Product Analytics: The HEART Framework

Focus: Measuring the Quality of User Experience (UX).

What Is Product Analytics?

Product analytics is the practice of measuring how users interact with your product — what they click, where they drop off, which features they love, and which they ignore. Unlike marketing analytics (which asks "how do we get users?"), product analytics asks "once they're here, are we delivering value?"

Without product analytics, product decisions are made on instinct. With it, you can run experiments, prioritize roadmaps with data, and understand why retention is falling before it becomes a revenue problem.

Why the HEART framework? Developed by Google, HEART gives product teams a structured way to pick the right metrics for any feature or product goal. It prevents the classic mistake of optimizing one number (e.g., DAU) while silently destroying another (e.g., NPS).

The Five HEART Components

H — Happiness

What it is: Happiness measures how users feel about your product — their subjective satisfaction, trust, and emotional connection to the brand. It's the only HEART metric you can't observe from clickstream data alone; it requires asking users directly.

Why it matters: Unhappy users churn silently. They don't always complain — they just leave. A falling NPS is a leading indicator of future churn, often by 60–90 days. Conversely, highly satisfied users become referral engines.

Key metrics defined:

• NPS (Net Promoter Score): Ask "How likely are you to recommend us?" on a 0–10 scale. Promoters (9–10) minus Detractors (0–6) = NPS. World-class B2B SaaS: 40+. Consumer apps: 50+.
• CSAT (Customer Satisfaction Score): Post-interaction rating, typically 1–5. Measures satisfaction with a specific interaction (e.g., support ticket, onboarding call).
• App Store Rating: Public-facing happiness signal. Below 4.0 hurts organic discovery. A sudden drop almost always correlates with a specific release.
• Sentiment Score: NLP analysis of support tickets, reviews, and social mentions. Gives continuous happiness signal without surveying.

Types of analysis to conduct:

• Segment NPS by plan tier, acquisition channel, and cohort month — a 40 NPS overall can hide a −10 NPS from your enterprise segment
• Build an NPS trend chart against your release timeline — drops almost always trace to a specific deploy
• Correlate NPS with 90-day retention: users who score 9–10 retain at 2× the rate of 7–8 scores

What a change means:

NPS drops 10+ points in a single month: A specific release, pricing change, or policy update broke trust. Pull the verbatim comments from detractors — they'll tell you exactly what happened. Cross-reference with your release calendar.

NPS rises without a product change: Your organic referral loop is working. Users are self-selecting in and arriving with higher expectations already met. This is the signal to invest in a referral program.

CSAT falls on support interactions only: Product quality is fine; your support experience is the problem. Look at first-response time, resolution rate, and whether the same issues repeat.

Root cause analysis framework:

1. Segment first — is the drop universal, or isolated to mobile / new users / a specific plan?
2. Match timing to events — releases, pricing changes, competitor launches
3. Read 20 verbatim responses from detractors — qualitative beats quantitative for diagnosis
4. Check support ticket volume — rising tickets + falling NPS = active product bug

E — Engagement

What it is: Engagement measures the intensity of product use — how often users come back, how long they stay, and how deeply they interact with core features. It's the heartbeat of your product.

Why it matters: Engagement is the strongest predictor of retention and LTV. A highly engaged user is expensive to acquire but cheap to keep. Low engagement means users are not forming habits around your product — and habitual products are defensible products.

Key metrics defined:

• DAU/MAU Ratio (Stickiness): Daily Active Users ÷ Monthly Active Users × 100. Measures how much of your monthly audience returns daily. WhatsApp: ~85%. Healthy B2B SaaS: 20–30%. Consumer app: 40%+.
• Session Length: Average time per session. Context matters — a tax tool should have short sessions (efficiency). A learning app should have long ones (depth).
• Feature Frequency: How often a specific feature is used per active user per week. This reveals which features are genuinely habit-forming vs. occasionally discovered.
• Sessions per User per Week: How many times a user opens the product. A falling number means you're losing the habit loop.

Types of analysis to conduct:

• Plot DAU/MAU trend over 12 months — look for secular decline vs. seasonal patterns
• Build an engagement distribution: what % of users account for 80% of sessions? Highly skewed distributions signal a power-user problem.
• Cohort engagement decay: chart average sessions/user/week for each signup cohort over 12 weeks. A fast-decaying curve means your core loop is weak.

What a change means:

DAU/MAU falls steadily over 3 months: Product isn't forming habits. Investigate notification strategy, re-engagement emails, and whether the core loop has a natural daily trigger.

Session length spikes suddenly: Could be good (deep engagement with a new feature) or bad (users confused and spending time trying to find things). Cross-reference with task success metrics and support tickets.

Engagement high for new users, declines by month 3: Classic "exploration plateau" — users exhaust novelty but haven't formed a genuine habit. This is a core loop / feature depth problem.

Root cause analysis:

1. Segment by acquisition channel — organic users typically engage 2–3× more than paid users
2. Check if a feature was deprecated, hidden, or broken in a recent release
3. Compare engagement across device types — mobile vs. desktop engagement profiles differ significantly
4. Look for external seasonality — B2B products drop in August and December; this isn't a product failure

A — Adoption

What it is: Adoption measures whether users are discovering and successfully using new features or core product capabilities. It answers: "Did we build the right thing, and can users find it?"

Why it matters: Low adoption is the silent killer of product roadmaps. A team can spend 6 months building a feature that only 3% of users ever touch. Adoption metrics tell you whether R&D investment is translating into user value — before you double down on the wrong bets.

Key metrics defined:

• Feature Adoption Rate: (Users who used Feature X in the period) ÷ (Total active users in the period) × 100. A rate below 10% for a core feature is a red flag.
• Time-to-First-Use: Median time from signup to first activation of a feature. Shorter is generally better for core features; longer can indicate discoverability problems.
• Onboarding Completion Rate: % of new users who complete each step of the onboarding checklist. Each step's drop-off rate is a prioritization signal.
• Breadth of Adoption: Average number of distinct features used per active user per month. Rising breadth = deeper product stickiness.
• Feature Retention: Of users who tried Feature X, what % use it again in the next 30 days? This separates features users try once vs. features they depend on.

Types of analysis to conduct:

• Funnel analysis: Signup → Onboarding Step 1 → Step 2 → … → Aha! Moment. Find the step with the largest drop-off.
• Segment adoption by plan tier — power users on paid plans may adopt heavily while free users ignore a feature entirely
• Compare time-to-adoption across cohorts to measure onboarding improvements

What a change means:

Adoption rate falls after a UI redesign: You moved or hid a feature. Check if its placement in the navigation changed. Run a click heatmap on the area where it used to live.

Adoption plateaus at 15–20%: The feature solves a problem that only a subset of users have. This isn't a discoverability problem — it's an ICP (ideal customer profile) problem. Consider whether it belongs in a higher tier.

Time-to-adoption increases: Your onboarding isn't teaching users about the feature. Add an in-app tooltip, an empty-state CTA, or include it in the onboarding checklist.

Root cause analysis:

1. Is the feature visible in the primary navigation, or buried 3 levels deep?
2. Does the empty state explain the feature's value clearly?
3. Are there permission restrictions preventing free users from accessing it?
4. Segment by plan — if only premium users are adopting, the feature may be mis-tiered.

R — Retention

What it is: Retention measures whether users return to your product over time. It is the most important metric in any recurring-revenue or subscription business — and the hardest to move.

Why it matters: Retention is the foundation of unit economics. If you acquire 1,000 users per month at a CAC of $50 but retain only 10% of them at 6 months, your LTV is far lower than the model assumes — and you'll burn cash indefinitely trying to replace churned users. High retention means every dollar spent on acquisition compounds.

Key metrics defined:

• Day-N Retention (D1, D7, D30, D90): % of users who were active N days after their first session. D1 measures onboarding quality; D30 measures habit formation; D90 measures long-term value delivery.
• Churn Rate: (Users lost in period) ÷ (Users at start of period). Monthly churn of 2% = 22% annual churn. Monthly churn of 5% = 46% annual churn.
• Net Revenue Retention (NRR): (Starting MRR + Expansion MRR − Churned MRR − Contraction MRR) ÷ Starting MRR × 100. The north star for B2B SaaS. Above 120% means your existing customers grow revenue faster than others churn.
• Rolling Retention: % of a signup cohort still active in a given period. Plotted as a curve to reveal Product-Market Fit.

Retention curve interpretation:

Types of analysis to conduct:

• Cohort retention heatmap: rows = signup month, columns = period number, values = % retained. Dark cells = high retention. Look for a specific cohort that performs worse — it reveals a product regression.
• Segment churn by plan tier, acquisition channel, company size — churn is almost never uniform
• Compare features used by retained vs. churned users — what do retained users do differently in their first 7 days?

What a change means:

D1 retention drops 10+ points: Onboarding or activation is broken. The first-session experience isn't delivering value fast enough. Could be a technical bug, a UX regression, or an onboarding flow change.

D30 drops while D1 is stable: Users are activating (first session worked) but not forming a habit. The core loop or trigger (email digest, notification, daily use case) is broken.

Churn spikes in one cohort month: A specific change hurt that cohort. Map the churn spike to your release calendar. Check if pricing, a removed feature, or a policy change affected only those users.

Root cause analysis:

1. Exit surveys — ask churned users one question at the moment of cancellation: "What was the main reason you left?"
2. Feature usage before churn — did churned users stop using a key feature 2–3 weeks before cancelling?
3. Support ticket analysis — did churned users have unresolved issues?
4. Cohort comparison — which acquisition channel produces the highest D30 retention? Invest more there.

T — Task Success

What it is: Task Success measures whether users can efficiently accomplish what they came to do. It's the UX equivalent of "does the product actually work the way users expect it to?"

Why it matters: Users don't churn when a product is complex — they churn when a product is confusing. Every percentage point improvement in task completion reduces support tickets, increases confidence, and shortens time-to-value. It directly affects both retention and NPS.

Key metrics defined:

• Task Completion Rate: % of users who successfully complete a defined multi-step task (e.g., export a report, create a dashboard, submit a query). A completion rate below 70% on a core flow is a critical UX problem.
• Error Rate: % of user actions that produce an error. Errors destroy trust. Segment by error type: user errors (recoverable with better UX) vs. system errors (require engineering fixes).
• Time on Task: Median time to complete a task. Rising time means increasing friction — navigation became harder, or a new step was added. Falling time = UX improvement.
• Help Article Views: Volume of views for specific help articles. A spike after a release is a strong signal that the new UI is not self-explanatory.

Types of analysis to conduct:

• Funnel analysis on every multi-step core task: where exactly do users abandon?
• Segment error rates by device, OS version, and plan — a mobile-only error pattern reveals a responsive design bug
• Before/after analysis for UX changes: compare task completion rate in the 4 weeks before and after a redesign

What a change means:

Task completion drops after a redesign: The new UI obscured a step or moved a key action. Watch session recordings of users attempting the task on the new design.

Error rate spikes on a specific action: A backend bug, a validation rule that's too strict, or a recently changed API endpoint. Cross-reference with your error log and deployment history.

Help article views spike after a release: Users encountered something unexpected. The release assumed knowledge that users don't have. Consider adding contextual tooltips or an in-product walkthrough.

Root cause analysis:

1. Session recordings — watch 10 recordings of users who dropped off at the problem step
2. Heatmaps — where are users clicking that does nothing? (Rage clicks)
3. Error log classification — categorize errors by type and frequency
4. A/B test the flow — simplify the problematic step against the control