Issue No. 187·Data
The dashboard your CEO loves is hiding a 38% miss in the metric she actually measures.
A teardown of the most-cited "activation rate" chart in B2B SaaS, why its denominator quietly doubled in 2023, and what your growth team should be reporting instead. A 14-minute read with the underlying CSV.
Recent dispatches
Long-form arguments for analysts who'd rather be right than optimistic.
Four pieces from the past month — growth teardowns, dashboard dissections, post-mortems, and methodology critiques. Average post length: 2,400 words; each one fact-checked against an original dataset.
-
Growth teardown
Why your "weekly active users" number has been wrong since the cookie deprecation — and the query that fixes it.
By Devi Okafor · 11 min read · March 7, 2025
A post-mortem on three SaaS companies whose WAU charts quietly diverged by 30–60% once third-party identifiers disappeared. Includes the dbt model we used to reconstruct the cohort.
-
Dashboard dissection
The "engagement" chart that posted Notion to a 40% retention miss — annotated line by line.
By Miguel Saldaña · 9 min read · February 28, 2025
An examination of how a stacked-bar visualization conflated two distinct user states, why the underlying SQL grouped by an inconsistent time grain, and what we'd ship instead.
-
Post-mortem
A retail chain's "best quarter in five years" turned out to be a revenue-recognition error. We traced the line item back.
By Priya Raman · 17 min read · February 21, 2025
The single footnote in the 10-Q versus what the dashboards were reporting internally — and the meeting agenda that would have surfaced the discrepancy two quarters earlier.
-
Methodology critique
"AI-powered analytics" is mostly a regression with extra steps. A reproducible benchmark across 14 vendors.
By Tomás Lindgren · 13 min read · February 14, 2025
We ran the same 40 natural-language question benchmark across 14 "AI analytics" products. Average accuracy against ground truth SQL: 41%. We publish the prompts, the answers, and the grader.
The Dashboard Roast
A standing audit of dashboards other analysts won't touch.
Every quarter we publish anonymized, line-by-line critiques of dashboards submitted by readers — 4,200+ audited since launch. We grade on metric definition, denominator consistency, time-grain, and the question the chart claims to answer.
- Metric definition. Does the numerator match the denominator across every filter on the page?
- Time-grain honesty. Are daily, weekly, and monthly series computed on the same calendar boundaries?
- The leading question. Strip the chart — does the title still tell a single, falsifiable story?
- Vanity checks. Is any series present that the team would not, in writing, defend in a board meeting?
4,200+ dashboards audited · quarterly cadence · free to submit your own.
Why operators trust the brief
Third-party citations, verified numbers, and the press list we never paid for.
- 38,400 Verified newsletter subscribers · 62% open rate · 11% click rate
- 184,000 Monthly unique visitors · Similarweb-verified, Q1 2024
- 487 Long-form articles published in the first six years
- 22 mo. Average reader tenure · top 1% of independent newsletters tracked by Mailchimp
- Harvard Business Review — cited 14+ times since 2021
- The Economist — cited in three long-form data features
- Stratechery — linked in the weekly newsletter commentary
- HubSpot — "Best Marketing Blogs of 2024" (#7)
- CMI — "Must-Read B2B Newsletters" (#3)
- Analytics8 Annual Awards — "Best Independent Analytics Publication" 2023 & 2024
From the editors
"We started this publication because every analytics post we'd read that quarter was either a vendor brochure, a recycled Medium post, or a takedown by someone who'd never shipped a dashboard. We wanted the thing we wanted to read: long, evidence-first, written by people who've sat in the operator seat when a chart broke at 2 a.m. Anti-hype isn't a posture — it's an editorial standard."
Independently owned since 2019 · headquartered at 2401 East 6th Street, Suite 3030, Austin, TX · 9-person editorial team including 3 PhDs in statistics and a former Netflix data scientist.
Read our editorial standards