issue 042 / Q1 / volume 03

pmt.report

a quarterly dispatch of metrics, anomalies, and dissent

data dissent design

a personal report

est. 2026

executive summary

read this part if nothing else

user growth +184% YoY
12 quarter trend
tap
signal density 8.42 db / event
frequency spectrum
tap
dissent index 73 / 100 unfiltered
quarter volatility
tap
field note

the numbers alone never told the story; the noise around them did.

p95 latency 128ms -22% qoq
descending p95
tap
coverage 94.6% across 11 markets
94.6 / 100
tap

tap a hex. it morphs. that is the point.

anomalies & outliers

where the boring reports go quiet

anomaly 01

the tuesday spike

a 612% jump in dissent-index registered on a single tuesday at 14:07 utc. no campaign. no incident. just a sudden, collective sigh from the data.

peak Δ +612% 14:07 utc
single point spike
tap
duration 04:21 m:s
elapsed of 09:00 window
tap
anomaly 02

the friday drift

every friday between 16:30 and 17:00, signal density softened by a consistent 7.4%. no scheduled cause. we are calling it the early-weekend effect until told otherwise.

drift -7.4% recurring
five week echo
tap
confidence 88.1% two-sided
distribution silhouette
tap

these are the lines we did not redact.

methodology

how the sausage is made

01

collect

11 markets, 4.2M events, sampling rate 1 / 1024.

02

clean

drop bots, drop dupes, keep the awkward ones for later.

03

contest

we argue with the model. when it wins, we publish; when we win, we annotate.

04

render

hex first, table second. visuals are not garnish.

we keep the disagreements in the footnotes.

field notes

unfiltered, mostly

we removed three "growth opportunity" charts. they were not opportunities. they were ceilings with paint.

— ed.

the loudest stakeholder asked for a smiley face on the slide. we negotiated for a hexagon.

— a.

a metric that improves every quarter and has no negative correlate is not a metric. it is a vibe.

— j.

we trust the median more than the mean for one reason: the median can read the room.

— r.