Business-flow audit & SPI/Wipimo gap analysis
AI-generated content
This document was generated by an AI assistant. Verify accuracy before relying on the details.
An AI-agent audit (15 parallel workstreams, three waves — see audit/ORCHESTRATION.md) inventoried every ADB business flow against a frozen 109-item French gérance-locative taxonomy (T1–T18) and benchmarked it against two competitor suites, SPI (Septeo) and Wipimo (Orisha). The output lives entirely under audit/ at the repo root: twelve code-verification pages per module (audit/business-flows/WS-A1…WS-A12), two competitor grids and a taxonomy freeze (audit/benchmark/), and the synthesis (audit/gap-analysis/). This page is a pointer and summary — it does not duplicate the analysis; read the linked sources for anything load-bearing.
At a glance
| Fact | Value | Source |
|---|---|---|
| Taxonomy items scored | 109 (T1–T18), 100% coverage | taxonomy.md, coverage-matrix.md |
| P0 compliance gaps ([M] absent/partial) | 30 (matrix header still shows 31; corrected post-verification) | GAP_REPORT.md §3 |
| P1 competitive gaps ([S], competitor-confirmed) | 23 | GAP_REPORT.md §4 |
| Scope decisions (syndic, transaction, location saisonnière) | 19 matrix rows | GAP_REPORT.md §5 |
| Parity / ADB-ahead rows | 27 PARITY, 1 ADB-AHEAD (open API surface, T18.6) | coverage-matrix.md |
| Fully implemented [M] items | 5 (T3.1, T4.3, T4.4, T5.2, T12.1) | GAP_REPORT.md §1 |
| Verification method | Single-refuter adversarial pass (plan called for 3) | verification-log.md |
| Change-controlled modules affected | adb-accounting, adb-contracts (most P0 fixes land here) | root CLAUDE.md |
Details
What the audit found
ADB is a deep gérance-locative engine — an event-sourced 67-event double-entry ledger, a law-accurate rent/indexation lifecycle, ISO-20022 SEPA collection, and a broad open REST/OpenAPI surface neither competitor clearly matches — wrapped in an incomplete statutory-compliance shell. The four highest-stakes P0 clusters are: client-funds & accounting compliance (no bank segregation, FEC export, VAT posting, or entry attribution), lease-lifecycle statutory checks (décence, permis de louer, encadrement des loyers, deposit restitution), the statutory recovery chain (6-week clause résolutoire, CAF/guarantor notices), and charge régularisation. Most P0 fixes land in adb-accounting or adb-contracts — both change-controlled per root CLAUDE.md — so every critical-path fix needs engineer approval, and accounting changes additionally need an integration test.
Full detail, per-item legal anchors, and the suggested dependency-driven sequencing are in GAP_REPORT.md. The row-level evidence (status, evidence level E1–E5, code paths, SPI/Wipimo coverage) is in coverage-matrix.md. The twelve module-level code audits are under audit/business-flows/ (one per workstream, e.g. WS-A3-ledger-owner-accounting.md for accounting, WS-A1-lease-mandate-lifecycle.md for contracts).
Caveats carried over from the audit
- Verification depth. WS-C2 adversarial verification ran with a single refuter per claim; the audit plan specified three per P0 claim for a majority vote. Three rows were refuted and corrected (T3.6 état des lieux, T4.4 quittance de loyer, T13.2 PNO insurance — all moved from
absenttopartial/implemented), and T12.1 is flagged medium-confidence. Treat the P0/P1 counts as directionally right but not fully adversarially hardened. - Competitor evidence is marketing-sourced. SPI and Wipimo grids (
spi.md,wipimo.md) come from vendor pages and reviews, not spec sheets — many cells areunclear. A P1 classification means the taxonomy deems an item market-standard and at least one competitor confirms it, not that both were independently verified. - Legal anchors are not legal advice. Taxonomy anchors (
taxonomy.md) cite Légifrance/service-public sources but were compiled by an AI agent and should be reviewed by counsel before being treated as authoritative. - Scope decisions are not gaps. Syndic de copropriété (T9/T10), transaction (T11), and location saisonnière are presented in the gap report as strategic build/no-build questions, not missing features — see GAP_REPORT §5.
Where to go from here
- Read
GAP_REPORT.mdend to end for the human-facing conclusions and sequencing. - Read
coverage-matrix.mdfor any specific taxonomy item's evidence trail. - Cross-check any proposed
adb-accounting/adb-contractschange against the relevantaudit/business-flows/WS-A*.mdpage and the module'sMODULE_OWNERSHIP.mdentry before starting work.
Open questions
- No follow-up adversarial pass (3-refuter majority vote) has been scheduled; the audit plan's acceptance criterion for P0 rows is not yet fully met.
- No decision has been recorded yet on the three §5 scope questions (syndic, transaction, location saisonnière) — GAP_REPORT frames them but does not decide them.
- The audit branch (
claude/business-flow-audit-*perORCHESTRATION.md) has not been confirmed merged; verify which branch/commit this ingest reflects before citing specific evidence paths as current.