs. Those are your cost line. A confused user generates more of them than a happy one.

• LTV to CAC, until you have at least a year of real customers behind it.

MRR and runway stay in your life, but move them to the finance sheet. They are results of the five, and they move a month or more after the five do. Watching them on Monday is watching last month.

Building the one-page Monday view this week

None of this needs a data team. It needs one afternoon of a developer's time and one honest sentence from you.

• Write your activation event as one sentence with a time limit. For the example, approved a first month within two weeks of signup.

• Add three events to the product: account created, the activation event, and the core outcome event. Any analytics tool will do, or a plain table in your own database.

• Make a cohort sheet. One row per signup week, with columns for signups, activated, median hours to activation, and still active in week four.

• Replace the Monday page with five rows and three columns: this week, last week, and the four-week average. Next to each row, name the screen that moves it, such as Connect your bank or Review March.

• Every Monday, open five accounts from last week that signed up and did not activate. Find the last screen each one reached and write it down. After a month you will have a list, and it will repeat.

That last step is the one founders skip, because it has no deliverable. It is also the only one that tells you why a number moved.

Reading the five rows together

Single numbers rarely tell you what to do. Pairs usually do. Signups up while activation falls means a new channel is bringing the wrong people, or the first screen loses the ones it brings, so do not raise ad spend yet. Activation steady while week-four retention slides means the first visit works and the second one does not, which is a different design problem, and our guide to SaaS user retention UX covers it. Long time to value with low activation means too many steps stand before the first outcome. In the bookkeeping tool, that could mean showing a sorted sample month before asking anyone to connect a bank. Outcomes flat while users grow means the people who stayed have stopped getting much from it.

Each of those readings points at a screen. None of them points at the homepage headline, which is the useful surprise.

At Studio Maydit we design the stretch between signup and the first outcome for AI and SaaS founders in the US, UK and Europe, which is exactly the stretch a Monday page that ends at signup never shows. Our fixed-scope product design projects run three to four weeks and end with a diagnosis of what is leaking in the product, so you leave knowing which screen your missing users stopped on and what to change first. If your Monday numbers all sit before the signup button, book a 30-minute call and bring the page with you.