AARRR · Product Diagnosis · Checklists · PTOS

AARRR Clinic: A Step-by-Step Protocol for Product Problem Diagnosis

A detailed protocol for conducting an 'AARRR Clinic'—a session for rapid product problem diagnosis, identifying 'leaks,' selecting the right analytical lenses, and formulating tests with guardrails.

AARRR Clinic: A Step-by-Step Protocol for Product Problem Diagnosis

Your product is growing, but something isn't right? Metrics show alarming signals, but it's unclear exactly where the "leak" is and why? The "AARRR Clinic" is a structured diagnostic protocol that will help your team quickly localize the problem, choose the right analytical "lenses," and formulate specific tests with guardrail metrics.

The goal of this session is to emerge in 60–90 minutes not with "just another idea," but with 1–2 testable experiments for the week.

Input Data (without which everything else is guesswork)

Before starting, ensure you have:

Segment: A clear understanding of whom you are diagnosing for (not "all users").
Scenario / Context: In what situation a person "hires" your product.
Value-event: One specific action that signifies the user "received value."
Retention Window: Defined timeframes for retention (D7/D30 or W1/W4).
Event Dictionary: Strict definition of key events (what counts, what doesn't).

Step 0. Triage (5 minutes)

A quick check for readiness to diagnose. If there's no clarity on these points, stop and define the value-event. Otherwise, you'll be optimizing noise.

Is the Value-event defined in one phrase?
Is Activation = the first value-event?
Is Retention = repetition of value-event within the timeframe?
Are there "dead metrics" (like "opened the app") that don't reflect value?

Artifact: 3 lines: value-event, activation, retention-window.

Step 1. Localize the "leak" using AARRR (10–15 minutes)

Identify at which stage of the AARRR funnel (Acquisition, Activation, Retention, Referral, Revenue) the first massive drop-off occurs—where the proportion of users sharply declines.

Mini-rule:
- If the leak is before value-event → the problem is more often in Activation / Task Success.
- If the leak is after value-event → the problem is more often in Retention / repeating context.

Artifact: "AARRR skeleton"—one main event for each funnel pocket.

Step 2. Select Lenses (5 minutes)

After localizing the problem, choose appropriate analytical "lenses" to gain a deeper understanding of "why" and "how." Not "what to do," but "how to look."

HEART (UX quality): when you need to understand how well a key task is performed (Task Success, Time-to-success, Happiness).
CJM (Customer Journey Map): when you need to uncover barriers, doubts, expectations at a specific step of the journey.
JTBD (Jobs-to-be-Done): when you need to understand why a person chooses/doesn't choose this behavior (Push/Pull/Habit/Anxiety forces).
Cohorts: when Retention is "average across the board" and it's unclear what truly changes over time.
RFM (Recency-Frequency-Monetary): when there are purchases/payments and the quality of the base is important (recency, frequency, monetary value).
Unit Economics / CLV: when any "victory" in AARRR might turn out to be economically toxic.
Growth loops / Flywheel: when growth has stalled and a compound mechanism is needed, not "just another channel."

Artifact: A list of selected lenses with a brief justification.

Step 3. Problem Breakdown (20–30 minutes)

Choose one point of failure and uncover the mechanism.

If failure in Activation:
- Use HEART: Task Success + Time-to-success.
- Or CJM: 5–7 steps of the scenario before the first value-event (barriers/doubts/errors).
If failure in Retention:
- Use Cohorts: retention over time + breakdowns by channels/segments/first scenario.
- Or JTBD: repeating context ("when... I need... so that...") and forces (Habit/Anxiety).

Artifact: 3 cause-hypotheses (not "solutions"), ranked by probability.

Step 4. Test for the Week + Guardrails (10–15 minutes)

Formulate a minimal test that can be conducted within a week, and define guardrails—protective metrics.

Test format:
- Hypothesis: "People don't do X because Y."
- Change: "We will reduce Y like this..."
- Success Metric: (1) primary (e.g., activation-to-value), (2) guardrail (to avoid breaking retention/economics).
Guardrails are mandatory: They will protect you from improving one metric at the cost of worsening others (e.g., "Do not worsen Retention," "Do not drop margin/payback").

Artifact: 1 test + 1 guardrail (minimum).

Step 5. Session Output (2 minutes)

The session is considered successful if you have obtained:

1 localized "leak" in the AARRR funnel.
3 cause-hypotheses.
1 test for the week.
1 guardrail.
(Optional) 1 next research step.

Symptom → Lenses → Artifact → Test → Guardrail Matrix

This matrix will help you quickly choose the right tools depending on the symptoms:

1) Symptom: "Traffic/registrations are growing, but no value/money"

Where in AARRR: Acquisition ↑, then drop-off in Activation/Retention/Revenue.
Lenses: Cohorts + HEART (Task Success) + Unit Economics.
Artifact: AARRR skeleton + cohort retention by channels.
Test: Disable/reduce a "bad" channel for 7 days or change traffic qualification (message/offer).
Guardrail: Retention D7 does not drop; CAC/payback does not worsen.

2) Symptom: "Many visits/installs, but Activation is low"

Where: Activation.
Lenses: HEART (Task Success, Time-to-success) + CJM + JTBD (expected first progress).
Artifact: CJM for 1 scenario up to the first value-event.
Test: Reduce steps to value-event / remove the main blocker (form, KYC, import, permissions).
Guardrail: Quality of result (errors/failures) does not increase.

3) Symptom: "Activation OK, but Retention is falling apart"

Where: Retention.
Lenses: Cohorts + JTBD + (sometimes) HEART (Engagement/Task Success).
Artifact: Cohorts by first scenario + JTBD map of forces (habit/anxiety).
Test: Create a "natural reason to return" (contextual reminder, repeatable cycle, template).
Guardrail: Activation does not drop; complaints/errors do not increase.

4) Symptom: "Retention is normal, Referral is almost zero"

Where: Referral.
Lenses: JTBD (social/emotional motive) + CJM invitation + Loops.
Artifact: "Micro-CJM referral" (when a reason arises, what prevents it).
Test: Move referral to the moment of maximum value + reduce friction (1–2 clicks).
Guardrail: Do not worsen core-flow and do not increase spam/negativity.

5) Symptom: "Users, but no payments"

Where: Revenue.
Lenses: JTBD (what they "truly" pay for) + payment path + Unit Economics + (if purchases exist) RFM.
Artifact: Map of "payment trigger" (where need/risk/limit arises).
Test: Shift paywall to the moment of maximum value or sell "accelerated progress" instead of "access."
Guardrail: Retention and NPS/complaints do not drop; margin does not worsen.

6) Symptom: "Metrics seem okay, but growth is stuck"

Where: More often the system, not a single pocket.
Lenses: Growth loops + Flywheel + Economics.
Artifact: 1 loop diagram (Input → Action → Output → Reinvest) + list of friction points.
Test: Strengthen one loop (content/UGC/invites/integrations) and measure the loop coefficient.
Guardrail: Customer Acquisition Cost (CAC) does not grow faster than Customer Lifetime Value (LTV); base quality does not decline.

7) Symptom: "Average metrics are normal, but feeling 'everything is bad'"

Where: Hidden segmentation/compensation.
Lenses: Cohorts + segments + RFM (if money involved).
Artifact: Breakdown by segments (new/old, channel, scenario, role).
Test: Targeted repair of the most problematic segment instead of "improve everything for everyone."
Guardrail: Do not drop the key segment that holds revenue/value.

Using this protocol, you can not only find the "leak" in the product but also develop an effective plan to fix it, based on data and clear success criteria.