The Signal Method
Most software organizations operate on a fiction: that the project management layer is the ledger. It is not. It is the expense report. The real ledger lives in the code, in the meetings, in the decisions, and in the finance entries that nobody connects to each other.
The Signal Method is a four-layer framework I developed inside InteliG and apply inside every engagement. It is how a software org stops running on theater and starts running on truth.
The stack
Raw data, captured at the source.
Commits. Pull requests. Deployments. Meetings. Decisions. Finance entries. Customer signals. Everything that happens inside the org becomes a signal. The signal is captured at the source, not transcribed by humans into a separate tool.
For PE: the engineering org stops self-reporting. The data is the ledger, not a summary of it.
Declared intent. Initiatives. Outcomes.
What we said we would do. The strategic initiatives. The outcomes we committed to. The non-goals. Strategy is the layer that gives meaning to the signal stream. Without it, intelligence is just activity.
For PE: the value-creation thesis becomes a structured, queryable object. Not a Word doc.
Decisions, rationale, context preserved.
Why we did what we did. The architectural decisions. The trade-offs. The context the next person needs. Knowledge is where the organization stops forgetting. It is what makes the operating system continuous instead of episodic.
For PE: institutional memory survives CTO transitions, M&A, restructuring.
Cross-layer synthesis. Cognis.
The AI layer that reasons across all three lower layers. Connects signal to strategy to knowledge. Answers questions in plain English. Not retrieval. Causal understanding accumulated over time.
For PE: the CTO answer machine. Replaces the weekly status meeting and the quarterly board prep.
Architectural foundation
The Signal Method does not work on a CRUD architecture. It requires primitives that make truth, time, and context first-class. The three patterns below are non-negotiable.
Bounded contexts. Aggregates that enforce invariants. Ubiquitous language. The same framework that survives 10x team growth is the framework that survives 10x AI volume.
Separate read and write models. Optimize each for its purpose. Reads scale, writes stay consistent. Agents can read everything; mutation stays disciplined.
Every state change is an event. Replay the entire history. Audit by default. AI reasoning improves dramatically when given a temporal event stream instead of a current-state snapshot.
Core principles
Filter ruthlessly. A signal must be load-bearing or it is noise consuming attention.
Every action carries why it was taken and what it produced. Both ends matter.
Not a feature. A layer the org sits on top of.
The system surfaces what is, not what is convenient. Comfort kills companies.
What you refuse to build is as defining as what you build.
The Value Creation engagement track is built around installing the Signal Method in 90 days. Diligence sprints use a compressed version to read the target.
