Dispatch

Vol. 3 / Issue 12Feb 2026

The cloud has a thousand dashboards
and no memory.

Dispatch exists because data without narrative is just noise — and noise is what your competitors are drowning in.

Read the latest issue

Signal

Recent Editions

Three issues. Three theses. Judge the voice.

Issue 10Cloud Infrastructure8 min read

The Trend

"Enterprises running three or more cloud providers report 34% higher operational overhead — yet only 12% have a unified cost governance model."

34%

Higher operational overhead

in organizations running 3+ cloud providers (n=214 enterprises, Q3 2025)

Multi-cloud is not a strategy. It's a hedge against regret.

The multi-cloud narrative has been sold as resilience. What it actually delivers is complexity debt. When AWS goes down, your GCP failover doesn't activate because the team that built it left eighteen months ago. The contracts are there. The runbooks are not. We tracked 47 enterprise migration post-mortems from 2024 and found a consistent pattern: the orgs that survived cloud outages cleanly were not the ones with the most providers — they were the ones with the clearest ownership boundaries.

Read full issue

Issue 11SaaS Metrics11 min read

The Challenge

"The cohorts expanding at 130% NRR in 2022 are the same ones showing 18-month churn cliffs in 2025 data — expansion was borrowing from future retention."

18mo

Churn cliff timeline

median lag between peak NRR and first contraction event in high-growth cohorts

NRR above 120% is not a moat. It's a margin call waiting to happen.

Everyone celebrated net revenue retention as the defining SaaS metric of the expansion era. The problem is that NRR is a lagging indicator dressed as a leading one. When a customer expands from $50k to $200k ARR in year two, that looks like product-market fit. What it often is: a champion with budget authority and no internal procurement process. When that champion leaves — and they leave — the renewal conversation starts from zero. We've been tracking 38 public SaaS companies' cohort data against their stated NRR figures and the divergence is not subtle.

Read full issue

Issue 12Enterprise Data13 min read

The Prediction

"Gartner calls it consolidation. We call it what it is: $2.4B in annual lakehouse licensing that nobody's querying past the demo environment."

61%

Tables never re-queried

after 90-day deployment window across 1,200 enterprise lakehouse accounts

By Q4 2026, the data lakehouse will be the new shelfware.

The data lakehouse was supposed to end the warehouse-vs-lake debate. Instead it created a third thing that does both jobs adequately and neither job well. The tell is in the query logs. We pulled anonymized usage data from three lakehouse vendors (combined: 1,200 enterprise accounts) and found that 61% of tables created in the first 90 days of deployment had zero reads after month six. This is not a data quality problem. It is an organizational problem that vendors are happy to let you mistake for a tooling problem — because tooling problems have paid upgrades.

Read full issue

Dispatch

"True analysis doesn't tell you what happened. It tells you why it will happen again — and who's positioned to benefit."

4,200+

Weekly readers

3 yrs

Of uninterrupted publishing

Proof of depth, before you commit your inbox.

Three full issues, free to read. No gate. No form. Just the work.

Issue 09January 2026

The Kubernetes Tax: What Platform Teams Actually Cost

A forensic look at the hidden labor costs inside platform engineering orgs — and why the build-vs-buy calculus shifted in Q4 2025.

InfrastructurePlatform EngCost Analysis

14 pages

Read Issue

Issue 07December 2025

Churn Cohorts That Lie: Reading Between the NRR Lines

How three public SaaS companies reported 115%+ NRR while losing their largest customer segment — and what the SEC filings revealed six quarters later.

SaaS MetricsChurnFinance

18 pages

Read Issue

Issue 04October 2025

The Data Mesh Reckoning: Two Years After the Hype

We surveyed 82 data engineering leads who implemented data mesh architectures. 71% said the organizational model was the failure point, not the technology.

Data ArchitectureOrg DesignEnterprise

22 pages

Read Issue

Browse all 12 issues