Skip to content

issue-010.md

Latest commit

 

History

History
77 lines (47 loc) · 5.11 KB

issue-010.md

File metadata and controls

77 lines (47 loc) · 5.11 KB
issue_type Enhancement
linear_workflow Option A — parent issue + child issues T1–T4 (P0/P1); T5–T6 deferred to backlog until T3 validated
source_plan .cursor/plans/moneymirror_pm_ux_plan_08398100.plan.md (MoneyMirror PM UX Plan)
app apps/money-mirror

Issue Title

MoneyMirror Phase 3 — Unified multi-source dashboard, transaction-native insights, and expert AI coaching

Type: Enhancement

Problem (Current State → Desired Outcome)

Current state: MoneyMirror rehydrates dashboard data around a single-statement mental model; filters skew toward month/statement pickers rather than a global date range across all uploads. There is no transaction-level UI as ground truth. Insights skew bucket-level (“Wants is high”) rather than merchant-specific (e.g. Zomato ₹X over N transactions). Generative coaching is not yet anchored on server-computed facts JSON with strict schema + tone guardrails. Multi-account / bank + credit card uploads are not yet unified in one scope model with clear perceived vs actual behaviour under rollups.

Desired outcome: One unified dashboard for bank + credit card statements over time, with date + source scope, a transactions API + UI as source of truth, merchant-level rollups with Overview↔Insights coupling and deep links, and an AI track where rules + optional Gemini narrative only reference Layer A facts (no hallucinated amounts). P0–P2 phasing: T1→T4 first; T5–T6 after validation (per PM plan).

User

Primary: Gen Z Indians (₹20K–₹80K/month) using MoneyMirror on mobile, uploading multiple bank and/or card statements over time.

Secondary: PM/IC owner (you) tracking delivery via Linear parent + T1–T4 children without duplicate scope vs VIJ-25 (Sprint 4 backlog).

Why it Matters

Trust requires transaction truth before insights and AI can feel credible. Without unified scope, users cannot reason across accounts; without merchant evidence, coaching stays generic and North Star nudges (e.g. second upload) lack a concrete hook. Expert-style coaching needs deterministic numbers from the server, not free-form model invention.

Opportunity

Delivering Phase 3 positions MoneyMirror as multi-statement-native, filter-native, and evidence-backed — closer to a credible “finance guide” than a single-PDF demo. Linear stays PM-facing: one umbrella issue plus T1–T4 children you pull one at a time, aligned with /explore → /create-plan → /execute-plan when you start implementation (no code in this /create-issue run).

Hypothesis

If we ship transaction surface + unified scope + merchant rollups + facts-grounded AI coaching for MoneyMirror users with multiple statements, users will trust category and coaching views more, return to upload a second statement sooner, and reduce support-like confusion (“why doesn’t this match my PDF?”) because the UI always points to line-level evidence.

Risks / Open Questions (for /explore and /create-plan)

  • PDF / parser variance across issuers; merchant normalization v1 (heuristic vs Gemini post-pass — cost/latency).
  • Regulatory / product safety: generative copy must stay educational, not advice; docs/COACHING-TONE.md must extend with expert examples as generative surface grows.
  • Performance: cross-statement aggregation and pagination; rate limits if queries move beyond in-memory assumptions.
  • Perceived vs actual under all-sources rollup — explicit PRD decision (single blended number vs per-account) before UI lock-in.
  • VIJ-25 overlap: F3, G2–G3, H3 — merge or supersede into Phase 3 children to avoid duplicate tickets (stakeholder checklist in source plan).

Task map (repo contract — T1–T6)

ID Priority Theme
T1 P0 Transaction surface + API (foundation)
T2 P1 Unified scope model — single dashboard, multi-source
T3 P1 Merchant rollups + Overview↔Insights coupling
T4 P1 AI / coaching upgrade — expert guide (facts + structured Gemini)
T5 P2 IA + growth (URL-backed tabs, desktop share) — defer until T1–T3 validated
T6 P2 Compare months + hygiene — defer

Detailed acceptance themes live in the source plan; formal AC belongs in /create-plan output (experiments/plans/plan-010.md).

Next step

/explore done — see experiments/exploration/exploration-010.md (Build). Run /create-plan for PRD + architecture + AC per T1–T6. No /execute-plan until you explicitly start implementation.