Most early dashboards lie to you. Not because the tools are broken—because you do not have enough traffic for percentages to mean anything, and founders still stare at them like they are a weather report.
You open Mixpanel or a free GA4 property, see twelve sessions, a 3% "conversion rate," and a bounce chart that looks like a heart attack. Then you spend half a day debating whether the signup form needs another A/B test. With under a hundred users, that is theater. The numbers are too thin to decide with, but thick enough to distract you from the three questions that actually matter.
This post is for indie hackers who want product metrics that drive decisions before the first 100 users—not a SaaS growth stack that assumes you already have volume. If you are still hunting for those early buyers, pair this with how to get your first 10 paying customers without a launch day. Metrics without people is just a spreadsheet hobby.
What metrics are supposed to do before 100 users
Before you pick tools, pick jobs. At this stage metrics only earn their keep if they help you do three things:
Prove activation. Someone new can reach the outcome your product exists for without you sitting next to them.
Catch silent churn in a tiny cohort. The five people who loved you last week are still around—or you know why they vanished before you invent a retention campaign.
Decide what to ship next. Not from a feature voting board. From the friction and themes you see in usage and conversations.
If a chart does not serve one of those jobs, it is decoration. Pageviews, referrer pie charts, and "engagement score" composites can wait. You are not running a media site. You are trying to learn whether a small group of humans gets value twice.
Vanity metrics feel productive because they move. Decision metrics feel boring because they often sit still until you fix the product.
Metrics that matter before you have traffic
Keep the list short. Four signals beat twenty half-tracked events.
1. Activation: time-to-first-value and the core job
Define one "aha" in plain language. Not "completed onboarding." Something like: created their first report, invited a teammate, generated the first export a client could use, or finished the workflow that replaces the spreadsheet they hate.
Track two numbers by hand if you have to:
Activation rate: of people who signed up in the last 14 days, how many hit that aha within 24–48 hours?
Time-to-first-value: median minutes (or days) from signup to that aha for the ones who got there.
If activation is weak, stop optimizing launch posts. The same instinct shows up when people soft-launch an AI product before public launch week: prove a small set can get value before you scale the invite list. Empty funnels are usually activation problems wearing a marketing costume.
A useful related habit: obsess over the first ten minutes in the product, not the hero section copy—the argument behind fixing the first 10 minutes instead of the landing page.
2. Weekly active among people who activated
Raw DAU with 40 users is noise. Better: of the people who ever activated, how many came back in the last 7 days and did the core job again?
That ratio tells you whether value sticks. If five people activated and four came back, you have a product problem worth shipping for. If forty signed up and two activated, you have an onboarding problem—and your "retention" chart will scare you for the wrong reason.
3. The qualitative "would you notice if this died?" test
Once a week, ask a few activated users one blunt question: if this product disappeared tomorrow, how annoyed would you be—shrug, mild inconvenience, or real pain?
You are not hunting for a Net Promoter Score. You are hunting for language. "Mild inconvenience" plus no workaround stories means you are a nice-to-have. "I already told my cofounder we need this" means you are closer to something that can convert later. Write the quotes down. They become social proof later—and if you are stuck writing that proof with almost no logos, see how to write social proof for an indie launch when you have almost no users.
4. Support and interview themes
Tag every support message and call note with 1–3 themes: confused by setup, missing export, pricing anxiety, "where do I put my data," etc. After two weeks, sort by frequency.
Themes are leading indicators. They show up before a churn chart does. Three people asking the same "where is X" question is a product decision. One person asking for a niche integration is a maybe-later sticky note.
Metrics that can wait (seriously, wait)
Founders waste months dressing up pre-product-market-fit companies like Series B analytics orgs. Park these until you have real volume or a clearer ICP:
Funnel conversion % with n=12. A 25% vs 33% signup rate on twelve visitors is a coin flip, not a strategy.
CAC and LTV. You do not have stable acquisition channels or enough paid months. Rough cost per conversation is fine; LTV models are fan fiction.
Fancy cohort charts. Weekly cohorts of three people look profound in Amplitude and mean nothing in a board memo to yourself.
Heatmaps on empty pages. Watching three sessions rage-click your pricing page is not research. Sit on a call instead.
North-star vanity composites. "Engagement score" that mixes sessions, clicks, and emoji reactions. You will game yourself.
Ignore does not mean forever. It means: do not let these steal the weekly review while activation is still shaky.
A founder-friendly stack that will not drown you
You do not need five tools. You need a habit.
Option A — spreadsheet + honesty. Columns: user, signup date, activated? (Y/N), first-value date, last core action, source (friend / tweet / directory), notes. Update it when someone signs up and once a week on Friday. Ugly. Accurate enough.
Option B — one simple event table. PostHog, Plausible custom events, or a tiny table in your own DB. Cap yourself at 3–5 events:
signed_upactivated(your aha, fired once)core_action(the repeatable job)invited_teammateorshared_output(optional buying signal)canceledorchurned_manualif you already charge
Do not instrument every button. Every extra event is a future lie you will misread.
Weekly review ritual (30 minutes, calendar it):
How many new signups, how many activated, median time-to-value?
Of activated users, who went quiet? Message two of them personally.
Top support/interview themes—pick one friction to ship against next week.
One sentence: "Next week we will know we improved if ___."
If the ritual takes two hours, your stack is too clever. Shrink the events until the review fits in a coffee.
How to read noisy numbers without lying to yourself
Small n makes founders dramatic. Three users bounce and you delete a feature. One power user loves a dead-end workflow and you build a roadmap around them. Both mistakes come from treating sparse data like truth.
Rules that keep you sane:
Do not kill a feature after three users. Watch for a pattern across two weeks and at least a handful of people in your ICP—or a clear qualitative reason ("everyone who tries this gets stuck on step 2").
Triangulate with conversations. A dip in core_action plus two emails saying "I got busy" is different from a dip plus "I could not find how to export." Same chart, different fix.
Separate curious clickers from activated users before you talk about retention. Otherwise you will "fix retention" by blaming people who never got value.
Prefer rates inside the activated set over sitewide percentages. "4 of 7 activated users returned" is a decision. "12% weekly active of all signups" hides whether activation ever happened.
Write the decision before you open the dashboard. "If activation < 40% this week, rewrite empty state." Looking at charts without a question is doomscrolling with SQL.
Noise is normal. Your job is not to eliminate it. Your job is to refuse to ship panic based on it.
A 14-day metrics sprint checklist
Use this if your analytics setup is a junk drawer or a blank page.
Day 1–2: Define
Write the one-sentence aha / core job.
List the three jobs metrics must serve (activation, silent churn, ship next).
Pick spreadsheet or one event tool—not both.
Day 3–5: Instrument lightly
Ship 3–5 events max. Test them yourself end-to-end.
Add a simple admin view or sheet of last 30 users with activated flag.
Start a themes column for support notes.
Day 6–10: Talk to humans
Message five activated users: would-you-notice question + one open "what almost made you quit?"
Watch two people go through signup to aha on a call (or Loom). Note where they hesitate.
Do not add new events yet. Fix the worst friction you saw.
Day 11–14: Review and decide
Run the 30-minute weekly ritual twice (end of week 1 and week 2).
Write one page: activation rate, return rate among activated, top themes, next ship.
Delete or ignore any chart you did not use in those reviews.
At the end of fourteen days you should know whether people reach value, whether they come back, and what to build—not whether your bounce rate looks "industry standard."
Ship distribution after activation is real
Metrics are not a substitute for customers. They are a way to stop guessing which customers are actually getting value before you pour energy into directories, launch weeks, and reply-guy marketing.
When activation is consistently happening for the people you care about, then turn the volume up. Until then, treat every new acquisition channel as more people you might confuse. Get the first-value path honest, keep a short list of signals, review weekly, and let conversations veto the charts when the sample is tiny.
That is product metrics before 100 users: fewer numbers, clearer jobs, and enough discipline to ignore the rest.
