Insights · The RevenuePoint quarterly

Notes on orchestration.

Essays on growing-business operations, the systems gap, and how we’re closing it at RevenuePoint — from the people doing the work.

Cover: the CRM is where your team works; the warehouse holds the truth; the right pipeline closes the gap

Featured · long read

CRMArchitecture

Building a single source of truth in your CRM (and keeping it true)

The CRM is where your team works every day. The warehouse is where the truth is assembled. The hard part is the pipeline that closes the gap — and stays closed. Here’s the recommended pattern: warehouse-first, canonicalized, synced back via reverse ETL, monitored end to end.

Thomas Jones · Apr 27, 2026 · 11 min read

An agent writes a report; a QC stage verifies it through severity, confidence, and status checks; a named analyst signs off
AIResearch

Agent writes, QC verifies: the codes and checks behind a research report you can defend

AI can produce a confident-sounding briefing on any subject in thirty seconds. Most of them are wrong in ways that take an analyst a day to find. Here’s the QC pipeline we run — every check, every code, every gate — that makes the difference between a report that ships and a report that comes back.

Apr 26, 2026 · 11 min read

Five-stage AI reporting pipeline — question, plan, query, QC, result — with a provenance badge attached to the result
AIReporting

How to do AI reporting right: the multi-step pipeline behind a number you can trust

Plain-English questions are easy for a model to answer. The hard part is building the pipeline around the model so the answer is actually correct — and defensible. Here’s the step-by-step approach we use.

Mar 31, 2026 · 9 min read

A time series with a noise band; one point lifted above the band and labeled as the anomaly
Anomaly DetectionMonitoring

How anomaly detection actually works: the methods behind an alert you can trust

Static thresholds don’t survive contact with real business metrics — the data has trend, seasonality, and noise. Here are the three algorithm families that do the actual work, and how we choose between them.

Feb 24, 2026 · 10 min read

Three slightly different records — Acme, Inc., ACME Inc, acme inc. — converging into one resolved entity
Entity ResolutionData Quality

Entity resolution: why your CRM thinks you have three customers when you have one

Every business above a certain age has the same hidden bug: the same customer appears two or three times across systems, just different enough to look like separate records. Here’s what that costs you, and the disciplined way to fix it in the warehouse layer.

Jan 13, 2026 · 8 min read

A semantic layer band sits between warehouse tables at the bottom and dashboards, reports, agents, and AI chat at the top — every consumer reads through the layer, not the tables
Semantic LayerAI

The semantic layer: where 'revenue' gets defined exactly once (and why AI can't work reliably without it)

Five teams. Five definitions of “revenue.” Five different numbers in five different reports. The semantic layer is the layer that fixes this — and the reason an AI on top of your warehouse goes from confidently wrong to defensibly right.

Nov 18, 2025 · 9 min read

Sources flow through three pattern lanes — batch, CDC, webhooks — into a DAG, then to dev, staging, and prod warehouses with promotion arrows
Data PipelinesInfrastructure

How data actually gets into your warehouse: batch, CDC, webhooks — and the DAGs that keep it all honest

Every warehouse is only as trustworthy as the pipelines feeding it. Here are the three patterns that move data from source systems into the warehouse, the tradeoffs between them, and how we use DAGs to orchestrate the whole thing — including moves between environments.

Oct 7, 2025 · 10 min read

Five source systems converge into a central warehouse, then radiate out to dashboards, reports, and agents
Data WarehouseInfrastructure

The data warehouse: why every other layer of your business depends on it

Dashboards, automations, agents, and AI chat all collapse on top of the same thing: a warehouse that has your business’s data in one place, defined the same way, and kept safe. Here’s what that actually buys you.

Aug 26, 2025 · 9 min read

Three labeled stages: agent proposes, human approves, system executes
OrchestrationAI

What orchestration actually means (and why it's the safe way to let AI act on your business)

A plain-English walkthrough of how AI agents, tools, and systems of record fit together — and why the human approval step is the whole reason this approach is safe.

Jul 15, 2025 · 8 min read

Newsletter · monthly

Get new posts in your inbox.

Occasional essays on orchestration, data infrastructure, and the work of running a modern business. No spam — unsubscribe anytime.