LogoPulse
// analytics briefing
Get Free Briefing
LIVE·pulse.signal.io / diagnostics2026-02-27 22:17 UTC

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

pulse-diagnostics — funnelRUNNING
events/sec2,400
latency12ms
signals0 flagged

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.

Issue #148FUNNEL ANALYSIS
raw_signal.sql

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;
periodvolumerate
22:004,8218.4%
23:004,9038.1%
00:005,1126.5%
01:004,9876.2%
pulse_read.md

Instrumentation noise, not UX regression

confidence94%
Drop correlates exactly with deploy at 23:47 UTC
purchase_complete event missing properties.session_id after deploy
Amplitude deduplication silently dropping ~22% of events
Actual conversion unchanged — your pipeline broke, not your funnel

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;
Issue #147RETENTION CURVE
raw_signal.sql

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;
periodvolumerate
Nov W13,24119.2%
Nov W23,48918.8%
Nov W33,10219.1%
Nov W43,67818.9%
pulse_read.md

Activation gap — users never reaching aha moment

confidence87%
D1 retention is healthy (42%) — users like what they see initially
D7 cliff is a classic "I forgot this existed" pattern
No re-engagement email or push in D2–D6 window in your event stream
Top D30 retained users all triggered core_feature_used within 48h of signup

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;

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.

pulse-diagnostic-free — v2.4.1

QUICK_EXAMPLES:

$

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.

AmplitudeMixpanelPostHog
Amplitude3h ago✓ resolved

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...

SR
Siddharth RaoGrowth Eng @ Series B
47💬 14
→ resolved via Pulse Issue #146
PostHog7h ago

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...

MC
Megan CallowayPM @ PLG SaaS
31💬 9
Mixpanel1d ago✓ resolved

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"...

JO
James OkonkwoData Lead @ Fintech
89💬 28
→ resolved via Community answer

+ 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.

Weekly decoded dashboard (1 real case)
Community-sourced anomaly with root cause
One SQL snippet you can run today
Access to free diagnostic tool

No credit card · No spam · Unsubscribe anytime

14-DAY TRIALNO CREDIT CARD

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.

Full diagnostic tool (all 3 analyzers)
148-issue searchable archive
Community forum + Slack channel
Weekly live Q&A with editors
Custom SQL snippet library
Platform-specific playbooks

No credit card required · Cancel anytime before day 14