What to Measure Before You Have Numbers: Leading Indicators of Pre-PMF Progress

Before you have revenue or daily-active-user numbers worth trusting, track three things: whether small early cohorts keep returning without a push, how many existing users would be "very disappointed" to lose the product, and whether people describe it in their own words, unprompted. Everything else pre-PMF is progress theater.

Skip revenue and DAU pre-PMF — both are lagging and easy to manufacture. Instead track retention-curve flattening, the Sean Ellis score (roughly 40%+ "very disappointed"), and unprompted qualitative pull from real users.

Why Revenue and DAU Are Lagging Indicators Before PMF

Revenue and daily-active-user counts describe what already happened, not what's about to happen. Pre-PMF, both can be manufactured by founder effort rather than earned through demand — a single generous buyer, a launch-day spike, or a notification campaign can move either number without a single new habit forming.

That's the core problem with treating growth metrics as proof of fit: they're outcome metrics, and outcomes lag the underlying behavior that causes them by weeks or months. By the time revenue confirms fit, you've usually already spent the runway finding out the hard way. A leading indicator has to move before the outcome, not alongside it.

Consider what actually inflates each number pre-PMF:

• Revenue can come from friends, favors, a discount that overrides real willingness to pay, or one enterprise buyer humoring a founder they like.
• DAU/MAU can be inflated by onboarding email sequences, push notifications, or a paid-acquisition burst that brings in users who were never going to stick around.
• Signup totals measure curiosity, not commitment — a viral tweet or a Product Hunt launch fills a funnel with people who click once and vanish.
• Press mentions and app-store rank correlate with marketing timing and spend, not with anyone forming a habit around the product.

The table's pattern repeats everywhere: every growth metric has a retention or behavior-based sibling that answers the harder, more honest question. This is also the core argument behind why discovery should outrank delivery before you've found fit — shipping faster only compounds a metric that was never measuring the right thing. If you're a solo or early founder wearing the PM hat, this is also where a founder-led PM's first ninety days should be spent building the instrumentation habit, not chasing a growth number that isn't earned yet.

The Retention Curve: The Single Best Leading Indicator of Fit

A retention curve tells you whether usage is becoming a habit or decaying toward zero — and unlike revenue, it can't be faked by a single big customer or a marketing push. Pre-PMF, the shape of the curve matters more than its height: a curve that flattens, even at a low percentage, beats one that keeps sliding toward the x-axis.

Reforge co-founder Brian Balfour has written extensively about what he calls the "smile" or "flattening" curve: retention should decline sharply after signup, then bend and stabilize into a roughly horizontal line rather than continuing to decay. A curve that never stops falling means the product hasn't found a repeatable reason for anyone to come back — no amount of acquisition volume fixes that shape.

How to Read Your Curve Before You Have Scale

You don't need thousands of users to see this pattern; you need enough cohorts, tracked consistently, to see whether the line bends. A few practical rules:

1. Track cohorts by signup week or month, not by calendar date — mixing cohorts of different ages hides the curve entirely.
2. Plot percentage retained, not raw counts, so a cohort of 12 people is as legible as one of 1,200.
3. Watch weeks 4 through 8 specifically — most consumer and many B2B products show their first real flattening signal in that window.
4. Don't panic at a low number. A curve that flattens at 8% is a real signal; one that keeps declining from 40% is not, even though 40% looks better on day one.

Mapping this against the actual customer journey helps explain why a cohort drops where it does — a flattening curve paired with a journey map showing exactly where the emotional dip happens gives you both the "what" and the "where to fix it."

The Sean Ellis Score: Turning "I'd Be Disappointed" Into a Number

The Sean Ellis score asks existing users one question — how would you feel if you could no longer use this product? — and turns qualitative pull into a single number you can track over time. Growth advisor Sean Ellis, who coined the term "growth hacking," built the test from surveying hundreds of startups and found that crossing roughly 40% answering "very disappointed" correlated strongly with products that went on to find durable growth.

The survey itself is short and cheap to run:

• How would you feel if you could no longer use [product]? (Very disappointed / Somewhat disappointed / Not disappointed)
• What type of person do you think would benefit most from [product]?
• What is the main benefit you get from [product]?
• How can we improve [product] for you?

The number matters less than what you do with the other three answers. Rahul Vohra, founder of Superhuman, publicly documented (via First Round Review) how his team refined the raw Ellis test into what he called a "product-market fit engine": segmenting respondents by the benefit they named, then doubling down on features for the segment already closest to "very disappointed," rather than trying to please everyone at once.

The instinct when a score comes back low is to broaden the product. The instinct that actually works is to narrow it — find the users already closest to disappointed-without-you, and build harder for exactly them.

That segmentation step is where most founders stop too early. A single aggregate score of 22% tells you little; the same 22% broken out by user type might reveal one segment sitting at 55% and dragging everyone else down with it — which is a completely different, far more actionable finding.

Qualitative Pull: The Signals No Dashboard Captures

Qualitative pull is the language, timing, and behavior of users volunteering enthusiasm you didn't ask for — someone forwarding your product to a colleague unprompted, a support ticket that reads like a love letter, a user describing their problem in the exact words you used in your positioning. No analytics tool captures this by default; you have to go looking.

This is the layer that a pure dashboard obsession misses entirely, and it's why interviews stay essential long after you've instrumented the basics. Structuring those conversations around a jobs-to-be-done framework turns scattered anecdotes into a comparable signal — you're listening for the same "job" language across interviews, not just collecting isolated quotes.

Concrete qualitative signals worth logging, not just noticing:

• Users referencing your product unprompted in a different context (a tweet, a Slack message a customer shares with you, a mention in someone else's blog post).
• Someone asking "can I pay you more" for a feature you haven't built yet — a much stronger signal than a feature request.
• Support tickets or sales calls where the user teaches you your own product's value back to you in language you didn't write.
• A user churning back in after leaving — reactivation after a real gap is a stronger fit signal than never leaving at all, because it survived a deliberate choice to stop.
• Someone referring a friend or colleague without a referral incentive attached.

Paul Graham's essay "Do Things That Don't Scale" makes the case, echoed throughout Y Combinator's advice to founders, that manually talking to a handful of early users produces more useful signal pre-PMF than any dashboard — because a founder in the room catches tone, hesitation, and enthusiasm that a metrics pipeline structurally cannot. This is also where founder bias creeps in hardest: it's worth reading how founder opinion can quietly override real signal before you decide which five conversations "prove" the product works.

The Acquisition Trap: Why Growth Before Retention Backfires

Spending on acquisition before retention holds amplifies the wrong number — you're paying to fill a bucket that's already leaking, which makes the leak bigger and harder to see under the noise of new signups. Growth spend on a low-retention product doesn't buy you time; it burns runway while making the underlying curve look temporarily busier.

The mechanism is straightforward:

1. Paid acquisition adds users to the top of the funnel, which briefly lifts DAU and total signups.
2. Those users churn at the same underlying rate as your organic cohort, because acquisition spend doesn't change the product.
3. The churn is masked by the new-user spike for a few weeks, creating a false read that things are improving.
4. When the spend stops, the mask comes off, and the underlying (unchanged) retention curve is all that's left — often alongside a smaller bank account.

Marc Andreessen's original definition of product-market fit — being in a good market with a product that can satisfy that market — implies the reverse order too: you can't manufacture fit by spending harder into a bad market. Fit is something a market pulls out of a startup; it isn't something acquisition spend can push into existence.

A simple governor: don't increase paid acquisition spend materially until week-4 retention has stayed flat or improved across at least two consecutive cohorts. That single rule prevents the most common pre-PMF mistake — reading a temporary DAU bump as validation and pouring more budget behind it.

The Short List: Five Signals Worth Instrumenting First

You don't need a full analytics stack before PMF — you need five specific signals, tracked consistently, even if the "tracking" is a shared spreadsheet updated by hand. Instrumenting the wrong ten metrics is worse than instrumenting the right five, because it buries the signal that matters under dashboards that don't.

Keeping a Weak-But-Real Signal From Going Unwatched

The hardest part of this list isn't defining the five signals — it's actually revisiting them on a cadence when there's no dashboard forcing you to. A retention curve that looked promising in month one is easy to forget checking again in month three, once a founder's attention gets pulled toward a dozen other fires.

This is a scheduling problem as much as an analytics one, and it's the specific gap Prodinja's Reminders are designed to close: you can set a recurring nudge to pull up a small cohort's retention numbers on a set cadence, so a weak-but-real early signal doesn't quietly go stale between the moments you happen to remember it exists. It's not a substitute for building the analytics — it's a way to make sure the analytics you already have actually get looked at.

If you're building out this measurement discipline as a new founding PM, it's worth anchoring it to the broader operating rhythm laid out in the complete guide to founder-led product management — pre-PMF metrics work best as one thread in that larger practice, not a bolt-on report nobody reads.

Key Takeaways

• Revenue and DAU are lagging indicators pre-PMF — both can be manufactured by founder effort or spend without a single new habit forming underneath them.
• Retention-curve shape matters more than its height. A curve that flattens at 8% is a stronger signal than one still declining from 40%.
• The Sean Ellis score — the share of users who'd be "very disappointed" without your product — becomes actionable once segmented by user type, not read as one aggregate number.
• Qualitative pull (unprompted referrals, users teaching your value back to you, reactivation after churn) is a real signal a dashboard structurally can't capture.
• Don't scale acquisition spend before retention holds. Paid growth on a leaking product just makes the leak more expensive and harder to see for a few weeks.
• Five signals are enough pre-PMF: retention curve shape, Sean Ellis score, activation rate, qualitative pull events, and organic/referral share.
• Revisiting weak signals on a cadence matters as much as collecting them — a promising early cohort is easy to forget checking again once the next fire shows up.