git:20260729.4a3a4d0 to git:20260729.8dd5ad0

---

name: activation-aha-moment

description: "The statistical method for finding a product's activation/aha moment — the action that, taken early, separates retained users from churned ones (the \"kink in the retention curve\"). Distinguishes setup-moment vs aha-moment vs habit-moment, and how to find and validate the metric rather than guessing it. Used by analytics-analyst and metrics-architect."

---

## Codex host binding

- Treat references to Claude slash workflows as the equivalently named Codex skill.

- Before delegating to any specialist, read the `great-pm-runtime` skill and the selected packaged role file.

- Treat "invoke", "assign", "delegate", "spawn", and source Agent-tool instructions as a required Codex `spawn_agent` call with that role and a bounded assignment.

- Store every returned agent identifier. Never call a wait tool until a spawn has returned an identifier, and wait only on identifiers returned by successful spawns.

- If `spawn_agent` is unavailable or a spawn fails, report BLOCKED; do not impersonate the specialist or wait on an empty agent set.

- Resolve bundled paths from the installed GreatPM plugin root.

- Ignore Claude-only model aliases, colors, turn limits, and tool allowlists.

- Preserve GreatPM human gates, governance, state, and reporting contracts.

> **Provenance.** great-pm-original, authored 2026-05-29, grounded in the

> cited sources (Reforge, Aakash Gupta, Brian Balfour). Host agents:

> analytics-analyst, metrics-architect.

# Activation & the Aha Moment — finding it statistically

Activation is the single highest-leverage point in a consumer funnel:

the moment a user **first experiences the core value** of the product.

Everything downstream — retention, expansion, referral, revenue — is

gated on users reaching it. A product with a leaky activation step

cannot be fixed by pouring more acquisition into the top; you are just

filling a bucket with a hole in it.

The mistake almost everyone makes is **declaring the aha moment by

intuition** ("obviously it's when they invite a teammate") and then

optimizing onboarding toward a guess. The discipline this skill teaches

is the opposite: the aha moment is a number you **find in your data** —

the early action that puts a measurable *kink in the retention curve* —

and then *pressure-test for causation* before you bet the onboarding on

it.

## 1. The three activation states (Reforge)

Activation is not one event; it is a sequence of three distinct states.

Confusing them is the most common analytical error.

| State | Definition | The question it answers | Example (Slack) |

|-------|-----------|--------------------------|------------------|

| **Setup moment** | User has completed the steps *required to be able to* experience value | "Are they technically ready?" | Created a workspace, invited the team |

| **Aha moment** | User has *first experienced* the core value | "Did the value land?" | The team actually sent messages and replied |

| **Habit moment** | User has established a *recurring usage pattern* — value is now a routine | "Will they keep coming back?" | 2,000 messages sent across the team |

- **Setup ≠ aha.** A user can complete every onboarding step (setup

done) and *never feel the value* (no aha). "% who finished onboarding"

is a setup metric masquerading as activation. This is why onboarding-

completion charts look healthy while retention bleeds.

- **Aha → habit is where retention is won.** The aha proves value *once*;

the habit moment is the point at which usage becomes self-sustaining.

Slack's famous "2,000 messages" is a *habit* threshold, not the first

aha — by then a team is hooked.

- Your activation *metric* is usually the **aha moment**, expressed as a

measurable event within a time window. The setup and habit moments

bracket it: setup is what onboarding must deliver to *reach* the aha;

habit is what retention work must drive *after* it.

## 2. Finding the aha quantitatively — the kink in the retention curve

The aha moment is **the early action most correlated with long-term

retention.** Operationally: find the action where users who did it (in

their first session / first N days) retain *materially better* than users

who didn't. That gap — visible as a kink or fork in the retention

curve — is your signal.

The classic results everyone cites came from exactly this analysis:

- **Facebook** — "7 friends in 10 days." Users who reached 7 friends in

their first 10 days retained dramatically better; below it, they

churned.

- **Slack** — "2,000 messages sent" (team-level habit threshold).

- **Dropbox** — put at least one file in one folder on one device.

- **Twitter/X** — follow ~30 accounts.

These are not guesses. They are the output of a correlation sweep. Here

is the method, step by step:

### The analysis (6 steps)

1. **List candidate actions.** Enumerate every meaningful early action a

user can take — not vanity events. For each, also decide the

plausible *time window* (first session, first 24h, first 7 days). The

window matters: "7 friends" was worthless without "in 10 days."

2. **Build the did-vs-didn't comparison.** For each candidate action ×

window, split users into two groups: those who performed it early and

those who didn't. (Restrict to users old enough to have a retention

outcome — e.g., only users who signed up ≥ 30 days ago if you measure

D30.)

3. **Measure downstream retention for each group.** Compute D7 / D14 /

D30 retention (or your product's natural cadence) for did vs didn't.

The **size of the gap** is the candidate's strength.

4. **Find the kink.** Rank candidates by retention lift. Also look for a

**threshold** within a count-based action — plot retention against

"number of X done in first week" and find where the curve bends

(1 friend → little lift; 7 → big lift; 15 → flattens). The bend is the

target number.

5. **Pressure-test for causation.** This is the step amateurs skip.

Correlation is necessary but not sufficient — high-intent users do

*everything* more, so the action may be a *symptom* of an already-

engaged user, not a *cause* of engagement. Tests:

- **Is the action plausibly causal?** Does experiencing it actually

deliver value, or is it just something engaged people happen to do?

- **Does it survive controlling for general activity?** Compare users

matched on overall early activity who did vs didn't do *this

specific* action.

- **The gold standard: an experiment.** Nudge a random subset toward

the action in onboarding and see if *induced* completion lifts

retention. If it does, it's causal. (Hands to `experiment-design`.)

6. **Lock the metric + minimize Time-to-Value (TTV).** Once you have a

validated aha (action + count + window), the activation *rate* =

% of new users who reach it. Then the entire onboarding job becomes:

**compress the steps and time to that aha.** Faster TTV → higher

activation → higher retention.

### Qual + quant, always

The number tells you *what* action; qualitative work tells you *why it is

the value*. Interview users who hit the aha and users who stalled just

before it. If you can't articulate, in a sentence, the value the user

*felt* at the aha, you have a correlation you don't understand — and

you'll optimize the wrong thing. (e.g., "7 friends" wasn't about the

number; it was the point at which the feed had enough content to be

worth returning to.)

## 3. When to use — and when NOT to

| Use it when… | Because… |

|---------------|----------|

| Retention is the strategic problem | Activation is the upstream lever |

| You're about to "optimize onboarding" | You need a *target* for onboarding first |

| Someone declared an aha by intuition | It needs a statistical kink test |

| Defining metrics pre-build (metrics-architect) | Decide *candidate* activation events to instrument now |

| Post-launch with ≥ a few hundred users + one retention cycle of history | The correlation sweep needs enough N and time |

**When NOT to use:**

- **Too little data.** With < a few hundred activated users or < one full

retention cycle of history, the curves are noise. Use qualitative

discovery and `hooked-ux` reasoning to *hypothesize* an aha instead;

come back to the statistical method when data accrues.

- **No retention problem.** If retention is already excellent and stable,

finding the aha is academic — spend effort elsewhere (acquisition,

monetization).

- **One-time-use or transactional products** (e.g., a tax-filing tool

used once a year). "Retention" isn't the right frame; use completion /

task-success metrics instead.

- **Don't substitute this for the fix.** Finding the aha is diagnosis;

use `funnel-diagnostics` to find *where* users stall before it and

`experiment-design` to *validate the fix*.

## 4. Worked example — Acme

**Context.** Acme's core promise (instructions.md §1) is "crystal-

clear visibility into cash flow." The strategic North Star is retention /

money-awareness habit. So the aha is almost certainly *the moment a user

first sees their own spending, accurately categorized, in a clear view* —

but we must find it, not assume it.

**Step 1 — candidate actions** (with windows), instrumented as events:

- `first_source_connected` (CSV / PDF / Gmail) — first 24h

- `first_statement_parsed` (transactions extracted) — first 24h

- `first_insight_viewed` (saw the categorized cash-flow view) — first session

- `first_override_saved` (re-bucketed a transaction — §7) — first 7 days

- `second_source_connected` — first 7 days

**Step 2–4 — did-vs-didn't on D7/D30.** Suppose the sweep shows:

| Candidate action (in window) | D30 retention: did | didn't | Lift |

|------------------------------|-------------------:|-------:|-----:|

| Connected a source | 41% | 9% | ×4.6 |

| Statement parsed | 44% | 8% | ×5.5 |

| **Viewed first categorized insight** | **52%** | **11%** | **×4.7** |

| Saved a first override | 61% | 22% | ×2.8 |

| Connected a 2nd source | 58% | 30% | ×1.9 |

Two candidates jump out: **"viewed first categorized insight"** (big lift,

large group) and **"saved a first override"** (highest absolute retention,

smaller group).

**Step 5 — causation test.** "Saved an override" has the highest

retention but is suspicious: only already-engaged users who *care enough

to correct* a category do it — it may be a *symptom* of an engaged user,

not a cause. "Viewed first categorized insight" is plausibly causal: it

*is* the value delivery (the §1 promise), and a larger share of users can

reach it. Pressure-test: control for early activity → the insight-view

lift survives; the override lift shrinks toward the engaged-user

baseline. An onboarding experiment that *induces* reaching the insight

view lifts D30 → confirms causation.

**Decision.** Activation metric = **% of new users who view their first

accurate categorized cash-flow insight within the first session.** The

override is a *deepening* signal (a habit-direction metric), not the aha.

**Step 6 — minimize TTV.** Now onboarding's entire job is "get every new

user to that first insight, fast": pre-seed a sample statement so a user

sees a populated view *before* connecting anything; one-tap CSV import;

make Gmail-sync a fast path; show a partial categorized view the instant

the first few transactions parse rather than waiting for the whole file.

Every step removed between signup and that insight raises activation —

and, because of the kink, raises D30 retention.

## 5. Common pitfalls

| Pitfall | Why it fails | Fix |

|---------|--------------|-----|

| Declaring the aha by intuition | You optimize onboarding toward a guess; if wrong, you compress TTV to the wrong destination | Run the correlation sweep; find the kink in data |

| Confusing setup completion with the aha | "% finished onboarding" rises while retention bleeds — value never landed | Measure value *experienced* (aha), not steps *completed* (setup) |

| Mistaking correlation for causation | High-intent users do everything more; you chase a symptom of engagement, not its cause | Control for general activity; confirm with an onboarding nudge experiment |

| Ignoring the time window | "7 friends" is meaningless without "in 10 days"; an action eventually done by everyone has no predictive power | Always pair the action with a window; find the count threshold |

| One aha for segments that activate differently | New SMB vs enterprise, CSV-uploader vs Gmail-sync user may have different value moments | Segment the analysis; allow per-segment activation metrics |

| Quant with no qual | You know the action but not the value it delivers, so you can't redesign around it | Interview users at and just-before the aha; name the felt value in one sentence |

| Stopping at diagnosis | Finding the aha doesn't move retention by itself | Hand to funnel-diagnostics (where users stall) + experiment-design (validate the fix) |

## 6. Ethical boundary

Optimize for the user reaching **genuine value faster**, not for a

metric-defined event that *looks* like value. Inducing the aha-action via

dark patterns (forced friend-invites, manipulative nudges) inflates the

activation number while degrading the experience — and the retention lift

evaporates because the *felt value* never happened. The kink only holds

when the action causes real value. If you can't name the value the user

feels at the aha, you don't have an activation metric — you have a vanity

trap with a confidence interval.

## 7. Quick diagnostic

| Question | If "no" → |

|----------|-----------|

| Do you have a *named* activation event (action + count + window)? | Run §2 correlation sweep |

| Is it the **aha** (value experienced), not **setup** (steps done)? | Re-define around value, not completion |

| Have you tested it for **causation**, not just correlation? | Control for activity; run a nudge experiment (experiment-design) |

| Can you state, in one sentence, the value the user *feels* at the aha? | Do the qualitative pass |

| Is onboarding ruthlessly optimized to minimize TTV to that aha? | Map the pre-aha funnel (funnel-diagnostics) and cut steps |

| Did you check whether segments activate differently? | Segment the sweep |

## 8. Cross-links & boundary

- **`metrics-design`** — names "activation" in one line of AARRR; this

skill is the *method to find the actual activation metric*. Use

metrics-design to place activation in the full KPI tree.

- **`north-star-input-tree`** — activation rate is almost always one of

the 3–5 input metrics under a retention North Star. Build the aha here,

slot it into the tree there.

- **`improve-retention`** — broad retention *tactics*; this skill finds

the *upstream activation lever* those tactics should target first.

- **`hooked-ux`** — habit-loop design (trigger→action→reward→investment);

complements the *habit moment* (§1) and helps *hypothesize* an aha when

you lack data. This skill is the quantitative validation layer.

- **`cohort-analysis`** — produces the retention curves you sweep across;

the kink shows up *in* cohort retention data.

- **`funnel-diagnostics`** — once you know the aha, this finds *where*

users drop before reaching it.

- **`experiment-design`** — the causation gold standard: the onboarding

nudge experiment that proves the aha *causes* retention.

- **`brainstorm-okrs`** — activation rate makes an excellent input-metric

OKR once validated.

**Differs from `improve-retention` because** it doesn't list retention

tactics — it isolates the *one early action* that predicts retention.

**Differs from `cohort-analysis` because** that skill *computes* the

curves; this skill *interprets the kink* into a decision. **Differs from

`metrics-design` because** that defines a metric set; this discovers the

single most important one.

## Sources

- Reforge, "Define your aha moment" — https://www.reforge.com/guides/define-your-aha-moment

- Reforge, "Define user activation moments" (setup / aha / habit) — https://www.reforge.com/guides/define-customer-activation-moments

- Aakash Gupta, "Ultimate Guide: Activation" — https://www.news.aakashg.com/p/ultimate-guide-activation

- Brian Balfour on product habits & retention (churn.fm) — https://www.churn.fm/episode/product-habits-retention

- Chamath Palihapitiya / Andy Johns on Facebook's "7 friends in 10 days" growth analysis (widely documented; the canonical kink-in-the-curve case)