Commercial Memory Layer — Strategic Decision & Entry Plan

Internal decision & alignment document. Not a pitch.

Executive summary — the decision

The call: SERVICES-FIRST. AUSTRALIA-FIRST. Fund Stage 0 only — do not build the product yet.

This document does not present a plan to admire. It presents a chain of cheap tests. Each is a hypothesis: if it validates, we advance; if it fails, we exit or downgrade — chained until we either find the business or walk away from the market. The company earns the right to exist one test at a time.

What we are deciding today is not whether to build a company. It is whether to fund a 30-day Australian Recovery Diagnostic sprint that tests the first three hypotheses for real money.

What we are / what we are not

Whenever we say “commercial memory for mid-market contractors,” we mean this precisely — defined by what we refuse to be.

We are:

• A productised commercial-recovery service for Australian mid-market commercial / fit-out contractors and specialist subcontractors.
• The system of memory over the recovery loop — the place a commercial event is captured, carried into entitlement and quantum, turned into recovered cash, and remembered as cost.
• Sold to the commercial / QS / contracts function — the people who chase the money, not the project-management seat.
• Software underneath a service. A data-compounding business.

We are not:

• Not a field diary / daily-log app (Raken, Fieldwire).
• Not document AI / contract review (Document Crunch).
• Not takeoff / estimating (Kreo, Civils.ai).
• Not a PM system of record (Procore).
• Not just a claims consultancy — the software must compound, or we stop.
• Not “AI for construction” in general.
• Not a day-1 venture SaaS.

How we got here — the reasoning journey

This call is the output of a reasoning chain, not a hunch:

1. We mapped the field. Across ~60 competitors, a pattern: everyone reads documents (capture, docs, takeoff, QA, scheduling); almost nobody recovers money or pools cost.
2. We tested the wedge directly — and corrected ourselves. Our first claim, “no one recovers money,” was false. Recovery is contested (Magra and others quantify and recover). So the edge cannot be “we do claims”; it must be accountability plus proprietary outcome data.
3. We realised the product is the contract regime. Entitlement and quantum are jurisdiction-specific. So “first market” is not a branding choice — it is wherever evidence turns into cash fastest and most testably.
4. We de-biased geography. We scored four markets, and two independent reviews — an external analytical pass and our own geo research — converged on Australia and services-first.

(Full evidence inventory is in the appendix. This section is the why , not the what.)

How we’re choosing the market

We optimise for the fastest evidence-to-cash proof, not the biggest TAM. Four criteria:

1. Will buyers pay enough for recovery?
2. Does the contract / payment regime turn evidence into cash quickly?
3. Is the competitive field validating but not already closed?
4. Can we get messy records and commercial buyers without a platform dependency?

The product logic — the Commercial Event loop

Capture → Commercial Event → Entitlement → Quantum → Recovery → Cross-firm cost (which feeds the next tender).

Recoverable margin is lost when site records never become entitlement, quantum, and recovery. The atomic unit is the Commercial Event — an instruction, variation, delay, disruption, access denial, or design change: the moment money is won or lost. Owning the event (not the diary, the document, or the takeoff) is what makes us none of the things in “what we are not.”

The buyer moment

The pain is not missing records. It is records that never become money.

A fit-out contractor has 37 variations, an access-denial period, delay notices, scattered site records, and a disputed payment schedule. The commercial manager knows money is leaking but has no adjudication-grade event ledger. We reconstruct the Commercial Events, map entitlement, quantify recovery, and produce the recovery pack.

The mechanism: source-linked reconstruction plus human commercial review now turns messy records into an adjudication-grade event ledger faster than a QS / claims consultant doing it by hand — and defensibly.

The market gap

The field is crowded on capture and documents, thin on recovery, and empty on trusted cross-firm outcomes. Two areas of a 21-area market map stand out — Area 15 (change / claims / recovery) and Area 21 (cross-firm historical cost):

• Area 15 (recovery) is contested, not empty. Magra, ClaimDD/EOT, and Masin AI genuinely quantify and recover. The implication: the wedge is accountability + outcome data, not AI claims drafting.
• Area 21 (cross-firm cost) is empty by design. Pooling cost across firms is a trust problem — Rate QS markets “your private data, not cross-firm” as a feature. That is exactly why it is the durable moat for whoever solves the trust problem.

The competitor landscape

Three nearest players each validate one piece of the loop; none owns it:

• Magra (US) — the money / quantum half.
• Gather Insights (UK) — the front half: NEC/JCT compensation-event detection and notices. Stops before quantum.
• BuildPass (AU) — capture plus agentic AI. Stops before claims.

The wider field: crowded (Procore, Raken, Fieldwire, PlanRadar, OpenSpace, BuildPass); recovery pieces (Magra, Gather, ClaimLogic, SmartPM, nPlan, Nodes & Links); cost memory (Rate QS, Gauge, BenchIt — all single-firm). Full dossier-grounded table in the appendix.

The platform lesson: do not be an integration parasite. Pype→Autodesk, Payapps→Autodesk, Document Crunch→Trimble, and Trunk Tools’ cut-off from the Procore API all say the same thing. Use platforms as inputs; own the recovery decision record and the outcome data.

Decision 1 — Australia first

The US is the prize. Australia is the proof market.

Weighted beachhead ranking (30% willingness-to-pay · 30% recovery regime · 20% competition · 20% cold-start):

How we concluded this: the product is the contract regime → which regime turns evidence into cash fastest? → Australia’s Security of Payment statutory adjudication: short-cycle, evidence-heavy, with half the recovery loop legally mandated. That makes AU the most falsifiable first market — the place a recovery business can be proven or disproven quickest. Our independent geo scan reached the same answer.

US-first is the vanity move (bigger ACV, denser platforms, Magra closer, no clean QS buyer). AU-first is the falsifiability move.

The economics

Australia supports a strong specialist business. The venture case only appears if the proof travels.

• SAM 1,500–2,300 firms · ACV A$35–80k (recovery audit) · first-market revenue ceiling A$15–30m.
• Scenarios: conservative niche ~A$3.2m · 5-year credible A$6–11.7m · strong A$16.5–30m.

This is not a day-1 venture SaaS raise. It is a real specialist company. The venture branch is a later hypothesis (H6/H7): it appears only if outcomes pool into a cross-firm moat and the motion transfers to other markets.

Decision 2 — Services-first

Services-first is not caution. It is how we earn trust, records, and outcomes.

How we concluded this: entitlement and quantum are adversarial — they touch cash, relationships, adjudication risk, and professional liability. Buyers will not trust automated output without source-linked evidence and human review, so a self-serve SaaS launch fails. And the moat (outcome data) does not exist on day one — only services generate it. Therefore: service first, software underneath, SaaS only after repeatability.

The shape: a buyer-facing AU Recovery Diagnostic, sitting on human commercial judgement and a legal/professional review boundary, sitting on reusable software assets, sitting on messy source records.

The discipline: automating adversarial entitlement before trust is naive; doing services without extracting reusable assets is just consulting. We do neither.

The first paid offer — AU Recovery Diagnostic

The first SKU is a diagnostic, not a SaaS login.

• Price test: A$5k–A$15k · Timeline: 1–2 weeks
• Buyer: Commercial / Contracts Manager, QS, or subcontractor owner
• Trigger: payment dispute, final-account leakage, variation backlog, weak EOT / loss-and-expense pack
• Inputs: diaries, emails, WhatsApp exports, photos, payment claims, payment schedules, programme, CVR / cost ledgers
• Outputs: Commercial Event ledger, entitlement map, quantum schedule, evidence pack, recovery priority memo
• Pass signal: 3 paid pilots / LOIs in 30 days

The anti-consultancy discipline

Every diagnostic must produce reusable software assets, or this becomes consulting. Each engagement creates: Commercial Event schema rows, evidence-extraction patterns, entitlement clause mappings, quantum templates, outcome labels, and benchmark candidates.

Kill-criterion: if after 10 diagnostics more than 60% of the work is still bespoke expert labour → stop, or reposition as a claims-services firm.

The four bets, sequenced

Recovery goes first. Capture and tender are inputs. Benchmarking is the prize, not the wedge.

• NOW — Claims / entitlement recovery — the revenue engine.
• NEXT — Tender baseline — the context layer for later event variance.
• LATER — Field evidence capture — the data-completion layer (integrations + thin UI).
• LAST — Cross-firm cost benchmark — the moat.

How we concluded the order: capture-first becomes another low-ACV workflow tool; benchmark-first dies in a trust cold-start; recovery is the only wedge that pays and creates outcome data.

The moat & trust architecture

The moat is permissioned realized-outcome data — not modelling.

The trust path: raw records (client-owned) → structured Commercial Events → outcome-labelled recovery records → anonymised / aggregated / thresholded → cross-firm benchmarks (useful only after density and governance) → tender and recovery intelligence.

A precise caveat: Australia’s public Security-of-Payment adjudication data seeds the event taxonomy and outcome logic (claim failure modes, evidence patterns, entitlement reasoning). It does not replace realized private cost data — rates, productivity loss, prelims, margins, settlement behaviour. That is exactly why services-first matters: services are how we earn the private data the moat requires.

The strategy as a hypothesis chain

We do not “decide to build.” We run cheap tests in order. Each either advances us or exits us.

Terminal states: EXIT MARKET (H1–H3 fail) · AU SPECIALIST BUSINESS (H4–H5 yes, H6–H7 no) · VENTURE-SCALE (all yes).

The staged, gated plan

Each stage is funded only by the hypothesis the last one validated.

• Stage 0 — 30-day beachhead validation (tests H1–H3)
• Stage 1 — Productised Recovery Diagnostic (months 1–3; tests H4)
• Stage 2 — Commercial Memory / tender baseline setup (3–6)
• Stage 3 — Live Commercial-Event capture, thin UI + integrations (6–12; tests H5)
• Stage 4 — Recovery Workbench (9–18)
• Stage 5 — Private → cross-firm benchmark (18+; tests H6) → expansion (H7)

Every gate is two-branch: validate → fund the next stage; fail → exit, or hold at the current terminal state.

Stage 0 — the 30-day validation contract

The next decision is not company funding. It is whether three buyers pay and hand over records.

The honest risk: our voice-of-user evidence is biased toward field/admin users — it proves adoption, not QS/commercial willingness-to-pay for recovery. Stage 0 closes exactly that gap. The five tests (H1–H3 made concrete):

1. 3 paid pilots / LOIs.
2. 5 firms share real redacted project records.
3. NSW / VIC / QLD counsel clears the service and success-fee model.
4. 3 historical back-tests improve recovery logic.
5. A commercial director says the output would change a recovery / payment decision.

The decision

Approve Stage 0 under hard gates — otherwise stop.

Stage 0 funded → paid demand (H1)? if no, stop → data access (H2)? if no, stop or advisory-only → legal route clean (H3)? if no, redesign or stop → back-tests improve recovery? if no, stop → all yes → Stage 1.

No SaaS build until a buyer pays for recovery and hands over records. Fund Stage 0. If the gates fail, stop.

Appendix

1. Research corpus inventory — ~60 competitors, 12 full dossiers, 7 category scans, 4 geographies, 2 independent strategic reviews that converged.
2. Dossier-grounded competitor table — the full 19-tool table (coverage, talk-vs-ship gap, voice-of-user, threat/partner/absorb, per-geo presence) in the strategy memo.
3. Geo sizing math — AU SAM / SOM / ACV assumptions and the expansion sequence (AU → UK → Canada/Ontario → US).
4. Commercial Event schema — detailed fields by AU / UK / Canada / US regime.
5. Legal / professional boundary notes — SoPA, claims-advice limits, success-fee constraints, adjudication-support boundary.
6. Hypothesis register — H1–H7 with explicit pass/fail thresholds and the action on each branch.