I once sat in a pitch meeting where a founder pulled up a dashboard with 34 different metrics. Revenue, page views, session duration, bounce rate, email open rate, social impressions, feature adoption for twelve separate features, and a handful of custom scores that even he could not explain on the spot. The investor across the table listened politely, then asked one question: "What percentage of users come back after the first week?" The founder did not know. That meeting lasted eight more minutes.
Most early-stage founders track too many metrics or the wrong ones entirely. They set up Mixpanel or Amplitude on day one, create a dozen dashboards, and then spend hours every week staring at numbers that tell them almost nothing about whether their startup is actually growing. The problem is not a lack of data. It is a lack of focus.
If you are building an AI startup and you are pre-Series A, there are exactly seven numbers you need to care about. Everything else is noise until you have proven that people want what you are building and that you can get more of those people efficiently.
Why Most Dashboards Are Full of Vanity Metrics
A vanity metric is any number that goes up and to the right but does not correlate with sustainable growth. Total signups is the classic example. It feels great to watch that counter climb, but if nobody sticks around after day two, you are filling a leaky bucket. Page views, total downloads, social media followers, email list size — these are all vanity metrics at the early stage.
Here is my rule of thumb: if a metric can only go up, it is probably a vanity metric. The metrics that matter can go up or down, and the direction tells you something actionable.
The 7 Metrics That Actually Predict Growth
1. Activation Rate
Activation rate measures what percentage of new signups complete the action that correlates with long-term retention. For an AI writing tool, that might be generating their first document. For an AI data analysis platform, it might be connecting a data source and running their first query. The key is identifying your activation event — the moment where the user first experiences the core value of your product.
How to measure it: Pick your activation event. Count the number of users who complete it within their first 48 hours divided by total new signups. A healthy activation rate for an AI product is 40-60%. Below 25% and you have an onboarding problem that needs fixing before you do anything else.
I helped one AI startup raise their activation rate from 18% to 52% by doing one thing: reducing the steps between signup and first output from seven to two. That single change did more for their growth than three months of marketing experiments. I talk about this more in my piece on product-led growth for AI products — the activation moment is everything.
2. Week-1 Retention
Of all the users who sign up this week, what percentage come back and use the product at least once in the following seven days? This is the single most important metric at the pre-product-market-fit stage. Everything else is secondary.
How to measure it: Cohort analysis. Take all users who signed up in a given week. Seven days later, count how many logged in or performed a meaningful action. Divide. For AI products, good week-1 retention is 30-45%. If you are above 50%, you likely have strong product-market fit signals. Below 20%, your product is not solving a real problem or is not solving it well enough.
Week-1 retention is the metric I look at first when I start working with a new startup. If it is below 20%, I do not touch marketing. We fix the product first.
3. Organic Growth Rate
What percentage of your new users come from channels you are not paying for? This includes word of mouth, direct traffic, organic search, and referrals. Organic growth rate tells you whether your product is generating its own momentum.
How to measure it: Tag your acquisition sources. Count users from non-paid sources divided by total new users. At the pre-Series A stage, you want this above 40%. If 90% of your users come from paid ads or manual outreach, you have not built something people talk about yet.
When I was helping one AI startup figure out their first 1,000 users, we noticed that 60% of signups in week three were coming from sources we could not trace — meaning people were sharing the product directly. That was the signal that told us to stop experimenting with new channels and start doubling down on what we had.
4. Time-to-Value
How long does it take from the moment a user signs up to the moment they experience the core value of your product? For AI products specifically, time-to-value should be under 5 minutes for most use cases. The best AI products deliver value in under 60 seconds.
How to measure it: Timestamp the signup event and the activation event. Calculate the median time between them. Median, not average — you do not want outliers distorting the picture.
5. Revenue Per User
If you are charging, this is straightforward: monthly recurring revenue divided by active users. If you are pre-revenue, substitute a proxy: how much usage does each active user generate? The number itself matters less than the trend. Are power users emerging who use the product significantly more than the median? Those power users are your product-market fit signal. Talk to them.
6. CAC Payback Period
How many months does it take for a customer to generate enough revenue to cover what you spent acquiring them? At the early stage, most of your acquisition cost is your time, so be honest about valuing it.
How to measure it: Total acquisition cost (ad spend + tools + your time valued at a reasonable hourly rate) divided by number of customers acquired = CAC. Then divide CAC by monthly revenue per customer = payback period in months. Before Series A, investors want to see payback under 12 months. Under 6 months is excellent. If you are building your go-to-market strategy, this is the number that determines which channels are sustainable.
7. NPS / Referral Rate
I actually prefer a harder metric than NPS: what percentage of your users have actually referred someone? NPS measures intent. Referral rate measures behavior. At the early stage, behavior is what matters.
How to measure it: If you have a referral program, check the data. If you do not, send a simple one-question survey to active users: "Have you told anyone else about [product]?" A referral rate above 15% is strong. Above 30% is exceptional and usually means you have nailed product-market fit.
Which Metrics Matter at Which Stage
Pre-Product-Market Fit
Focus on three metrics only: activation rate, week-1 retention, and time-to-value. These three tell you whether the product works. Nothing else matters until they are healthy.
Do not waste time optimizing CAC or referral rate when your retention is 15%. You are trying to fill a bucket that has no bottom. Fix the bucket first.
Post-Product-Market Fit
Once your retention curve flattens (meaning some cohort of users sticks around indefinitely), shift focus to: organic growth rate, revenue per user, CAC payback period, and referral rate. These tell you whether you can scale the thing that is working.
The One Metric That Matters Framework
Even within these seven metrics, trying to improve all of them simultaneously is a mistake. At any given time, you should have one metric that matters — one number that the entire team is focused on moving. This is the metric where improvement will have the biggest cascading effect on everything else.
Usually it is obvious. If your activation rate is 15%, that is your OMTM. If activation is strong but week-1 retention is 10%, retention is your OMTM. If both are healthy but you are spending $500 to acquire each customer, CAC is your OMTM.
The power of the OMTM framework is not just focus. It is alignment. When everyone on the team knows that this month we are trying to move week-1 retention from 25% to 35%, every decision gets easier.
How to Present Metrics to Investors
Investors at the pre-Series A stage are looking for one thing: evidence that you have found something that works and that it can scale. They do not want 34 metrics. They want a story told in three to five numbers.
- Retention curve. Show a cohort chart. If the curve flattens, you win.
- Growth rate. Month-over-month user or revenue growth. Show the trend line over at least three months.
- Unit economics. CAC, revenue per user, and payback period. Even rough estimates show that you are thinking about sustainability.
- Activation rate. This shows you understand your product and your users.
- One qualitative metric. A quote from a user, a referral story, a screenshot of someone raving about your product. Numbers tell the story. Qualitative evidence makes it real.
Do not hide bad numbers. Investors have seen thousands of decks. They can spot when you are cherry-picking. Instead, show the bad number and explain what you are doing about it. "Our week-1 retention was 18% in January. We identified onboarding friction as the cause, shipped three changes, and February cohorts are retaining at 31%." That story is more compelling than a deck full of perfect numbers that nobody believes.
Common Mistakes That Kill Early-Stage Growth
Optimizing for the Wrong Metric
The most common version: spending all your energy on acquisition when your retention is broken. I see it constantly. Founders who run ads, do cold outreach, post on every social platform, and burn through their runway getting signups — while 90% of those signups churn within a week.
Measuring Too Early
If you have 30 users, your metrics are meaningless. A single power user or a single churned user will swing your retention rate by 10 percentage points. You need at least 100 to 200 users in a cohort before the numbers start to stabilize. Before that threshold, talk to users directly.
Dashboard Paralysis
I have worked with founders who spend two to three hours every morning reviewing dashboards. They can tell you their metrics to two decimal places but they have not shipped a product update in three weeks. Metrics are a tool for making decisions, not a substitute for making decisions. Check your numbers once a week. Spend the rest of your time building, talking to users, and running experiments.
The antidote to dashboard paralysis is the OMTM framework. One number. One focus. One weekly check-in to see if it moved. Everything else is a distraction.
Comparing Yourself to Benchmarks That Do Not Apply
A 60% week-1 retention rate is incredible for a complex B2B analytics tool and mediocre for a consumer social app. Benchmarks only make sense within your category, your stage, and your business model. The best benchmark is your own number from last month. Are you improving? That is what matters.
If you are in the thick of this right now — trying to figure out what is working, what to measure, and where to focus — start with week-1 retention. Measure it for your last three cohorts. If it is below 20%, stop everything else and fix the product. If it is above 30%, congratulations — you probably have something real, and it is time to figure out how to pour fuel on it.