Structured problem understanding

Before you solve it, understand it.

Phase 1 maps the system behind your problem — causes, dependencies, unknowns, and assumptions — before any solution is on the table. Clear thinking starts here.

How it works

From raw observations to a causal map

Describe what you're seeing. Phase 1 builds the structure automatically — grouping evidence into clusters, identifying causal relationships, and surfacing what you still don't know.

1

Describe the problem

State what you observe, what you want instead, and the boundaries you're working within.

2

Build the model

The system groups your observations into clusters and maps causal relationships between them.

3

Investigate

Answer targeted questions to validate hypotheses, fill gaps, and sharpen weak links in the model.

4

Confirm clarity

When structural readiness is met and you feel confident in the map, close Phase 1 and move forward.

Under the hood

Observations become structure

You add what you observe. The system finds the shape. Every causal claim is traceable back to specific evidence — the model never invents structure.

Step 1

Observations

You describe what you're seeing — facts, events, measurements, and reported beliefs. Each observation is typed and embedded for grouping.

Measurement / factEventReported belief

Step 2

Clusters

Related observations are grouped into clusters — each representing one underlying phenomenon. Cluster identity is stable: it's determined by which observations belong to it, not by its label.

ML grouping → LLM reviewObserved & desired anchors fixedLabels are display names only

Step 3

Relationships

The system ranks candidate causal pairs and classifies each one. Confidence is computed deterministically from the evidence types behind each link.

Measurement
Event
Reported belief

What you get

A structural picture, not a guess

Every output is evidence-backed and confidence-rated. Nothing is invented — if it's in the model, there's a chain of observations behind it.

Causal clusters

Observations are grouped into clusters that describe the same underlying phenomenon, with relationships mapped between them.

Confidence ratings

Each relationship carries a confidence score derived from the types of evidence behind it — facts carry more weight than beliefs.

Named unknowns

Gaps in the causal map are surfaced explicitly — you know what you don't know, and the system asks the questions most likely to resolve them.

Load-bearing assumptions

Assumptions that heavily influence the model are flagged for validation before they silently shape every downstream decision.

Targeted questions

One question at a time, ranked by expected information gain — weak relationships, causal gaps, and key assumptions first.

Structural readiness

A clear signal when the model has enough coverage to support moving forward — not just when it feels intuitively "done."

Use cases

The kind of problems Phase 1 helps with

Any problem where you're not sure you're looking at the right thing — where the symptom is obvious but the system behind it isn't.

Customer success

Rising churn in the 60–90 day window despite positive onboarding scores

AssumedOnboarding quality or product complexity

Model surfaced

A misalignment between what sales promised and the workflow customers actually adopted — diverging at first meaningful use.

Sales–product gapFirst meaningful use

Operations

Slow cross-team handoffs delaying every design-to-engineering delivery

AssumedCommunication breakdowns and tool fragmentation

Model surfaced

Three independent blockers — spec ambiguity, async review lag, and missing decision authority — each sufficient on its own.

Spec ambiguityReview lagNo decision owner

Leadership

Two senior engineers are quietly disengaged and no one knows why

AssumedCompensation, career growth, or team dynamics

Model surfaced

Work had become invisible — impact felt only through its absence, with no formal recognition pathway.

Invisible impactNo recognition path

Product

Ideas keep getting built halfway and stalling before they ship

AssumedResource constraints and competing priorities

Model surfaced

No one had authority to cut scope — everything stayed in progress rather than shipping or being killed.

Scope lockDecision vacuum

Personal

Motivation has dropped since the company started using AI more heavily

AssumedJob security fears and increased workload

Model surfaced

Two distinct clusters — loss of craft identity and anxiety about relevance — each needing a different response.

Craft identity lossRelevance anxiety

Confidence model

Not all evidence is equal

Relationship confidence is computed deterministically from the types of observations supporting each causal link. A relationship backed only by reported beliefs is automatically flagged as weak — and becomes a question target.

Measurement or fact
Quantitative data, documented events, verified records
High
Event
Observed occurrences, behavioral patterns, incidents
Medium
Reported belief
Opinions, anecdotes, second-hand accounts
Lower

The investigation loop

One question at a time, ranked by impact

Questions are generated from exactly three sources — weak causal links, gaps in the causal map, and load-bearing assumptions. Nothing else. Each answer updates the model before the next question is selected.

Relationship validation

Validate a hypothesized causal link

“Does unclear ownership between product and platform directly slow decision-making, or are they happening in parallel?”

Targets relationships evidenced only by beliefs, or flagged as hypothesized. A yes/no answer directly updates the relationship's confidence and status.

Causal gap fill

Explain a cluster with no incoming causes

“What's the most likely driver of unclear ownership — is it role definition, historical context, or something about how decisions get escalated?”

Targets clusters that appear to be root causes but have no upstream explanation. Your answer becomes a new observation, routed into the model.

Assumption check

Validate a high-impact assumption

“The model assumes the team is aware of the ownership gap. Is that actually true, or is this invisible to most people involved?”

Targets assumptions whose validity would significantly reshape the causal map. Contradicting an assumption triggers a re-evaluation of dependent relationships.

“If I had an hour to solve a problem, I'd spend 55 minutes thinking about the problem and 5 minutes thinking about solutions.”

Attributed to Albert Einstein

Stop mistaking symptoms for causes

Phase 1 gives you the structure to understand a problem deeply before anyone has to agree on what to do about it. Start with one real problem, today.

Start your first inquiry →