// analytics intelligence
Stop watching
dashboards.
Start reading them.
One decoded dashboard, one community anomaly, one SQL snippet — every week. Built for growth engineers, PMs, and data leads who need signal, not noise.
step_1→2
91.2%
step_2→3
78.4%
step_3→4
41.1%
step_4→5
88.9%
12,400+
engineers & PMs
148
weekly issues
4.8★
avg rating
// what's inside each issue
Raw data problem → Pulse interpretation
Every issue trains your eye to see the gap between noise and signal. The problems escalate. The free tool's limits become visible.
Checkout conversion dropped 22% overnight
SELECT
date_trunc('hour', timestamp) as hour,
COUNT(*) FILTER (WHERE event='checkout_start') as starts,
COUNT(*) FILTER (WHERE event='purchase_complete') as completions,
ROUND(
COUNT(*) FILTER (WHERE event='purchase_complete')::numeric /
NULLIF(COUNT(*) FILTER (WHERE event='checkout_start'), 0) * 100, 1
) as cvr_pct
FROM events
WHERE timestamp > NOW() - INTERVAL '48 hours'
GROUP BY 1 ORDER BY 1;Instrumentation noise, not UX regression
VERIFY_WITH.sql
-- Quick verify: check event property completeness
SELECT
DATE_TRUNC('hour', timestamp) as hr,
COUNT(*) as total,
COUNT(properties->>'session_id') as has_session_id,
ROUND(COUNT(properties->>'session_id')::numeric / COUNT(*) * 100, 1) as pct_complete
FROM events WHERE event = 'purchase_complete'
AND timestamp > NOW() - INTERVAL '6 hours'
GROUP BY 1 ORDER BY 1;D7 retention stuck at 19% for 3 cohorts
SELECT
cohort_week,
COUNT(DISTINCT user_id) as cohort_size,
COUNT(DISTINCT CASE WHEN days_since_signup = 7
THEN user_id END) as retained_d7,
ROUND(
COUNT(DISTINCT CASE WHEN days_since_signup = 7
THEN user_id END)::numeric /
COUNT(DISTINCT user_id) * 100, 1
) as d7_retention
FROM user_activity
WHERE cohort_week >= '2025-11-01'
GROUP BY 1 ORDER BY 1;Activation gap — users never reaching aha moment
VERIFY_WITH.sql
-- Find the activation event separating retained vs churned
SELECT
first_event,
AVG(CASE WHEN d7_retained THEN 1 ELSE 0 END) as retention_rate,
COUNT(*) as users
FROM (
SELECT u.user_id,
FIRST_VALUE(e.event) OVER (
PARTITION BY u.user_id ORDER BY e.timestamp
) as first_event,
MAX(CASE WHEN e.days_since_signup = 7 THEN 1 ELSE 0 END) as d7_retained
FROM users u JOIN events e USING (user_id)
) sub GROUP BY 1 ORDER BY 2 DESC LIMIT 10;// free diagnostic tool
Paste a metric scenario. Get an instant read.
Describe what you're seeing in your dashboard. Pulse classifies whether it's churn, activation failure, or instrumentation noise — before you spend 3 hours investigating.
QUICK_EXAMPLES:
// community threads
Conversations you want to be in
12,400+ engineers and PMs debugging real problems in real time. Every thread is indexed against 148 issues of pattern history.
Why does my funnel show higher conversion than Stripe?
Been debugging this for 2 days. Amplitude shows 8.4% checkout CVR but Stripe only shows 6.1% successful charges. The gap is consistent across 3 weeks...
Identify() calls not stitching anonymous → known users in cohorts
Our activation cohort analysis is completely broken. Anonymous session events aren't being attributed to identified users even though we call identify() on signup...
p-value on A/B test is 0.03 but my CEO wants "99% confidence" — how do I explain this?
Board meeting tomorrow. Test ran for 14 days, n=12,400 per variant, primary metric +7.2% with p=0.03. My CEO saw a tweet saying "p-value doesn't mean probability of being right"...
+ 340 active threads this week · Full access with 14-day trial
Join the Community →Get the Free Briefing
One decoded dashboard, one community anomaly, one SQL snippet — every Thursday at 9am UTC. No noise, no filler. Unsubscribe in one click.
No credit card · No spam · Unsubscribe anytime
Start 14-Day Full Access
Full tool suite, community access, and the complete archive of 148 decoded dashboards. Cancel before day 14 and pay nothing.
No credit card required · Cancel anytime before day 14