06 execercise

Build prompt: M&A Integration Control Tower

Copy everything below this line into Claude with the data folder attached.

You are my product builder and integration-management analyst. I am Adrian Vale, SVP of Strategic Initiatives at Aurex. Aurex acquired four capability businesses in 2025—Northwind (now Aurex Prime), Treasura, Tideway, and Vaultline—and must integrate products, customers, systems, risk controls, and teams while capturing the strategic rationale of each deal.

Build an executive-ready M&A Integration Control Tower from the synthetic files in data/. The tool should help an executive find exceptions and unblock decisions; it is not a generic project-management board.

Working method

Inspect all source files first. Report file purpose, row counts, date range, missing values, and join keys. State a short plan before implementation. Do not silently fill gaps or convert an unknown into a green status. Use the source files as the only source of truth.

Inputs

• data/integration_workstreams.csv: one row per acquisition/workstream, including schedule, progress, status, dependency, blocker, decision, and synergy fields.
• data/integration_milestones.csv: detailed milestones and acceptance criteria.
• data/decision_log.md: synthetic notes for decisions already made or still open.

Display “SYNTHETIC DATA — FOR DEMO ONLY” prominently in the product and generated brief.

Definitions and calculations

Use 2026-03-06 as the as-of date.

• A workstream is overdue if target_date is before the as-of date and percent_complete < 100.
• A milestone is overdue if its due_date is before the as-of date and its status is not Complete.
• Planned progress: linearly interpolate from 0% on baseline_date to 100% on target_date, capped between 0% and 100%.
• Schedule variance equals percent_complete - planned_progress. Show percentage points, not percent.
• A synergy target is counted once per acquisition + synergy type. Because it repeats across workstream rows, deduplicate it before summing. Sum realized value the same way using the maximum value per acquisition + synergy type, and document this rule.
• Forecast synergy confidence:
  • High: realized at least 80% of target, or all related workstreams are Green;
  • Medium: realized 40–79% of target and no related workstream is Red;
  • Low: any related workstream is Red, or realized below 40% when the target is nonzero.
• An item requires executive attention if any of these are true: Red status; overdue; schedule variance at or below -15 points; an open decision is present; blocker has remained open more than 14 days based on blocker_opened_date; or the next milestone is due within 14 days and the item is below 80% complete.

Product requirements

Create a self-contained local web app with as few dependencies as practical. If a local server is unavailable, provide a single-file HTML app that opens directly.

Include:

• Portfolio KPI cards: overall completion, Red/Amber/Green counts, overdue milestones, open executive decisions, FY26 synergy target, synergy realized, and synergy confidence.
• One acquisition summary card each for Aurex Prime, Treasura, Tideway, and Vaultline, with completion, schedule variance, RAG mix, top blocker, next milestone, and synergy progress.
• A cross-deal heat map with acquisitions as columns and workstreams as rows. Every colored cell must also carry a text/icon label.
• A milestone timeline with overdue and upcoming-30-day views.
• A deduplicated synergy view by acquisition and synergy type, showing target, realized, gap, and confidence.
• An executive attention queue, ordered by severity and due date, with the source field(s) that triggered attention.
• A decision center separating open, decided, and superseded decisions. Link workstream decisions to the corresponding entry in the decision log when IDs match; flag unmatched IDs.
• Filters for acquisition, workstream, executive sponsor, RAG, owner, status, and “attention required.”
• A data-quality panel for duplicate keys, impossible dates, percentages outside 0–100, missing owners, unmatched decision IDs, and contradictory milestone/workstream states.

The design should feel like a serious operating review: compact, calm, legible, and optimized for scanning. Include metric definitions in an “About” panel. Do not create a false portfolio average by averaging acquisition percentages; calculate the overall figure as the average of the individual workstream completion values and label it clearly.

Deliverables

Create:

1. the working app;
2. README.md with exact run/open instructions, expected input schemas, calculation rules, and steps to refresh the dashboard weekly;
3. WEEKLY_EXECUTIVE_MEMO.md, maximum 600 words, containing overall status, changes that matter, synergy outlook, top five exceptions, decisions required (with owner and date), and next-two-week milestones;
4. VALIDATION.md with row counts, join results, synergy-deduplication checks, schedule-variance spot checks, and interaction tests.

Guardrails and quality bar

• Treat RAG values and notes as reported management data, not objective truth.
• Never infer legal, regulatory, accounting, security, or personnel approval.
• Clearly separate a source-data fact, a deterministic calculation, and an AI recommendation.
• Test every filter, link, sort, modal/drawer, reset state, and empty state. Confirm all displayed totals come from the input files.
• If something cannot be completed, disclose it in the README with a concrete next step.

At the end, tell me what you built, how to open it, which three synthetic issues most need Adrian's attention, and what live exports should replace these files first.